We connect two sides of the video game industry that are mutually beneficial: game developers and content creators

NOIZ

noiz

PARIS, France, April 10th, 2024/Chainwire/-- Sui Basecamp, the inaugural global conference for the Sui ecosystem, opened its doors today in Paris, welcoming developers and entrepreneurs from around the world as they celebrate Paris Blockchain Week. A two-day event across two venues, Sui Basecamp features speakers across web2 and web3, spanning industry verticals from payments to gaming to e-commerce to major league sports. A celebration and exploration of the latest developments in web3, attendees are invited to presentations on cryptography from some of the world’s leading technologists, fireside chats on increasing market liquidity, and panels on real-world technological use cases. Highlight announcements from the conference included: Introducing the SuiPlay0x1 – a lightweight handheld gaming device designed in collaboration with Playtron. SuiPlay0x1 runs Playtron’s device-agnostic gaming operating system, which is building up compatibility with various hardware configurations, gaming storefronts and direct-download game partners. The SuiPlay is the first handheld gaming device with native web3 capabilities and will be available in stores worldwide in 2025. First Digital Labs is launching FDUSD, the fastest-growing stablecoin in crypto, on Sui, becoming the ecosystem’s first native stablecoin. Originally deployed on Ethereum and BNB, with plans to expand to other blockchains, Sui becomes the first blockchain FDUSD has expanded to since its inception. FDUSD launched in August 2023 and has already amassed a market cap of over $3.5 billion. Over the last week, FDUSD has seen the fifth-highest trading volume across the entire industry, including over $8 billion of volume in a single day. Enoki – Mysten Labs, the original contributor to Sui, has announced the launch of the Enoki platform, the gateway to next-gen customer experiences. Inside the Enoki portal, enterprises can access, leverage, and embed public ledger solutions inside their applications, products, and services, thanks to SDKs powered by Sui’s native feature, zkLogin. Partners integrating Enoki at launch include: Bluefin: A perpetual DEX with blazing-fast settlement and a seamless trading experience. Drife: A decentralized ride-hailing platform powered by blockchain with the intent of empowering both drivers and commuters. Quantum Temple: The digital platform for immersive cultural travel. The event is a two-day affair that begins at 9:30 AM local time on Wednesday, April 10, and concludes at 4:15 PM CET on Thursday, April 11. On Wednesday night, registered guests are invited to attend the Sui Soiree, an evening of cocktails, refreshments, and dancing with a performance by German electronic music record producer, songwriter, and DJ, Boys Noize. For more information, please visit sui.io/basecamp. About Sui Sui is a first-of-its-kind Layer 1 blockchain and smart contract platform designed from the bottom up to make digital asset ownership fast, private, secure, and accessible to everyone. Its object-centric model, based on the Move programming language, enables parallel execution, sub-second finality, and rich on-chain assets. With horizontally scalable processing and storage, Sui supports a wide range of applications with unrivaled speed at low cost. Sui is a step-function advancement in blockchain and a platform on which creators and developers can build amazing, user-friendly experiences. Learn more: https://sui.io About Mysten Labs Mysten Labs is a team of leading distributed systems, programming languages, and cryptography experts whose founders were senior executives and lead architects of pioneering blockchain projects. The mission of Mysten Labs is to create foundational infrastructure for web3. Learn more: https://mystenlabs.com About First Digital Labs First Digital Labs is the brand name of FD121 Limited, a Hong Kong-registered subsidiary under the First Digital Group. First Digital Labs focuses on cutting-edge research and development, specializing in the innovation and advancement of digital assets. First Digital Labs is the issuer of the FDUSD stablecoin. To learn more about First Digital Lab, visit https://firstdigitallabs.com/. About Playtron Playtron is building a lightweight gaming OS optimized for a new generation of powerful handheld gaming PCs and beyond. Play all your games from every store: Steam, Epic, GOG, and more. Playtron's founders built Android for 1B users while simultaneously overseeing a multi-hundred-person operational system team in an attempt to unseat the incumbent mobile duopoly. Playtron is backed by Samsung Next, Polychain, Circle, Mysten Labs, Alumni Ventures and more. Playtron-powered devices will start to ship worldwide later this year, ready to onboard the next 100 million core and casual gamers. Contact Sui Foundation media@sui.io This story was distributed as a release by Chainwire under HackerNoon’s Business Blogging Program. Learn more about the program . here

Gaming, Stablecoins And Product Innovation Take The Stage At Sui Basecamp

গেমিং, স্টেবলকয়েন এবং প্রোডাক্ট ইনোভেশন সুই বেসক্যাম্পে স্টেজ নেয়

Gaming, Stablecoins und Produktinnovationen stehen im Mittelpunkt des Sui Basecamps

Los juegos, las monedas estables y la innovación de productos suben al escenario en Sui Basecamp

Les jeux, les pièces stables et l'innovation produit entrent en scène au Sui Basecamp

गेमिंग, स्टेबलकॉइन और उत्पाद नवाचार सुई बेसकैंप में मंच पर हैं

ゲーム、ステーブルコイン、製品イノベーションがSui Basecampで登場

게임, 스테이블코인 및 제품 혁신이 Sui Basecamp에서 무대에 오릅니다.

Jogos, stablecoins e inovação de produtos sobem ao palco no Sui Basecamp

Игры, стейблкоины и инновации в продуктах выходят на сцену Sui Basecamp

Oyun, Stablecoinler ve Ürün İnovasyonu Sui Basecamp'ta Sahneye Çıkıyor

Trò chơi, Stablecoin và đổi mới sản phẩm chiếm ưu thế tại Sui Basecamp

游戏、稳定币和产品创新在 Sui Basecamp 上亮相

In Facebook’s recent earnings call with investors and analysts, CEO Mark Zuckerberg repeatedly mentioned the word “metaverse.” He announced his expectation for Facebook to transition into a metaverse company within the upcoming years. This is not the first time the Facebook CEO has expressed his ambitious new venture. In June, Zuckerberg announced to his employees that Facebook would go further than just building interconnected social apps. Zuckerberg’s future of the internet would be a maximalist version of Facebook, converging the worlds of socializing, work, commerce, entertainment, and more. Putting aside the number of times Zuckerberg has raised public concern, from Facebook’s data and privacy breaches to allegations that Facebook is contributing to vaccine misinformation, there is good reason for him to push Facebook towards becoming the biggest player in a new technological era - developing the metaverse. What is the Metaverse? First coined by science fiction author Neal Stephenson in his dystopian novel “Snow Crash,” the metaverse is described as a virtual shared space for people to interact using avatars. It is a virtual reality experience that, nearly three decades later, becomes an entity that every major tech company wants to commercialize. Microsoft’s CEO Satya Nadella described the metaverse “as the virtual and physical worlds converge” - a platform for simulated environments and mixed reality. Instead of attending a Zoom meeting, you can attend your meeting in your avatar form in a simulated office space, surrounded by colleagues from anywhere in the world. When the work meeting is over, you can choose to attend a virtual concert with your friends or shop for items that are only available digitally in virtual stores. Everything is interconnected on the metaverse. There is no need to hop onto a different website or application. Matthew Ball, a venture capitalist and expert on the metaverse, said this of the metaverse, “More of our lives will be online, rather than just data relating to our lives.” Of course, when so many aspects of our lives can be lived online, this naturally translates to huge economic opportunities amounting to billions. Digital advertising is expected to hit $455 billion. The market for non-fungible tokens (NFT) - digital assets that can be exchanged for real-world cash, is rising in popularity and amounts to $2.5 billion in the first quarter of 2021. Together with other fields such as gaming, the metaverse is easily worth more than a trillion. The Metaverse Is Already Manifesting While big tech companies like Facebook and Microsoft are trying to incorporate new technology to form this futuristic utopia, other companies are already ahead of the game in creating metaverse experiences, if not developing advanced software and hardware that can accommodate these experiences. Discord has grown from a social media platform for gamers to a platform for its 150 million users to converge in topic-based spaces called servers. Users can use it for various ways of interaction - chat, send and receive files, watch live streams and listen to music together. Here, we can see an early version of the metaverse, especially in the socialization aspect. In terms of software, companies like Epic Games and Nvidia Corp. are leading with technology that can run virtual world simulations allowing large multiplayer experiences. Video gamers on Fortnite, Roblox, or Animal Crossing are already experiencing metaverse experiences with the ability to create their virtual worlds. Another essential building block for the metaverse is blockchain technology. It allows a seamless, secure process when transporting assets between platforms or from the real world to the digital world. Real-world assets can be transformed into NFTs, which can be easily purchased using cryptocurrency, increasing the liquidity of these assets. Hence companies incorporating blockchain are better positioned in ushering the metaverse than traditional big names in tech that have yet to embrace blockchain. Play-to-Earn Games Are Accelerating The Revolution While the metaverse in its full form will involve cutting-edge VR and AR technology, Internet users need to see the appeal in joining platforms that will eventually form the metaverse. The pandemic has accelerated the process of making people more comfortable in spending more aspects of their lives online, but the catalyst for mainstream adoption lies in play-to-earn games. For starters, gaming has been touted as a great use case for encouraging mass adoption of blockchain technology. Blockchain is already disrupting the $176 billion industry, allowing players to enjoy more immersive experiences and establishing new norms for in-game currency and trading gaming assets. Back then, people could spend a lot of money on in-game assets but never truly own them or make money out of them. With blockchain allowing true ownership of digital assets, play-to-earn games have seen a meteoric rise in popularity. Players can earn in-game assets such as native tokens or NFTs through game participation, which can be cashed out on open marketplaces. With the pandemic causing financial difficulties for many, earning an income through play-to-earn games from the comfort of their own homes instead of working day jobs have become a reality for many players in developing countries. For example, people in the Phillippines and Vietnam have quit their jobs to focus on trading NFT pets and native tokens on the NFT pet breeding game Axie Infinity, which saw daily users reaching 800,000 in July. More people become attracted to play-to-earn games and their open economies, bringing value to games while generating personal revenue. As a result, gaming companies receive investments from various institutions and investors, which drives the current bullish market. Gaming companies are also evolving in their competition to offer more social and immersive experiences apart from games. Decentral Games - Rising Star In The Metaverse Take blockchain entertainment company Decentral Games ( ), for example. The Ethereum-based brand started as the first community-owned metaverse gaming platform and has expanded its repertoire to involve hosting regular music and art events and a thriving NFT market demand. Within a few months from its official launch, DG has secured partnerships with some of the biggest names in the entertainment industry, such as and . DG video game pioneer Atari iconic nightclub Amnesia Ibiza How did DG achieve this success?  Unlike traditional gaming models where players and the house have what DG founder Miles Anthony called an “adversarial relationship,” DG users enjoy games through a provably fair, play-to-earn model where incentives are aligned between player and house. Notably, DG is the first DAO that owns, governs, and monetizes virtual land. DG players who own DG’s native token are essentially part of DG governance, allowing them to participate in DG’s development-related decisions. Combined with the 3D environment to play in, DG provides a whole new, unparalleled gaming experience where users are empowered to “be the house” while enjoying immersive, social experiences. Users can also find employment in this virtual space starting early this year, marking DG as the first metaverse employer. Within this month alone, DG has announced the launch of - BAYC, DexTools, and Aquarium, proving its appeal among brands in building metaverse venues. Notably, DG has also developed virtual headquarters for prominent crypto projects like Polygon, Coingecko, Rarible in the metaverse. three new B2B metaverse gaming venues in partnership with popular brands DG has started tapping into the wider entertainment scene, curating unique music and arts events for its users. Thousands of users turned up for electronic artist Dillon Francis’ live set during the . On the same day, DG also hosted the much-anticipated auction of Dillon Francis’ Gerald, the pinata-themed NFTs, designed by visual artist Merda. Following the success of this event, DG has gone on to host more successful events featuring prolific names in the dance music industry, such as German producer Boys Noize and Chicago-based trio Autograf. grand opening of Atari Casino powered by DG DG utilizes sophisticated virtual infrastructure on the backend combining Ethereum, Polygon (Matic), and the Binance Smart Chain networks, ensuring scalability concerns are addressed to accommodate more users. Backed by a team of experts and strategic partnerships, DG has specialized technological expertise that allows users to benefit from faster transaction speeds and lower transaction costs. Likely even before Facebook realizes its metaverse dream, DG’s innovative contributions across several verticals make it a promising player in this race to develop the metaverse. References: https://www.cnbc.com/2021/07/26/facebook-creates-exec-team-to-work-on-metaverse.html https://www.theverge.com/22588022/mark-zuckerberg-facebook-ceo-metaverse-interview https://www.technologyreview.com/2021/03/11/1020600/facebook-responsible-ai-misinformation/ https://www.crn.com/slide-shows/cloud/ceo-satya-nadella-s-12-boldest-remarks-at-microsoft-build-2021/11 https://cointelegraph.com/news/the-metaverse-mark-zuckerberg-s-brave-new-world https://www.emarketer.com/content/worldwide-digital-ad-spending-2021 https://www.reuters.com/technology/nft-sales-volume-surges-25-bln-2021-first-half-2021-07-05/ https://www.nasdaq.com/articles/best-stocks-etf-for-the-trillion-dollar-metaverse-2021-08-16 https://newzoo.com/insights/articles/global-games-market-to-generate-175-8-billion-in-2021-despite-a-slight-decline-the-market-is-on-track-to-surpass-200-billion-in-2023/ https://kr-asia.com/what-does-axie-infinitys-meteoric-rise-tell-us-about-the-play-to-earn-game-industry https://decentral.games/blog/decentral-games-launches-three-new-partnered-venues-alongside-bored-ape-yacht-club-dextools-and-vega Disclaimer: The author of this article owns $DG tokens. This article is part of The hosted by HackerNoon in partnership with Gaming Metaverse Writing Contest . The Sandbox Submit your #gaming-metaverse story today for your chance to win up to $2000 in SAND tokens.

Manifesting The Metaverse With Play-to-Earn Games

Pirates. Bounties and bounty hunters. Nope, not your new Netflix/Hulu series but trendy blockchain/bitcoin topic. I’ve been swimming in murky waters of bounty programs for ICOs last couple of weeks. While navigation is often times hard if you navigate properly you can come across shiny genuinely amazing programs which in all reality can earn you a chunk rather than a ton of money. However, as any experienced bounty hunter will tell you, there is a lot of garbage out there. And navigating through it can be an overwhelming task. But, what are bounties? Essentially, bounty is a marketing strategy used by blockchain startups interested in stirring up communities, getting the word out about them so to speak, by investing a part of their ‘ CryptoAssets’ into communities in return for spreading the word. In simple terms, blockchain startups offer users a fair compensation for promoting and/or improving their ICO product. Users are required to perform different tasks such as social media posting/sharing etc for which they get either ICOs native token, other cryptocurrencies, or even fiat currency. Huh, what? Let’s give just a few examples shall we? TeslaWatt Bounty Program TeslaWatt main goal is to build a solar farm to provide our new and existing customers access to independent energy. For the purpose of realizing their plans, they’ve built an infrastructure for a utility token TSW that will make this happen. TeslaWatt token #TSW is an Ethereum-based token of value. As a utility token, TSW will enable holders to continuously use TeslaWatt’s solutions, as well as to trade the utilities. A SOCIAL MEDIA BOUNTY CAMPAIGN (on Twitter, Telegram, BitcoinTalk forum, Facebook and LinkedIn) is live, with the total allocation of their 250.000 TSW tokens! Find out more at the official . BitcoinTalk page here BlockScart Bounty Program BlockScart is set to be a global e-commerce marketplace powered by blockchain technology. They operate with BLXS coin and participants will get their hands on BLXS which is a Waves Sponsored Asset at a rate traded via the DEO (decentralized exchange offering)at $0.05. They are running their campaign on Telegram, Facebook, Twitter, Instagram and Reddit. Find out more at the official . BitcoinTalk page here While it is one of the best ways for blockchain startups get the word out, for those participating in the program how can you be certain this isn't a scam? While I don’t have all the answers in my experience these things work in defining the good ones from the rest:- First and foremost study the rules and the length they’ve taken to explain and promote purpose of their bounty program.- Do a research before you jump on their bandwagon. Find out everything you can that is official about the company behind bounty (headquarters address, check Google map for it, Google my Business listing, official press releases and articles etc).- Dig a bit deeper about the company and people running it and I’m not talking about their official website and soc profiles. Check out the social noizz, forum talk etc. See if there is a into what they are trying to promote.- Use their Telegram channel to get familiar with them. Ask the dumbest question you can think off and them go into tech specifics.- Check out what is the main purpose of their token. Ask around if their explanation makes sense.- you either follow or have in your community.- Check out who shared/re-twitted/posted etc. You’ll be surprised how many fake profiles do that (usually low in number of followers/friends). Once you find real profiles send them a message. Ask what they did and why. consistency of interest ASK EXPERTS Having done all that won’t make you a successful bounty hunter in an instant. But it’ll give you a nice beginners framework which can help you become one. BTW, a well-versed and competent bounty hunters are in demand right now. Seize the moment and happy hunting. Read on…the links :-) There are people who could explain things far better than me and whose posts helped me a lot. While I share experience they share the knowledge. Tokenomics: A Business Guide to Token Usage, Utility and Value The ultimate guide to ICO bounty campaigns by Pascal Thellmann Utility vs. Security Tokens: Why Not Both? How To Make Money From Crypto Bounty Programs The Complete Guide by Mindmaster If you enjoyed this story, please click the 👏 button, follow me at , and and share it to help others find it! Medium Twitter Facebook Have I missed something? Let me know. Share your thoughts in the comment section below.

Colorful World Of Bounty Programs

In this story I will go in depth and try to give you an all-around picture how we prepared for non-paid efforts in order to generate traction, exposure, grow the community and ride that wave for our SaaS tool Reportz.io Before we start… is a clean and simple KPI reporting that helps agencies, freelancers, entrepreneurs and all other save time, automate client reporting and slay mortal enemies! Reportz.io tool badasses We created it for our in-house needs since at time and all the tools we tried (and we tried them all) were not serving our needs as a full stack agency we are or were just shamelessly upselling every click on their platform — which we really didn’t like… we were serving more than 300 clients on a monthly basis Bootstrapping. Reportz is just one of our tools, and every one of them is 100% bootstrapped. We never wanted somebody to sit on our head telling us what to do, and since everything is paid from our pocket — our main strategy is to: “USE EVERY CHANNEL OF ACQUISITION WE CAN” …before jumping into burning that PPC budget. You know how it goes… We also have that situation that we are competing with some of the biggest and well-funded companies, (the main difference between us and Klipfolio is that we didn’t want that 12 million seed funding investment 😛) so we really had in order to stand a chance which you will get the opportunity to find out later in the text since I really covered it all. to focus on doing “that something” better than everyone else Pre-Preparation. The first campaign we did was on Product Hunt. After our Product Hunt launch, we acquired an initial 531 users. The campaign was executed really well, with a little help from our friend Nichole Elizabeth De Mere (you know her from GrowthHackers and Zest.is), we ended as 🏅, but in the end, less than 100 people out of those 531 became paid users, because our initial onboarding + nurturing efforts were a disaster. 💩 3 Most Upvoted Product of the day Product Hunt helped us see the issues, collect feedback and modify our onboarding. It also helped me realize that we have to before jumping into any acquisition effort create a clear vision of where we are and what are the necessities in order to reach our goals… Our onboarding was shit and we had to improve it Mapping customers journey and evaluating the pain points Creating ICP (Ideal Customer Persona) and segmenting all the activities regarding unique segments Realizing that customer success will be a crucial part of our conversion % raise since nature/functionality/complexity of the tool. For a couple of months, we were , building integrations, building relationships with partners and working on the development preparing the educational material. We wanted to , and that meant as much as possible while making useful content and shitload of other educational material in order to help our users get on-board. avoid ending in SaaS Graveyard by any means optimizing the onboarding and reducing the general complexity You don’t want to end up there… Preparation. Industry issues. As mentioned, we had to figure out that will make us stand out from the fierce competition. what is the single thing we do better Much like Shakira's hips, data doesn’t lie. , we realized that there are 5 main issues people are not happy with, regarding reporting tools: After surveying and talking with our users Complexity Prices Upsells Educational material Customer success & support How we solved that? We didn’t create predefined metrics, and instead just integrated RAW metrics everyone could recognize Our pricing model is so a user is just paying the “ ” pay per use amount of dashboard he uses on a monthly level — whether you are a freelancer or enterprise, which perfectly aligns with our pricing model We have NO upsells Everything is included, We created a strong base of educational material in various formats + we have a “Demo with Founder” option which is free for all users Every person from CEO to support manager has slots on Intercom, so if there are users that are “too much” for rookies, somebody experienced can jump in Mapping channels of promotion Traditionally, we mapped channels as: Owned Blog Newsletter Earned Facebook groups Influencers Platforms LinkedIn Quora Reddit YouTube Guest Posts Publishes Partnerships Paid Facebook Ads AdWords BingAds Quora Ads PR articles As mentioned, our main goal is to make the most out of 1. and 2. and to play smart. To make all hassle worth it, we needed a and non-paid efforts, something that will help us generate that will help us foundation for our future campaigns advocacy, receive constant feedback and make the necessary noize get the attention of big shots. It came to my mind just like that. Since I am the creator of , we knew exactly that owning an engaged community of raving fans will help us a lot. the biggest e-commerce community in the Balkan I also know that the best way to create a community is to: Make it an interesting place Make it a valuable place Find a big source of traffic that you will redirect to it Usually, we would do a as always… quick influencer campaign … but this time we needed something different. P.S. We wrote a complete “How To Article” about building a community on Reportz blog , check it out if you want to know more. AppSumo Since said “Fuck you Facebook”, and went to pursue his young-boy-dreams, AppSumo was one of the most profitable sectors of Sumo Inc. Noah Khagan Fun fact: During our AppSumo preparation, I found a case study with some interesting results I am sure you are familiar with AppSumo, but in case you don’t know, it basically is a Groupon for entrepreneurs, offering crazy discounts on various SaaS tools for their audience… Now, personally, I am not a fan of discount campaigns, but is not a bad start, I have to admit. getting in front of 1M+ entrepreneurs worldwide Introductions were made, and I ended up on a 45-minute call with Olman, Head of Business Development of the AppSummo. Everything was arranged fast since the preparations were done way before the call happened. I do. We went live on May 7th. Note: I created a dedicated post regarding our AppSumo launch, so I am not going to jump into details about AppSumo here, but… Before we go any way further you should know a thing or two about… Running a deal on AppSumo: AppSumo takes from 60%-70% of the profit They run ads, hire influencers and affiliates Most common price per coupon is $49 for LTD (lifetime deal) Average sales per campaign vary from 2.5k — 3.5k of coupons Average refund rate varies from 12–16% No question refund period lasts for 60 days You can offer a “stackable deal” so that one user can buy more of them Most of the audience are freelancers and small size teams but there are some exceptions Also, be sure to: Hire and train support people since it will get crazy Check your server status since it will crash if you don’t have enough capacity Prepare ALL the help articles, materials and handouts regarding your tool Train your support team and arm them with patience and empathy Monitor all LTD communities and engage with users, reply to questions and provide them with help Create a (scratch) public roadmap since a lot of guys will ask for it Learn about that specific audience, know their pain points and value P.S. We helped quite a few of guys land on the AppSumo after our campaign and skyrocket their product. If you want me to help you regarding AppSumo, reach me out HERE . Our stats: Coupons sold AFTER the refund rate: 5,304 Refund rate - 11.69% And with we ended as the at the moment. Reportz.io best campaign for 2018 They are talking about us… TRUE VALUE If you think that 💸 is the reason you should go to the AppSumo, you should scroll a bit up and read my AppSumo part once more… The true value of these campaigns comes from: Getting enormous exposure and credibility The chance to create a community of early adopters, advocates and affiliates that will help you execute other campaigns The chance to get testimonials and feedback about your product, integrations and feature updates …and generally ridding that wave in order to reach your higher, long-term goals. 🌊 Riding the wave 1. Getting enormous exposure and credibility ( Stats shown through Reportz.io) Creating that momentum is essential for your outreach, partnership, affiliate and influencer efforts… From May 1st to June 10th, we got… Of course I am using Reportz.io We used this exposure to “push” some other activities. My good friend Rad Basta, who is also a CEO of Reportz.io got published on Forbes on May 29th. The exposure after the AppSumo promo gave us enough credibility to outreach our B2B targets and attracted some bigger fishes… FE. Closing big deals is fun. 100 regular dashboards + my generous offer of 150 dashboards = $560 + $599=$1159 per month just from this single user…which is $13.908 per year, just from this single user… Got me into podcast with one of my favourite LinkedIn “value bombers”, Andrei Zinkevich And many other cool things… 2. Chance to create a community of early adopters, advocates and affiliates that will help you execute every other campaign. Reportz Community grew 100% organicly from 5 to 987 members. We got a lot of feedback how to improve the standard features… We used the “voice of the community” to get in touch with guys from other tools we plan integrations with… We used it to create the inital base of affiliates that will promote Reportz… And we used it as one of the main channels of communication and nurture efforts, beside newsletter. Also, durign the AppSumo launch, and post launch, we got 100+ testemonials (listing just the favourite ones)… We don’t speak about competition but our users do. We are known as generous guys… Listen to this man. Indeed. :) Listen to this man. Remember when I told you about our Industry issues… Complexity Prices Upsells Educational material Customer success & support Well, from previous screenshots you can see that we managed to overcome all of them…yet, the biggest issue was the sole complexity of the tool. That is why we still do . FREE DEMO WITH FOUNDERS And in 30 days it looked like this… In 60 days we had: 264 demoes booked 227 demos done During launch and after-launch we had: And generally, we really pivoted all of our focus on users and their learning curve reduction. Ending word. Don’t look on AppSumo as your profit source. Sure you will make some numbers, but remember that AppSumo takes the biggest part + this is the “once of” payment, which we don’t like… Instead, try to use that AppSumo brings to make your other moves work. the momentum We used AppSumo and massive impact it brings to help us execute: B2B outreach campaigns Social campaigns Approach and negotiate with inlfuencers Get PR coverage for free Approach and create initial affiliate base Grow our community Grow the newsletter and AppSumo was just the wave we needed. 🌊 If you want to give Reportz.io a spin, you are more then welcome to do so. Be sure to reach out to me and I will personally walk you through everything, show you tips and tricks and help you put that client reporting on autopilot! Cheers! 🍻

mnaHow We Acquired 7000+ Users in Less than a Month

Authors: (1) Ruohan Zhang, Department of Computer Science, Stanford University, Institute for Human-Centered AI (HAI), Stanford University & Equally contributed; zharu@stanford.edu; (2) Sharon Lee, Department of Computer Science, Stanford University & Equally contributed; zharu@stanford.edu; (3) Minjune Hwang, Department of Computer Science, Stanford University & Equally contributed; zharu@stanford.edu; (4) Ayano Hiranaka, Department of Mechanical Engineering, Stanford University & Equally contributed; zharu@stanford.edu; (5) Chen Wang, Department of Computer Science, Stanford University; (6) Wensi Ai, Department of Computer Science, Stanford University; (7) Jin Jie Ryan Tan, Department of Computer Science, Stanford University; (8) Shreya Gupta, Department of Computer Science, Stanford University; (9) Yilun Hao, Department of Computer Science, Stanford University; (10) Ruohan Gao, Department of Computer Science, Stanford University; (11) Anthony Norcia, Department of Psychology, Stanford University (12) Li Fei-Fei, 1Department of Computer Science, Stanford University & Institute for Human-Centered AI (HAI), Stanford University; (13) Jiajun Wu, Department of Computer Science, Stanford University & Institute for Human-Centered AI (HAI), Stanford University. Table of Links Abstract & Introduction Brain-Robot Interface (BRI): Background The NOIR System Experiments Results Conclusion, Limitations, and Ethical Concerns Acknowledgments & References Appendix 1: Questions and Answers about NOIR Appendix 2: Comparison between Different Brain Recording Devices Appendix 3: System Setup Appendix 4: Task Definitions Appendix 5: Experimental Procedure Appendix 6: Decoding Algorithms Details Appendix 7: Robot Learning Algorithm Details Appendix 7: Robot Learning Algorithm Details Object and skill learning details We utilize pre-trained R3M as the feature extractor. Our training procedure aims to learn a latent representation of an input image for inferring the correct object-skill pair in the given scene. The feature embedding model is a fully-connected neural network that further encodes the outputs of the foundation model. Model parameters and training hyperparameters are summarized in Table 7. Collecting human data using a BRI system is expensive. To enable few-shot learning, the feature embedding model is trained using a triplet loss [80], which operates on three input vectors: anchor, positive (with the same label as the anchor), and negative (with a different label). Triplet loss pulls inputs with the same labels together by penalizing their distance in the latent space, as well as pushes inputs with different labels away. The loss function is defined as: where a is the anchor vector, p the positive vector, n the negative vector, f the model, and α = 1 is our separation margin. We test our algorithm in the following generalization settings: Generalization test set. For position generalization, we randomize the initial positions of all objects in the scene with fixed orientation and collect 20 different trajectories. For the pose generalization, we randomize both the initial positions and orientations of all objects. Position and pose. . The context-level generalization refers to placing the target object in different environments, defined by different backgrounds, object orientations, and the inclusion of different objects in the scene. We collect 20 different trajectories with these variations. Context . The instance generalization aims to assess the model’s capability to generalize across different types of objects present in the scene. For our target task (MakePasta), we collect 20 trajectories with 20 different kinds of pasta with different shapes, sizes, and colors. Instance . To understand the separability of latent representations generated by our object-skill learning model, we visualize the 1024-dimensional final image representations using the t-SNE data visualization technique [81]. Results for the MakePasta pose generalization test set are shown in Figure 7. The model can well separate each of the different stages of the task, allowing us to retrieve the correct object-skill pair for an unseen image. Latent representation visualization One-shot parameter learning details We empirically found that using DINOv2’s ViT-B model, alongside a 75x100 feature map and a 3x3 sliding window, with cosine similarity as the distance metric, resulted in the best performance for our image resolutions. Design choices. We test the generalization ability of our algorithm on 1008 unique training and test pairs, encompassing four types of generalizations including 8 position trials, 8 orientation trials, 32 context trials, and 960 instance trials. Generalization test set The position and orientation generalizations, shown in Fig. 8 and Fig. 9 respectively, are tested in isolation, e.g. when the position is varied, the orientation is kept the same. Position and orientation The context-level generalization, shown in Fig. 10, refers to placing the target object in different environments, e.g. the training image might show the target object in the kitchen while the test image shows the target object in a workspace. Here, we allow for position and orientation to vary as well. Context To test our algorithm’s capability of instance-level generalization, shown in Fig. 11, we collected a set of four different object categories, each containing five unique object instances. Our object categories consist of mug, pen, bottle, and medicine bottle, whereas the bottle and medicine bottle categories consist of images from both the top-down and side views. We test all permutations within each object category including train and test pairs with different camera views. Here, we allow for position and orientation to vary as well. Instance We test our method against baselines on 1080 unique training and test pairs, encompassing four types of generalizations including 8 position trials, 8 orientation trials, 32 context trials, 960 instance trials, 48 trials where we vary all four generalizations simultaneously, and 24 trials from the SetTable task. Test set for comparing our method against baselines Finally, we test our algorithm’s ability to generalize when all four variables differ between the training and test image, shown in Fig. 12. Here, the only object category we use is a mug. Position, orientation, context, and instance simultaneously. This paper is under CC 4.0 license. available on arxiv

NOIR: Neural Signal Operated Intelligent Robots for Everyday Activities: Appendix 7

Authors: (1) Ruohan Zhang, Department of Computer Science, Stanford University, Institute for Human-Centered AI (HAI), Stanford University & Equally contributed; zharu@stanford.edu; (2) Sharon Lee, Department of Computer Science, Stanford University & Equally contributed; zharu@stanford.edu; (3) Minjune Hwang, Department of Computer Science, Stanford University & Equally contributed; zharu@stanford.edu; (4) Ayano Hiranaka, Department of Mechanical Engineering, Stanford University & Equally contributed; zharu@stanford.edu; (5) Chen Wang, Department of Computer Science, Stanford University; (6) Wensi Ai, Department of Computer Science, Stanford University; (7) Jin Jie Ryan Tan, Department of Computer Science, Stanford University; (8) Shreya Gupta, Department of Computer Science, Stanford University; (9) Yilun Hao, Department of Computer Science, Stanford University; (10) Ruohan Gao, Department of Computer Science, Stanford University; (11) Anthony Norcia, Department of Psychology, Stanford University (12) Li Fei-Fei, 1Department of Computer Science, Stanford University & Institute for Human-Centered AI (HAI), Stanford University; (13) Jiajun Wu, Department of Computer Science, Stanford University & Institute for Human-Centered AI (HAI), Stanford University. Table of Links Abstract & Introduction Brain-Robot Interface (BRI): Background The NOIR System Experiments Results Conclusion, Limitations, and Ethical Concerns Acknowledgments & References Appendix 1: Questions and Answers about NOIR Appendix 2: Comparison between Different Brain Recording Devices Appendix 3: System Setup Appendix 4: Task Definitions Appendix 5: Experimental Procedure Appendix 6: Decoding Algorithms Details Appendix 7: Robot Learning Algorithm Details Appendix 6: Decoding Algorithms Details For both SSVEP and MI, we select a subset of channels and discard the signals from the rest, as shown in Figure 6. They correspond to the visual cortex for SSVEP, and the motor and visual areas for MI (with peripheral areas). For muscle tension (jaw clenching), we retain all channels. . To predict the object of interest, we apply Canonical Correlation Analysis (CCA) as shown in [77] to the collected SSVEP data. As each potential object of interest is flashing at a different frequency, we are able to generate reference signals Yf for each frequency fi: SSVEP n where f is the sampling frequency and N is the number of samples. s s By calculating the maximum correlation ρfn for each frequency f used for potential objects of interest, we are then able to predict the output class by finding argmax (ρf ) and matching the result to the object of interest with that frequency. n fn n Furthermore, we are able to return a list of predicted objects of interest in descending order of likelihood by matching each object to a list of descending maximum correlations ρf . n Motor imagery. To perform MI classification, we first band-pass filter the data between 8Hz - 30Hz, as that is the frequency range that includes the µ-band and β-band signals relevant to MI. The data is then transformed using the Common Spatial Pattern (CSP) algorithm. CSP is a linear transformation technique that applies a rotation to the data to orthogonalize the components where the over-timestep variance of the data differs the most across classes. We can then use the logvariance of each time series after rotation as features and perform QDA. Thereafter, we extract features by taking the normalized variance of this transformed data (called “CSP-space data”). We then perform Quadratic Discriminant Analysis (QDA) on this data. To calculate our calibration accuracy, we perform K-fold cross validation with KCV = 4, but we use the entire calibrate dataset to fit the classifier for deployment at task-time. Facial muscle tension results in a very significant high-variance signal across almost all channels that is very detectable using simple variance-based threshold filters without having to perform any frequency filters. Recall that we record three 500ms-long trials for each class (“Rest”, “Clench”). In short, for each of the calibration time-series, we take the variance of the channel with the median variance; call this variance m. Then, we just take the mid-point between the maximum m between the rest samples, and the minimum m between the clench samples, and have this be our threshold variance level. This paper is under CC 4.0 license. available on arxiv

NOIR: Neural Signal Operated Intelligent Robots for Everyday Activities: Appendix 6

Authors: (1) Ruohan Zhang, Department of Computer Science, Stanford University, Institute for Human-Centered AI (HAI), Stanford University & Equally contributed; zharu@stanford.edu; (2) Sharon Lee, Department of Computer Science, Stanford University & Equally contributed; zharu@stanford.edu; (3) Minjune Hwang, Department of Computer Science, Stanford University & Equally contributed; zharu@stanford.edu; (4) Ayano Hiranaka, Department of Mechanical Engineering, Stanford University & Equally contributed; zharu@stanford.edu; (5) Chen Wang, Department of Computer Science, Stanford University; (6) Wensi Ai, Department of Computer Science, Stanford University; (7) Jin Jie Ryan Tan, Department of Computer Science, Stanford University; (8) Shreya Gupta, Department of Computer Science, Stanford University; (9) Yilun Hao, Department of Computer Science, Stanford University; (10) Ruohan Gao, Department of Computer Science, Stanford University; (11) Anthony Norcia, Department of Psychology, Stanford University (12) Li Fei-Fei, 1Department of Computer Science, Stanford University & Institute for Human-Centered AI (HAI), Stanford University; (13) Jiajun Wu, Department of Computer Science, Stanford University & Institute for Human-Centered AI (HAI), Stanford University. Table of Links Abstract & Introduction Brain-Robot Interface (BRI): Background The NOIR System Experiments Results Conclusion, Limitations, and Ethical Concerns Acknowledgments & References Appendix 1: Questions and Answers about NOIR Appendix 2: Comparison between Different Brain Recording Devices Appendix 3: System Setup Appendix 4: Task Definitions Appendix 5: Experimental Procedure Appendix 6: Decoding Algorithms Details Appendix 7: Robot Learning Algorithm Details Appendix 5: Experimental Procedure . In our experiments, we use the 128-channel HydroCel Geodesic SensorNet from Magstim EGI, which has sponge tips in its electrode channels. Prior to experiments, the EEG net is soaked in a solution containing dissolved conductive salt (Potassium Chloride) and baby shampoo for 15 minutes. After the soaking, the net is worn by the experiment subject, and an impedance check is done. This impedance check entails ensuring that the impedance of each channel electrode is ≤ 50.0 kΩ, by using a syringe to add more conductive fluid between the electrodes and the scalp. We then carefully put on a shower cap to minimize the drying of conductive fluid over the course of the experiment. EEG device preparation Before commencing the experiments, subjects are given instructions on how to execute the SSVEP, MI, and muscle tension (jaw-clenching) tasks. For SSVEP, they are instructed to simply focus on the flickering object of interest without getting distracted by the other objects on the screen. For MI, similar to datasets such as BCI Competition 2003 [75], and as per extensive literature review [76], we instruct subjects to either imagine continually bending their hands at the wrist (wrist dorsiflexion) or squeezing a ball for the hand actions (“Left”, “Right”), and to imagine depressing a pedal with both feet (feet dorsiflexion) for the “Legs” action. For the “Rest” class, as is common practice in EEG experiments in general, we instruct users to focus on a fixation cross displayed on the screen. Subjects were told to stick with their actions of choice throughout the experiment, for consistency. For muscle tension, subjects were told to simply clench their jaw without too much or too little effort. Instructions to subjects. . For SSVEP, subjects are told in writing on the screen to focus on the object of interest. Thereafter, a scene image of the objects with flickering masks overlaid on each object is presented, and we immediately begin recording the EEG Data over this period of time. For MI, the cues are different during calibration and task-time. During calibration, subjects are presented with a warning symbol (.) on screen for 1 second, before being presented with the symbol representing the action they are to imagine (<-: “Left”, ->: “Right”, v: “Legs”, +: “Rest”), which lasts on screen for 5500 ms. We record the latter 5000 ms of EEG data. After which, there is a randomized period of rest the lasts between 0.5 and 2 seconds, before the process repeats for another randomly chosen action class. This is done in 4 blocks of 5 trials per action, for a total of 20 trials per action. This procedure is again similar to datasets like BCI Competition 2003 [75], that use non-linguistic cues and randomization of rest / task. At task-time, similar to SSVEP, subjects are told in writing on the screen to perform MI to select a robot skill to execute. Thereafter, a written mapping of a class symbol ({<-, ->, v, +}) to skill ({pick from top, pick from side, ...}) is presented, and we begin recording EEG Data after a 2-second delay. For muscle tension, there is also a calibration phase, similar to MI, which entails collecting three 500ms-long trials for each class (“Rest”, and “Clench”) at the start of each experiment. The cues are written on the screen in words. At task time, when appropriate, written prompts are also presented on the screen (e.g. “clench if incorrect”), followed by a written countdown, after which the user has a 500ms window to clench (or not). Interface This paper is under CC 4.0 license. available on arxiv

NOIR: Neural Signal Operated Intelligent Robots for Everyday Activities: Appendix 5

Authors: (1) Ruohan Zhang, Department of Computer Science, Stanford University, Institute for Human-Centered AI (HAI), Stanford University & Equally contributed; zharu@stanford.edu; (2) Sharon Lee, Department of Computer Science, Stanford University & Equally contributed; zharu@stanford.edu; (3) Minjune Hwang, Department of Computer Science, Stanford University & Equally contributed; zharu@stanford.edu; (4) Ayano Hiranaka, Department of Mechanical Engineering, Stanford University & Equally contributed; zharu@stanford.edu; (5) Chen Wang, Department of Computer Science, Stanford University; (6) Wensi Ai, Department of Computer Science, Stanford University; (7) Jin Jie Ryan Tan, Department of Computer Science, Stanford University; (8) Shreya Gupta, Department of Computer Science, Stanford University; (9) Yilun Hao, Department of Computer Science, Stanford University; (10) Ruohan Gao, Department of Computer Science, Stanford University; (11) Anthony Norcia, Department of Psychology, Stanford University (12) Li Fei-Fei, 1Department of Computer Science, Stanford University & Institute for Human-Centered AI (HAI), Stanford University; (13) Jiajun Wu, Department of Computer Science, Stanford University & Institute for Human-Centered AI (HAI), Stanford University. Table of Links Abstract & Introduction Brain-Robot Interface (BRI): Background The NOIR System Experiments Results Conclusion, Limitations, and Ethical Concerns Acknowledgments & References Appendix 1: Questions and Answers about NOIR Appendix 2: Comparison between Different Brain Recording Devices Appendix 3: System Setup Appendix 4: Task Definitions Appendix 5: Experimental Procedure Appendix 6: Decoding Algorithms Details Appendix 7: Robot Learning Algorithm Details Appendix 4: Task Definitions For systematic evaluation of task success, we provide formal definitions of our tasks in the format of BEHAVIOR Domain Definition Language (BDDL) language [69, 71]. BDDL is a predicate logic-based language that establishes a symbolic state representation built on predefined, meaningful predicates grounded in physical states [71]. Each task is defined in BDDL as an initial and goal condition parametrizing sets of possible initial states and satisfactory goal states, as shown in the figures at the end of the appendix. Compared to scene- or pose-specific definitions which are too restricted, BDDL is more intuitive to humans while providing concrete evaluation metrics for measuring task success. This paper is under CC 4.0 license. available on arxiv

NOIR: Neural Signal Operated Intelligent Robots for Everyday Activities: Appendix 4

Authors: (1) Ruohan Zhang, Department of Computer Science, Stanford University, Institute for Human-Centered AI (HAI), Stanford University & Equally contributed; zharu@stanford.edu; (2) Sharon Lee, Department of Computer Science, Stanford University & Equally contributed; zharu@stanford.edu; (3) Minjune Hwang, Department of Computer Science, Stanford University & Equally contributed; zharu@stanford.edu; (4) Ayano Hiranaka, Department of Mechanical Engineering, Stanford University & Equally contributed; zharu@stanford.edu; (5) Chen Wang, Department of Computer Science, Stanford University; (6) Wensi Ai, Department of Computer Science, Stanford University; (7) Jin Jie Ryan Tan, Department of Computer Science, Stanford University; (8) Shreya Gupta, Department of Computer Science, Stanford University; (9) Yilun Hao, Department of Computer Science, Stanford University; (10) Ruohan Gao, Department of Computer Science, Stanford University; (11) Anthony Norcia, Department of Psychology, Stanford University (12) Li Fei-Fei, 1Department of Computer Science, Stanford University & Institute for Human-Centered AI (HAI), Stanford University; (13) Jiajun Wu, Department of Computer Science, Stanford University & Institute for Human-Centered AI (HAI), Stanford University. Table of Links Abstract & Introduction Brain-Robot Interface (BRI): Background The NOIR System Experiments Results Conclusion, Limitations, and Ethical Concerns Acknowledgments & References Appendix 1: Questions and Answers about NOIR Appendix 2: Comparison between Different Brain Recording Devices Appendix 3: System Setup Appendix 4: Task Definitions Appendix 5: Experimental Procedure Appendix 6: Decoding Algorithms Details Appendix 7: Robot Learning Algorithm Details Appendix 3: System Setup The robot we use in our tabletop manipulation task is a standard Franka Emika robot arm with three RealSense cameras. For mobile manipulation, we use a Tiago++ model from PAL Robotics, with an omnidirectional base, two 7-degrees-of-freedom arms with parallel-yaw grippers, a 1-degree-of-freedom prismatic torso, two SICK LiDAR sensors (back and front of the base), and an ASUS Xtion RGB-D camera mounted on the robot’s head, which can be controlled in yaw and pitch. All sensors and actuators are connected through the Robot Operating System, ROS [74]. The code runs on a laptop with an Nvidia GTX 1070 that sends the commands to the onboard robot computer to be executed. Robot platform. A list of primitive skills along with their parameters can be found in Table 6, eight for Franka (16 tasks) and five for Tiago (four tasks). Human users can accomplish all 20 tasks, which are long-horizon and challenging, using these skills. Primitive skills list. This paper is under CC 4.0 license. available on arxiv

NOIR: Neural Signal Operated Intelligent Robots for Everyday Activities: Appendix 3

Authors: (1) Ruohan Zhang, Department of Computer Science, Stanford University, Institute for Human-Centered AI (HAI), Stanford University & Equally contributed; zharu@stanford.edu; (2) Sharon Lee, Department of Computer Science, Stanford University & Equally contributed; zharu@stanford.edu; (3) Minjune Hwang, Department of Computer Science, Stanford University & Equally contributed; zharu@stanford.edu; (4) Ayano Hiranaka, Department of Mechanical Engineering, Stanford University & Equally contributed; zharu@stanford.edu; (5) Chen Wang, Department of Computer Science, Stanford University; (6) Wensi Ai, Department of Computer Science, Stanford University; (7) Jin Jie Ryan Tan, Department of Computer Science, Stanford University; (8) Shreya Gupta, Department of Computer Science, Stanford University; (9) Yilun Hao, Department of Computer Science, Stanford University; (10) Ruohan Gao, Department of Computer Science, Stanford University; (11) Anthony Norcia, Department of Psychology, Stanford University (12) Li Fei-Fei, 1Department of Computer Science, Stanford University & Institute for Human-Centered AI (HAI), Stanford University; (13) Jiajun Wu, Department of Computer Science, Stanford University & Institute for Human-Centered AI (HAI), Stanford University. Table of Links Abstract & Introduction Brain-Robot Interface (BRI): Background The NOIR System Experiments Results Conclusion, Limitations, and Ethical Concerns Acknowledgments & References Appendix 1: Questions and Answers about NOIR Appendix 2: Comparison between Different Brain Recording Devices Appendix 3: System Setup Appendix 4: Task Definitions Appendix 5: Experimental Procedure Appendix 6: Decoding Algorithms Details Appendix 7: Robot Learning Algorithm Details Appendix 2: Comparison between Different Brain Recording Devices We use the EGI NetStation EEG system which uses rapid application 128-channel saline-based EGI SensorNets. Here we justify our choice of using non-invasive, saline-based EEG as the recording device for brain signals. A comparison of different brain reading devices (gel-based EEG, dry EEG, MEG, fMRI, fNIRS, implant) and their advantages and disadvantages are shown in Table 5, using our device as the baseline. Two noticeable alternatives are functional magnetic resonance imaging (fMRI) and invasive implants. fMRI measures the small changes in blood flow that occur with brain activity, which has a very high spatial resolution hence fine-grained information such as object categories and language [31] can be decoded from it. But fMRI suffers from low temporal resolution, and the recording device is extremely costly and cannot be used in daily scenarios. Brain implants have a very good signal-to-noise ratio and have great potential. However, the main concern is that it requires surgery to be applied, and health-related risks are not negligible. This paper is under CC 4.0 license. available on arxiv

NOIR: Neural Signal Operated Intelligent Robots for Everyday Activities: Appendix 2

NOIR: Neural Signal Operated Intelligent Robots for Everyday Activities: Appendix 1

NOIR: Neural Signal Operated Intelligent Robots: Acknowledgments & References

Authors: (1) Ruohan Zhang, Department of Computer Science, Stanford University, Institute for Human-Centered AI (HAI), Stanford University & Equally contributed; zharu@stanford.edu; (2) Sharon Lee, Department of Computer Science, Stanford University & Equally contributed; zharu@stanford.edu; (3) Minjune Hwang, Department of Computer Science, Stanford University & Equally contributed; zharu@stanford.edu; (4) Ayano Hiranaka, Department of Mechanical Engineering, Stanford University & Equally contributed; zharu@stanford.edu; (5) Chen Wang, Department of Computer Science, Stanford University; (6) Wensi Ai, Department of Computer Science, Stanford University; (7) Jin Jie Ryan Tan, Department of Computer Science, Stanford University; (8) Shreya Gupta, Department of Computer Science, Stanford University; (9) Yilun Hao, Department of Computer Science, Stanford University; (10) Ruohan Gao, Department of Computer Science, Stanford University; (11) Anthony Norcia, Department of Psychology, Stanford University (12) Li Fei-Fei, 1Department of Computer Science, Stanford University & Institute for Human-Centered AI (HAI), Stanford University; (13) Jiajun Wu, Department of Computer Science, Stanford University & Institute for Human-Centered AI (HAI), Stanford University. Table of Links Abstract & Introduction Brain-Robot Interface (BRI): Background The NOIR System Experiments Results Conclusion, Limitations, and Ethical Concerns Acknowledgments & References Appendix 1: Questions and Answers about NOIR Appendix 2: Comparison between Different Brain Recording Devices Appendix 3: System Setup Appendix 4: Task Definitions Appendix 5: Experimental Procedure Appendix 6: Decoding Algorithms Details Appendix 7: Robot Learning Algorithm Details 5 Results We seek to provide answers to the following questions through extensive evaluation: 1) Is NOIR truly general-purpose, in that it allows all of our human subjects to accomplish the diverse set of everyday tasks we have proposed? 2) Does our decoding pipeline provide accurate decoding results? 3) Does our proposed robot learning and intention prediction algorithm improve NOIR’s efficiency? Table 1 summarizes the performance based on two metrics: the number of attempts until success and the time to complete the task in successful trials. When the participant reached an unrecoverable state in task execution, we reset the environment and the participant reattempted the task from the beginning. Task horizons (number of primitive skills executed) are included as a reference. Although these tasks are long-horizon and challenging, NOIR shows very encouraging results: on average, tasks can be completed with only 1.83 attempts. The reason for task failures is human errors in skill and parameter selection, i.e., the users pick the wrong skills or parameters, which leads to non-recoverable states and needs manual resets. Decoding errors or robot execution errors are avoided thanks to our safety mechanism with confirmation and interruption. Although our primitive skill library is limited, human users find novel usage of these skills to solve tasks creatively. Hence we observe emerging capabilities such as extrinsic dexterity. For example, in task CleanBook (Fig. 1.6), Franka’s Pick skill is not designed to grasp a book from the table, but users learn to push the book towards the edge of the table and grasp it from the side. In CutBanana (Fig. 1.12), users utilize Push skill to cut. The average task completion time is 20.29 minutes. Note that the time humans spent on decision-making and decoding is relatively long (80% of total time), partially due to the safety mechanism. Later, we will show that our proposed robot learning algorithms can address this issue effectively. System performance. A key to our system’s success is the accuracy in decoding brain signals. Table 2 summarizes the decoding accuracy of different stages. We find that CCA on SSVEP produces a high accuracy of 81.2%, meaning that object selection is mostly accurate. As for CSP + QDA on MI for parameter selection, the 2-way classification model performs at 73.9% accuracy, which is consistent with current literature [36]. The 4-way skill-selection classification models perform at about 42.2% accuracy. Though this may not seem high, it is competitive considering inconsistencies attributed to long task duration (hence the discrepancy between calibration and task-time accuracies). Our calibration time is only 10 minutes, which is significantly shorter compared to the duration of typical MI calibration and training sessions by several orders of magnitude [21]. More calibration provides more data for training more robust classifiers, and allows human users to practice more which typically yields stronger brain signals. Overall, the decoding accuracy is satisfactory, and with the safety mechanism, there has been no instance of task failure caused by incorrect decoding. Decoding accuracy. We then answer the third question: Does our proposed robot learning algorithm improve NOIR’s efficiency? First, we evaluate object and skill selection learning. We collect a dataset offline with 15 training samples for each object-skill pair in MakePasta task. Given an image, a prediction is considered correct if both the object and the skill are predicted correctly. Results are shown in Table 3. While a simple image classification model using ResNet [72] achieves an average accuracy of 0.31, our method with a pre-trained ResNet backbone achieves significantly higher accuracy at 0.73, highlighting the importance of contrastive learning and retrieval-based learning. Using R3M as the feature extractor further improves the performance to 0.94. The generalization ability of the algorithm is tested on the same MakePasta task. For instance-level generalization, 20 different types of pasta are used; for context generalization, we randomly select and place 20 task-irrelevant objects in the background. Results are shown in Table 3. In all variations, our model achieves accuracy over 93%, meaning that the human can skip the skill and object selection 93% of the time, significantly reducing their time and effort. We further test our algorithm during actual task execution (Fig. 5). A human user completes the task with and without object-skill prediction two times each. With object and skill learning, the average time required for each object-skill selection is reduced by 60% from 45.7 to 18.1 seconds. More details about the experiments and visualization of learned representation can be found in Appendix 7.1. Object and skill selection results. One-shot parameter learning results. First, using our pre-collected dataset (see Appendix 7.2), we compare our algorithm against multiple baselines. The MSE values of the predictions are shown in Table 4. Random sample shows the average error when randomly predicting points in the 2D space. Sample on objects randomly predicts a point on objects and not on the background; the object masks here are detected with the Segment Anything Model (SAM) [73]. For Pixel similarity, we employ the cosine similarity and sliding window techniques used in our algorithm, but on raw images without using DINOv2 features. All of the baselines are drastically outperformed by our algorithm. Second, our one-shot learning method demonstrates robust generalization capability, as tested on the respective dataset; table 4 presents the results. The low prediction error means that users spend much less effort in controlling the cursor to move to the desired position. We further demonstrate the effectiveness of the parameter learning algorithm in actual task execution for SetTable, quantified in terms of saved human effort in controlling the cursor movement (Fig. 5). Without learning, the cursor starts at the chosen object or the center of the screen. The predicted result is used as the starting location for cursor control which led to a considerable decrease in cursor movement, with the mean distance reduced by 41%. These findings highlight the potential of parameter learning in improving efficiency and reducing human effort. More results and visualizations can be found in Appendix 7.2. This paper is under CC 4.0 license. available on arxiv

Decoding Human Intentions: NOIR's Performance and Learning Algorithms

Authors: (1) Ruohan Zhang, Department of Computer Science, Stanford University, Institute for Human-Centered AI (HAI), Stanford University & Equally contributed; zharu@stanford.edu; (2) Sharon Lee, Department of Computer Science, Stanford University & Equally contributed; zharu@stanford.edu; (3) Minjune Hwang, Department of Computer Science, Stanford University & Equally contributed; zharu@stanford.edu; (4) Ayano Hiranaka, Department of Mechanical Engineering, Stanford University & Equally contributed; zharu@stanford.edu; (5) Chen Wang, Department of Computer Science, Stanford University; (6) Wensi Ai, Department of Computer Science, Stanford University; (7) Jin Jie Ryan Tan, Department of Computer Science, Stanford University; (8) Shreya Gupta, Department of Computer Science, Stanford University; (9) Yilun Hao, Department of Computer Science, Stanford University; (10) Ruohan Gao, Department of Computer Science, Stanford University; (11) Anthony Norcia, Department of Psychology, Stanford University (12) Li Fei-Fei, 1Department of Computer Science, Stanford University & Institute for Human-Centered AI (HAI), Stanford University; (13) Jiajun Wu, Department of Computer Science, Stanford University & Institute for Human-Centered AI (HAI), Stanford University. Table of Links Abstract & Introduction Brain-Robot Interface (BRI): Background The NOIR System Experiments Results Conclusion, Limitations, and Ethical Concerns Acknowledgments & References Appendix 1: Questions and Answers about NOIR Appendix 2: Comparison between Different Brain Recording Devices Appendix 3: System Setup Appendix 4: Task Definitions Appendix 5: Experimental Procedure Appendix 6: Decoding Algorithms Details Appendix 7: Robot Learning Algorithm Details 4 Experiments NOIR can greatly benefit those who require assistance with everyday activities. We select tasks from the BEHAVIOR benchmark [69] and Activities of Daily Living [70] to capture actual human needs. The tasks are shown in Fig. 1, and consist of 16 tabletop tasks and four mobile manipulation tasks. The tasks encompass various categories, including eight meal preparation tasks, six cleaning tasks, three personal care tasks, and three entertainment tasks. For systematic evaluation of task success, we provide formal definitions of these activities in the BDDL language format [69, 71], which specifies the initial and goal conditions of a task using first-order logic. Task definitions and figures can be found in Appendix 4. Tasks. The human study conducted has received approval from Institutional Review Board. Three healthy human participants (2 male, 1 female) performed all 15 Franka tasks. Sukiyaki, four Tiago tasks, and learning tasks are performed by one user. We use the EGI NetStation EEG system, which is completely non-invasive, making almost everyone an ideal subject. Before experiments, users are familiarized with task definitions and system interfaces. During the experiment, users stay in an isolated room, remain stationary, watch the robot on a screen, and solely rely on brain signals to communicate with the robots (more details about the procedure can be found in Appendix 5). Procedure. This paper is under CC 4.0 license. available on arxiv

Human-Centric Evaluation of NOIR in Real-world Tasks

Authors: (1) Ruohan Zhang, Department of Computer Science, Stanford University, Institute for Human-Centered AI (HAI), Stanford University & Equally contributed; zharu@stanford.edu; (2) Sharon Lee, Department of Computer Science, Stanford University & Equally contributed; zharu@stanford.edu; (3) Minjune Hwang, Department of Computer Science, Stanford University & Equally contributed; zharu@stanford.edu; (4) Ayano Hiranaka, Department of Mechanical Engineering, Stanford University & Equally contributed; zharu@stanford.edu; (5) Chen Wang, Department of Computer Science, Stanford University; (6) Wensi Ai, Department of Computer Science, Stanford University; (7) Jin Jie Ryan Tan, Department of Computer Science, Stanford University; (8) Shreya Gupta, Department of Computer Science, Stanford University; (9) Yilun Hao, Department of Computer Science, Stanford University; (10) Ruohan Gao, Department of Computer Science, Stanford University; (11) Anthony Norcia, Department of Psychology, Stanford University (12) Li Fei-Fei, 1Department of Computer Science, Stanford University & Institute for Human-Centered AI (HAI), Stanford University; (13) Jiajun Wu, Department of Computer Science, Stanford University & Institute for Human-Centered AI (HAI), Stanford University. Table of Links Abstract & Introduction Brain-Robot Interface (BRI): Background The NOIR System Experiments Results Conclusion, Limitations, and Ethical Concerns Acknowledgments & References Appendix 1: Questions and Answers about NOIR Appendix 2: Comparison between Different Brain Recording Devices Appendix 3: System Setup Appendix 4: Task Definitions Appendix 5: Experimental Procedure Appendix 6: Decoding Algorithms Details Appendix 7: Robot Learning Algorithm Details 3 The NOIR System The challenges we try to tackle are: 1) How do we build a general-purpose BRI system that works for a variety of tasks? 2) How do we decode relevant communication signals from human brains? 3) How do we make robots more intelligent and adaptive for more efficient collaboration? An overview of our system is shown in Fig. 2. Humans act as planning agents to perceive, plan, and communicate behavioral goals to the robot, while robots use pre-defined primitive skills to achieve these goals. The overarching goal of building a general-purpose BRI system is achieved by synergistically integrating two designs together. First, we propose a novel modular brain decoding pipeline for human intentions, in which the human intended goal is decomposed into three components: what, how, and where (Sec. 3.1). Second, we equip the robots with a library of parameterized primitive skills to accomplish human-specified goals (Sec. 3.2). This design enables humans and robots to collaborate to accomplish a variety of challenging, long-horizon everyday tasks. At last, we show a key feature of NOIR to allow robots to act more efficiently and to be capable of adapting to individual users, we adopt few-shot imitation learning from humans (Sec. 3.3). 3.1 The brain: A modular decoding pipeline We hypothesize that the key to building a general-purpose EEG decoding system is modularization. Decoding complete behavioral goals (e.g., in the form of natural language) is only feasible with expensive devices like fMRI, and with many hours of training data for each individual [31]. As shown in Fig. 3, we decompose human intention into three components: (a) What object to manipulate; (b) How to interact with the object; (c) Where to interact. The decoding of specific user intents from EEG signals is challenging but can be done with steady-state visually evoked potential and motor imagery, as introduced in Sec. 2. For brevity, details of decoding algorithms are in Appendix 6. Upon showing the task set-up on a screen, we first infer the user’s intended object. We make objects on the screen flicker with different frequencies (Fig. 3a), which, when focused on by the user, evokes SSVEP [26]. By identifying which frequency is stronger in the EEG data, we may infer the frequency of the flickering visual stimulus, and hence the object that the user focuses on. We apply modern computer vision techniques to circumvent the problem of having to physically attach LED lights [27, 28]. Specifically, we use the foundation model OWL-ViT [32] to detect and track objects, which takes in an image and object descriptions and outputs object segmentation masks. By overlaying each mask of different flickering frequencies (6Hz, 7.5Hz, 8.57Hz, and 10Hz [33, 34]), and having the user focus on the desired object for 10 seconds, we are able to identify the attended object. Selecting objects with steady-state visually evoked potential (SSVEP). We use only the signals from the visual cortex (Appendix 6) and preprocess the data with a notch filter. We then use Canonical Correlation Analysis (CCA) for classification [35]. We create a Canonical Reference Signal (CRS), which is a set of sin and cos waves, for each of our frequencies and their harmonics. We then use CCA to calculate the frequency whose CRS has the highest correlation with the EEG signal, and identify the object that was made to flicker at that frequency. . The user then chooses a skill and its parameters. We frame this as a k-way (k ≤ 4) MI classification problem, where we aim to decode which of the k pre-decided actions the user was imagining. Unlike SSVEP, a small amount of calibration data (10-min) is required due to the distinct nature of each user’s MI signals. The four classes are: Left Hand, Right Hand, Legs, and Rest; the class names describe the body parts that users imagine using to execute some skills (e.g. pushing a pedal with feet). Upon being presented with the list of k skill options, we record a 5-second EEG signal, and use a classifier trained on the calibration data. The user then guides a cursor on the screen to the appropriate location for executing the skill. To move the cursor along the x axis, the user is prompted to imagine moving their Left hand for leftward cursor movement. We record another five seconds of data and utilize a 2-way classifier. This process is repeated for x, y, and z axes. Selecting skill and parameters with motor imagery (MI) For decoding, we use only EEG channels around the brain areas related to motor imagery (Appendix 6). The data is band-pass-filtered between 8Hz and 30Hz to include µ-band and β-band frequency ranges correlated with MI activity [36]. The classification algorithm is based on the common spatial pattern (CSP) [37–40] algorithm and quadratic discriminant analysis (QDA). Due to its simplicity, CSP+QDA is explainable and amenable to small training datasets. Contour maps of electrode contributions to the top few CSP-space principal components are shown in the middle row of Fig. 3. There are distinct concentrations around the right and left motor areas, as well as the visual cortex (which correlates with the Rest class). Safety is critical in BRI due to noisy decoding. We follow a common practice and collect electrical signals generated from facial muscle tension (Electromyography, or EMG). This signal appears when users frown or clench their jaws, indicating a negative response. This signal is strong with near-perfect decoding accuracy, and thus we use it to confirm or reject object, skill, or parameter selections. With a pre-determined threshold value obtained through the calibration stage, we can reliably detect muscle tension from 500-ms windows. Confirming or interrupting with muscle tension. 3.2 The robot: Parameterized primitive skills Our robots must be able to solve a diverse set of manipulation tasks under the guidance of humans, which can be achieved by equipping them with a set of parameterized primitive skills. The benefits of using these skills are that they can be combined and reused across tasks. Moreover, these skills are intuitive to humans. Since skill-augmented robots have shown promising results in solving longhorizon tasks, we follow recent works in robotics with parameterized skills [41–52], and augment the action space of our robots with a set of primitive skills and their parameters. Neither the human nor the agent requires knowledge of the underlying control mechanism for these skills, thus the skills can be implemented in any method as long as they are robust and adaptive to various tasks. We use two robots in our experiment: A Franka Emika Panda arm for tabletop manipulation tasks, and a PAL Tiago robot for mobile manipulation tasks (see Appendix for hardware details). Skills for the Franka robot use the operational space pose controller (OSC) [53] from the Deoxys API [54]. For example, Reaching skill trajectories are generated by numerical 3D trajectory interpolation conditioned on the current robot end-effector 6D pose and target pose. Then OSC controls the robot to reach the waypoints along the trajectory orderly. The Tiago robot’s navigation skill is implemented using the ROS MoveBase package, while all other skills are implemented using MoveIt motion planning framework [55]. A complete list of skills for both robots is in Appendix 3. Later, we will show that humans and robots can work together using these skills to solve all the tasks. 3.3 Leveraging robot learning for efficient BRI The modular decoding pipeline and the primitive skill library lay the foundation for NOIR. However, the efficiency of such a system can be further improved. During the collaboration, the robots should learn the user’s object, skill, and parameter selection preferences, hence in future trials, the robot can predict users’ intended goals and be more autonomous, hence reducing the effort required for decoding. Learning and generalization are required since the location, pose, arrangement, and instance of the objects could differ from trial to trial. Meanwhile, the learning algorithms should be sample-efficient since human data is expensive to collect. In NOIR, human effort can be reduced if the robot intelligently learns to propose appropriate object-skill selections for a given state in the task. Inspired by retrieval-based imitation learning [56–58], our proposed method learns a latent state representation from observed states. Given a new state observation, it finds the most similar state in the latent space and the corresponding action. Our method is shown in Fig. 4. During task execution, we record data points that consist of images and the object-skill pairs selected by the human. The images are first encoded by a pre-trained R3M model [59] to extract useful features for robot manipulation tasks, and are then passed through several trainable, fully-connected layers. These layers are trained using contrastive learning with a triplet loss[60] that encourages the images with the same object-skill label to be embedded closer in the latent space. The learned image embeddings and object-skill labels are stored in the memory. During test time, the model retrieves the nearest data point in the latent space and suggests the object-action pair associated with that data point to the human. Details of the algorithm can be found in Appendix 7.1. Retrieval-based few-shot object and skill selection. Parameter selection requires a lot of human effort as it needs precise cursor manipulation through MI. To reduce human effort, we propose a learning algorithm for predicting parameters given an object-skill pair as an initial point for cursor control. Assuming that the user has once successfully pinpointed the precise key point to pick a mug’s handle, does this parameter need to be specified again in the future? Recent advancement in foundation models such as DINOv2 [61] allows us to find corresponding semantic key points, eliminating the need for parameter re-specification. Compared to previous works, our algorithm is one-shot [62–66] and predicts specific 2D points instead of semantic segments [67, 68]. As shown in Fig. 4, given a training image (360 × 240) and parameter choice (x, y), we predict the semantically corresponding point in the test images, in which positions, orientations, instances of the target object, and contexts may vary. We utilize a pre-trained DINOv2 model to obtain semantic features [61]. We input both train and test images into the model and generate 768 patch tokens, each as a pixel-wise feature map of dimension 75 × 100. We then extract a 3 × 3 patch centered around the provided training parameter and search for a matching feature in the test image, using cosine similarity as the distance metric. Details of this algorithm can be found in Appendix 7.2. One-shot skill parameter learning. This paper is under CC 4.0 license. available on arxiv

The NOIR System: Building a General Purpose BRI System 

Authors: (1) Ruohan Zhang, Department of Computer Science, Stanford University, Institute for Human-Centered AI (HAI), Stanford University & Equally contributed; zharu@stanford.edu; (2) Sharon Lee, Department of Computer Science, Stanford University & Equally contributed; zharu@stanford.edu; (3) Minjune Hwang, Department of Computer Science, Stanford University & Equally contributed; zharu@stanford.edu; (4) Ayano Hiranaka, Department of Mechanical Engineering, Stanford University & Equally contributed; zharu@stanford.edu; (5) Chen Wang, Department of Computer Science, Stanford University; (6) Wensi Ai, Department of Computer Science, Stanford University; (7) Jin Jie Ryan Tan, Department of Computer Science, Stanford University; (8) Shreya Gupta, Department of Computer Science, Stanford University; (9) Yilun Hao, Department of Computer Science, Stanford University; (10) Ruohan Gao, Department of Computer Science, Stanford University; (11) Anthony Norcia, Department of Psychology, Stanford University (12) Li Fei-Fei, 1Department of Computer Science, Stanford University & Institute for Human-Centered AI (HAI), Stanford University; (13) Jiajun Wu, Department of Computer Science, Stanford University & Institute for Human-Centered AI (HAI), Stanford University. Table of Links Abstract & Introduction Brain-Robot Interface (BRI): Background The NOIR System Experiments Results Conclusion, Limitations, and Ethical Concerns Acknowledgments & References Appendix 1: Questions and Answers about NOIR Appendix 2: Comparison between Different Brain Recording Devices Appendix 3: System Setup Appendix 4: Task Definitions Appendix 5: Experimental Procedure Appendix 6: Decoding Algorithms Details Appendix 7: Robot Learning Algorithm Details Abstract We present Neural Signal Operated Intelligent Robots (NOIR), a general-purpose, intelligent brain-robot interface system that enables humans to command robots to perform everyday activities through brain signals. Through this interface, humans communicate their intended objects of interest and actions to the robots using electroencephalography (EEG). Our novel system demonstrates success in an expansive array of 20 challenging, everyday household activities, including cooking, cleaning, personal care, and entertainment. The effectiveness of the system is improved by its synergistic integration of robot learning algorithms, allowing for NOIR to adapt to individual users and predict their intentions. Our work enhances the way humans interact with robots, replacing traditional channels of interaction with direct, neural communication. Project website: https://noir-corl.github.io/ Brain-Robot Interface; Human-Robot Interaction Keywords: 1 Introduction Brain-robot interfaces (BRIs) are a pinnacle achievement in the realm of art, science, and engineering. This aspiration, which features prominently in speculative fiction, innovative artwork, and groundbreaking scientific studies, entails creating robotic systems that operate in perfect synergy with humans. A critical component of such systems is their ability to communicate with humans. In human-robot collaboration and robot learning, humans communicate their intents through actions [1], button presses [2, 3], gaze [4–7], facial expression [8], language [9, 10], etc [11, 12]. However, the prospect of direct communication through neural signals stands out to be the most thrilling but challenging medium. We present Neural Signal Operated Intelligent Robots (NOIR), a general-purpose, intelligent BRI system with non-invasive electroencephalography (EEG). The primary principle of this system is hierarchical shared autonomy, where humans define high-level goals while the robot actualizes the goals through the execution of low-level motor commands. Taking advantage of the progress in neuroscience, robotics, and machine learning, our system distinguishes itself by extending beyond previous attempts to make the following contributions. First, NOIR is general-purpose in its diversity of tasks and accessibility. We show that humans can accomplish an expansive array of 20 daily everyday activities, in contrast to existing BRI systems that are typically specialized at one or a few tasks or exist solely in simulation [13–22]. Additionally, the system can be used by the general population, with a minimum amount of training. Second, the “I” in NOIR means that our robots are intelligent and adaptive. The robots are equipped with a library of diverse skills, allowing them to perform low-level actions without dense human supervision. Human behavioral goals can naturally be communicated, interpreted, and executed by the robots with parameterized primitive skills, such as Pick(obj-A) or MoveTo(x,y). Additionally, our robots are capable of learning human intended goals during their collaboration. We show that by leveraging the recent progress in foundation models, we can make such a system more adaptive with limited data. We show that this can significantly increase the efficiency of the system. The key technical contributions of NOIR include a modular neural signal decoding pipeline for human intentions. Decoding human intended goals (e.g., “pick up the mug from the handle”) from neural signals is extremely challenging. We decompose human intention into three components: What object to manipulate, How to interact with the object, and Where to interact, and show that such signals can be decoded from different types of neural data. These decomposed signals naturally correspond to parameterized robot skills and can be communicated effectively to the robots. In 20 household activities involving tabletop or mobile manipulations, three human subjects successfully used our system to accomplish these tasks with their brain signals. We demonstrate that few-shot robot learning from humans can significantly improve the efficiency of our system. This approach to building intelligent robotic systems, which utilizes human brain signals for collaboration, holds immense potential for the development of critical assistive technologies for individuals with or without disabilities and to improve the quality of their life. This paper is under CC 4.0 license. available on arxiv

NOIR: Neural Signal Operated Intelligent Robots for Everyday Activities

NOIR: দৈনন্দিন কার্যকলাপের জন্য নিউরাল সিগন্যাল পরিচালিত বুদ্ধিমান রোবট

NOIR: Mit neuronalen Signalen betriebene intelligente Roboter für alltägliche Aktivitäten

NOIR: Robots inteligentes operados por señales neuronales para actividades cotidianas

NOIR : robots intelligents actionnés par des signaux neuronaux pour les activités quotidiennes

NOIR: रोजमर्रा की गतिविधियों के लिए न्यूरल सिग्नल संचालित इंटेलिजेंट रोबोट

NOIR: 日常活動のための神経信号で動作するインテリジェント ロボット

느와르(NOIR): 일상 활동을 위한 신경 신호로 작동되는 지능형 로봇

NOIR: Robôs Inteligentes Operados por Sinais Neurais para Atividades Diárias

NOIR: интеллектуальные роботы, управляемые нейронными сигналами, для повседневной деятельности

NOIR: Günlük Faaliyetler için Nöral Sinyalle Çalışan Akıllı Robotlar

NOIR: Robot thông minh vận hành bằng tín hiệu thần kinh cho các hoạt động hàng ngày

NOIR：用于日常活动的神经信号操作智能机器人

Rouge et Noir

North and South by Elizabeth Cleghorn Gaskell, is part of the HackerNoon Books Series. You can jump to any chapter in this book . CHAPTER XXXIII. here CHAPTER XXXIII. PEACE. “Sleep on, my love, in thy cold bed, Never to be disquieted! My last Good Night—thou wilt not wake Till I thy fate shall overtake.” Dr. King. Home seemed unnaturally quiet after all this terror and noisy commotion. Her father had seen all due preparation made for her refreshment on her return; and then sate down again in his accustomed chair, to fall into one of his sad waking dreams. Dixon had got Mary Higgins to scold and direct in the kitchen; and her scolding was not the less energetic because it was delivered in an angry whisper; for, speaking above her breath she would have thought irreverent, as long as there was any one lying dead in the house. Margaret had resolved not to mention the crowning and closing affright to her father. There was no use in speaking about it; it had ended well; the only thing to be feared was lest Leonards should in some way borrow money enough to effect his purpose of following Frederick to London, and hunting him out there. But there were immense chances against the success of any such plan; and Margaret determined not to torment herself by thinking of what she could do nothing to prevent. Frederick would be as much on his guard as she could put him; and in a day or two at most he would be safely out of England. “I suppose we shall hear from Mr. Bell to-morrow,” said Margaret. “Yes,” replied her father. “I suppose so.” “If he can come, he will be here to-morrow evening, I should think.” “If he cannot come, I shall ask Mr. Thornton to go with me to the funeral. I cannot go alone. I should break down utterly.” “Don’t ask Mr. Thornton, papa. Let me go with you,” said Margaret, impetuously. “You! My dear, women do not generally go.” “No; because they can’t control themselves. Women of our class don’t go, because they have no power over their emotions, and yet are ashamed of showing them. Poor women go, and don’t care if they are seen overwhelmed with grief. But I promise you, papa, that if you will let me go, I will be no trouble. Don’t have a stranger, and leave me out. Dear papa! if Mr. Bell cannot come, I shall go. I won’t urge my wish against your will, if he does.” Mr. Bell could not come. He had the gout. It was a most affectionate letter, and expressed great and true regret for his inability to attend. He hoped to come and pay them a visit soon, if they would have him; his Milton property required some looking after, and his agent had written to him to say that his presence was absolutely necessary; or else he had avoided coming near Milton as long as he could, and now the only thing that would reconcile him to this necessary visit was the idea that he should see, and might possibly be able to comfort his old friend. Margaret had all the difficulty in the world to persuade her father not to invite Mr. Thornton. She had an indescribable repugnance to this step being taken. The night before the funeral, came a stately note from Mrs. Thornton to Miss Hale, saying that, at her son’s desire, their carriage should attend the funeral, if it would not be disagreeable to the family. Margaret tossed the note to her father. “Oh, don’t let us have these forms,” said she. “Let us go alone—you and me, papa. They don’t care for us, or else he would have offered to go himself, and not have proposed this sending an empty carriage.” “I thought you were so extremely averse to his going, Margaret,” said Mr. Hale in some surprise. “And so I am. I don’t want him to come at all; and I should especially dislike the idea of our asking him. But this seems such a mockery of mourning that I did not expect it from him.” She startled her father by bursting into tears. She had been so subdued in her grief, so thoughtful for others, so gentle and patient in all things, that he could not understand her impatient ways to-night; she seemed agitated and restless; and at all the tenderness which her father in his turn now lavished upon her, she only cried the more. She passed so bad a night that she was ill prepared for the additional anxiety caused by a letter received from Frederick. Mr. Lennox was out of town; his clerk said that he would return by the following Tuesday at the latest; that he might possibly be home on Monday. Consequently, after some consideration, Frederick had determined upon remaining in London a day or two longer. He had thought of coming down to Milton again; the temptation had been very strong; but the idea of Mr. Bell domesticated in his father’s house, and the alarm he had received at the last moment at the railway station, had made him resolve to stay in London. Margaret might be assured he would take every precaution against being tracked by Leonards. Margaret was thankful that she received this letter while her father was absent in her mother’s room. If he had been present, he would have expected her to read it aloud to him, and it would have raised in him a state of nervous alarm which she would have found it impossible to soothe away. There was not merely the fact, which disturbed her excessively, of Frederick’s detention in London, but there were allusions to the recognition at the last moment at Milton, and the possibility of a pursuit, which made her blood run cold; and how then would it have affected her father? Many a time did Margaret repent of having suggested and urged on the plan of consulting Mr. Lennox. At the moment, it had seemed as if it would occasion so little delay—add so little to the apparently small chances of detection; and yet everything that had since occurred had tended to make it so undesirable. Margaret battled hard against this regret of hers for what could not now be helped; this self-reproach for having said what at the time appeared to be wise, but which after events were proving to have been so foolish. But her father was in too depressed a state of mind and body to struggle healthily; he would succumb to all these causes for morbid regret over what could not be recalled. Margaret summoned up all her forces to her aid. Her father seemed to have forgotten that they had any reason to expect a letter from Frederick that morning. He was absorbed in one idea—that the last visible token of the presence of his wife was to be carried away from him, and hidden from his sight. He trembled pitifully as the undertaker’s man was arranging his crape draperies around him. He looked wistfully at Margaret; and, when released, he tottered towards her, murmuring, “Pray for me, Margaret. I have no strength left in me. I cannot pray. I give her up because I must. I try to bear it: indeed I do. I know it is God’s will. But I cannot see why she died. Pray for me, Margaret, that I may have faith to pray. It is a great strait, my child.” Margaret sat by him in the coach, almost supporting him in her arms; and repeating all the noble verses of holy comfort, or texts expressive of faithful resignation, that she could remember. Her voice never faltered; and she herself gained strength by doing this. Her father’s lips moved after her, repeating the well-known texts as her words suggested them; it was terrible to see the patient struggling effort to obtain the resignation which he had not strength to take into his heart as part of himself. Margaret’s fortitude nearly gave way as Dixon, with a slight motion of her hand, directed her notice to Nicholas Higgins and his daughter, standing a little aloof, but deeply attentive to the ceremonial. Nicholas wore his usual fustian clothes, but had a bit of black stuff sown round his hat—a mark of mourning which he had never shown to his daughter Bessy’s memory. But Mr. Hale saw nothing. He went on repeating to himself, mechanically as it were, all the funeral service as it was read by the officiating clergyman; he sighed twice or thrice when all was ended; and then, putting his hand on Margaret’s arm, he mutely entreated to be led away, as if he were blind, and she his faithful guide. Dixon sobbed aloud; she covered her face with her handkerchief, and was so absorbed in her own grief that she did not perceive that the crowd, attracted on such occasions, was dispersing, till she was spoken to by some one close at hand. It was Mr. Thornton. He had been present all the time, standing, with bent head, behind a group of people, so that, in fact, no one had recognised him. “I beg your pardon,—but, can you tell me how Mr. Hale is? And Miss Hale, too? I should like to know how they both are.” “Of course, sir. They are much as is to be expected. Master is terribly broke down. Miss Hale bears up better than likely.” Mr. Thornton would rather have heard that she was suffering the natural sorrow. In the first place, there was selfishness enough in him to have taken pleasure in the idea that his great love might come in to comfort and console her; much the same kind of passionate pleasure which comes stinging through a mother’s heart, when her drooping infant nestles close to her, and is dependent upon her for everything. But this delicious vision of what might have been—in which, in spite of all Margaret’s repulse, he would have indulged only a few days ago—was miserably disturbed by the recollection of what he had seen near the Outwood station. “Miserably disturbed!” that is not strong enough. He was haunted by the remembrance of the handsome young man, with whom she stood in an attitude of such familiar confidence; and the remembrance shot through him like an agony, till it made him clench his hands tight in order to subdue the pain. At that late hour, so far from home! It took a great moral effort to galvanise his trust—erewhile so perfect—in Margaret’s pure and exquisite maidenliness into life; as soon as the effort ceased, his trust dropped down dead and powerless: and all sorts of wild fancies chased each other like dreams, through his mind. Here was a little bit of miserable, gnawing confirmation. “She bore up better than likely” under this grief. She had then some hope to look to, so bright that even in her affectionate nature it could come in to lighten the dark hours of a daughter newly made motherless. Yes! he knew how she would love. He had not loved her without gaining that instinctive knowledge of what capabilities were in her. Her soul would walk in glorious sunlight if any man was worthy, by his power of loving, to win back her love. Even in her mourning she would rest with a peaceful faith upon his sympathy. His sympathy! Whose? That other man’s. And that it was another was enough to make Mr. Thornton’s pale grave face grow doubly wan and stern at Dixon’s answer. “I suppose I may call,” said he coldly. “On Mr. Hale, I mean. He will perhaps admit me after to-morrow or so.” He spoke as if the answer were a matter of indifference to him. But it was not so. For all his pain, he longed to see the author of it. Although he hated Margaret at times, when he thought of that gentle, familiar attitude and all the attendant circumstances, he had a restless desire to renew her picture in his mind—a longing for the very atmosphere she breathed. He was in the Charybdis of passion, and must perforce circle and circle ever nearer round the fatal centre. “I dare say, sir, master will see you. He was very sorry to have to deny you the other day; but circumstances was not agreeable just then.” For some reason or other, Dixon never named this interview that she had had with Mr. Thornton to Margaret. It might have been mere chance, but so it was that Margaret never heard that he had attended her poor mother’s funeral. About HackerNoon Book Series: We bring you the most important technical, scientific, and insightful public domain books. This book is part of the public domain. Elizabeth Cleghorn Gaskell (2003). North and South. Urbana, Illinois: Project Gutenberg. Retrieved https://www.gutenberg.org/cache/epub/4276/pg4276-images.html This eBook is for the use of anyone anywhere at no cost and with almost no restrictions whatsoever. You may copy it, give it away or re-use it under the terms of the Project Gutenberg License included with this eBook or online at , located at . www.gutenberg.org https://www.gutenberg.org/policy/license.html

Stories, News & Mentions about NOIZ

Noiz’s Stories on HackerNoon