Starfish Space is giving life to on-orbit services

Starfish Space

THE LUNARY COPRIS; THE BISON ONITIS

GREAT SCENES ON THE MOON

LOVE AND SUCCESS

Over the past six years, a little-known private equity firm, Vista Equity Partners, has built an educational software empire that wields unseen influence over the educational journeys of tens of millions of children. Along the way, The Markup found, the companies the firm controls have scooped up a massive amount of very personal data on kids, which they use to fuel a suite of predictive analytics products that push the boundaries of technology’s role in education and, in some cases, raise discrimination concerns. One district we examined uses risk-scoring algorithms from a company in the group, PowerSchool, that incorporate indicators of family wealth to predict a student’s future success—a controversial practice that parents don’t know about—raising troubling questions. “I did not even realize there was anybody in this space still doing that [using free and reduced lunch status] in a model being used on real kids,” said Ryan Baker, the director of the University of Pennsylvania’s Center for Learning Analytics. “I am surprised and really appalled.” Vista Equity Partners, which declined to comment for this story, has acquired controlling ownership stakes in some of the leading names in educational technology, including EAB, which sells a suite of college counseling and recruitment products, and PowerSchool, which dominates the market for K-12 data warehousing and analytics. PowerSchool alone on more than 45 million children, including 75 percent of North American K-12 students. Ellucian, a recent Vista acquisition, says it . And EAB’s products are . But parents of those students say they’ve largely been left in the dark about what data the companies collect and how they use it. claims to hold data serves 26 million students used by thousands of colleges and universities “We are paying these vendors and they are making money on our kids’ data,” said Ellen Zavian, whose son was required to use Naviance, college preparation software recently acquired by PowerSchool, at Montgomery Blair High School in Silver Spring, Md. After growing concerned about the questions her son was being asked to answer on Naviance-administered surveys, Zavian and other members of a local student privacy group requested access in 2019 to the data the company holds on their children from the district under the Federal Educational Rights and Privacy Act (FERPA). But to date, she has received back only usernames and passwords. “Parents know very little about this process,” she said. The ed-tech companies in Vista’s portfolio appear to operate largely independently, but they have entered into a number of partnerships that deepen the ties of shared ownership. PowerSchool and EAB, for example, have a aimed at “delivering data movement solutions that drive value and save time for Districts.” data integration partnership The two companies also last year that made EAB the exclusive reseller of some PowerSchool products. signed another deal EAB did not respond to requests for comment. To piece together the extent of the companies’ data collection, The Markup reviewed thousands of pages of obtained through public records requests. contracts, user manuals, data sharing agreements, and survey questions We found that the companies, collectively, gather everything from basic demographic information—entered automatically when a student enrolls in school—to data about students’ citizenship status, religious affiliation, school disciplinary records, medical diagnoses, what speed they read and type at, the full text of answers they give on tests, the pictures they draw for assignments, whether they live in a two-parent household, whether they’ve used drugs, been the victim of a crime, or expressed interest in LGBTQ+ groups, among hundreds of other data points. Each Vista-owned company doesn’t necessarily hold all the data points listed here. Some of those data fields were recorded in the traffic between students’ computers and PowerSchool servers when students used their accounts. The Markup reviewed the accounts with students’ permission. Other data fields were listed in districts’ data privacy agreements with PowerSchool and the data library—a list of all available data fields—for one district’s PowerSchool database. Our review offers a more detailed picture of the company’s data operations than PowerSchool publicly discloses, but it is likely an incomplete portrait. According to with school districts, PowerSchool has the right to de-identify the data it holds on their behalf—by removing fields such as names and social security numbers—and use it in any way it sees fit to improve and build its own products. its contracts In some districts, such as Miami-Dade County Public Schools, recent have exceeded $2.5 million for a single year, according to copies of the deals obtained through public records requests. PowerSchool contracts “You don’t pay the oil company to come pump oil off your land; it’s the other way around.” Alex Bowers, Teachers College, Columbia University “It’s hard for me to understand how PowerSchool would not be paying for the privilege” of extracting so much student data, said Alex Bowers, a professor of educational leadership at Columbia University’s Teachers College. “You don’t pay the oil company to come pump oil off your land; it’s the other way around.” PowerSchool declined to answer specific questions about the data it collects and how it uses that information. “At PowerSchool, ensuring student equity, privacy, and access to good quality education is our top priority and is foundational to everything we do,” Darron Flagg, the company’s chief compliance and privacy officer, wrote in a brief statement to The Markup. “PowerSchool strictly and proactively follows legal, regulatory, and voluntary requirements for protecting student privacy including the Family Educational Rights and Privacy Act (FERPA), state regulations, and the Student Privacy Pledge. PowerSchool customers own their student and school data. We do not sell student or school data; we do not collect, maintain, use, or share student personal information beyond what is authorized by the district, parent, or student.” A Cautionary Tale: Elgin, Illinois Many of PowerSchool’s newer product lines, including its predictive analytics tools and personalized learning platform, require troves of student data to train the underlying algorithms. But experts who reviewed The Markup’s findings said that some of the data being used for those purposes is bound to lead to discriminatory outcomes. Consider School District U-46 in Elgin, Ill., which was the only district—out of 27 we submitted public records requests to—that provided a complete list of the data PowerSchool warehouses on its behalf. The district also provided documents detailing how PowerSchool’s predictive analytics algorithms draw on some of that data to influence students’ educational journeys. U-46’s PowerSchool database contains nearly 7,000 data fields about Elgin students, parents, and staff, according to a The Markup obtained. copy of the data library As early as first grade, algorithms from the company’s Unified Insights product line start generating predictions about whether students are at low, moderate, or high risk of not graduating high school on time, not meeting certain standards on the SATs, or not completing two years of college, among other outcomes. The district’s documents describe available via PowerSchool, although U-46 says it does not use most of them. dozens of different predictive models The district begins displaying student on-time graduation risk scores to teachers and administrators beginning in seventh grade, according to Matt Raimondi, Elgin’s assessment and accountability coordinator. Free and reduced lunch status—a proxy for family wealth—and student gender are among the most important factors in determining that risk score, according . At one point, Elgin’s models—developed by a company called Hoonuit that was acquired by PowerSchool in 2020 and rebranded as Unified Insights—also variable. to the documents incorporated student race as a heavily weighted Flagg, from PowerSchool, said race was removed from the models in 2017 before the company acquired Hoonuit. The predictive models also draw on data points like attendance, disciplinary history, and test scores. Learning analytics experts told The Markup that the use of demographic data like gender and free and reduced lunch status—attributes that students and school officials can’t change—to predict student outcomes is bound to encode discrimination into the predictive models. “I think that having [free and reduced lunch status] as a predictor in the model is indefensible in 2021,” said Baker of the University of Pennsylvania’s Center for Learning Analytics. Baker has consulted with BrightBytes, a competitor of PowerSchool in the K-12 predictive analytics space. “Unified Insights does provide the option for school districts to include free and reduced lunch status to enable districts to reduce dropout risk associated with economic hardship and identify additional social service supports that may be available to impacted students,” Flagg, from PowerSchool, wrote in an email. “Including these things that are not within the control of the family or the school is highly problematic,” said Bowers, from Columbia University Teachers College, because even the best-intentioned school cannot change all the systemic gender and wealth disparities that affect a particular student. Basing the risk scores so heavily on those factors therefore obscures the impact of other factors a school may be able to influence, he said. Raimondi said U-46 has chosen not to use many of the predictive models PowerSchool makes available because of their reliance on immutable student characteristics “Especially down at the early grades, we don’t even make it visible to any users besides myself and a programmer,” he said. “The models at the lower grades, they’re not that accurate and they rely a lot more heavily on demographic-type data.” Each year, Elgin’s dropout risk model misses about 90 students in each grade level, out of 3,000 students per grade, who do not go on to graduate on time, according to a by a PowerSchool data scientist and obtained by The Markup. presentation prepared “We have no comment on the sensitivity/specificity of the models,” U-46 spokesperson Karla Jiménez wrote in an email. The Markup on a similar dropout prediction tool EAB sells to colleges and universities. Some of those schools incorporated race as a “high impact predictor” of success, and their algorithms labeled Black students “high risk” at as much as four times the rate of their White peers, effectively steering students of color away from certain majors. has previously reported After our reporting, Texas A&M University as a predictive variable. dropped the use of race The Data Empire Is Growing Vista Equity Partners has been expanding its reach in the educational software industry for years. Along with that expansion, it’s put together a portfolio of companies that amass data and effectively track kids throughout their educational journeys. Since 2015, when Vista first purchased PowerSchool from Pearson for $350 million, Vista has been on a spending spree, acquiring other ed tech companies that collect different kinds of student data. In 2017, PowerSchool bought SunGard K-12, which provided human resources and payroll software for schools. In 2019, it purchased Schoology, a widely used learning management system that served as the digital backbone for many schools’ curricula and lesson plans. It acquired Hoonuit, which provides the predictive risk scoring used by districts like Elgin, in 2020. Last March, it completed the purchase of the college preparatory software Naviance, and in November, it purchased Kickboard, a company that collects data about students’ behavior and social-emotional skills. In , PowerSchool officials have said more acquisitions are a key part of the company’s growth plan. presentations to investors EAB has been on a similar purchasing spree, acquiring companies like Wisr, YouVisit, Cappex, and Starfish that are used for college recruitment, advertising, and tracking students on campus. It also announced the , a “next-generation data warehouse and analytics hub” designed to “break down data silos.” creation of Edify Last June, Vista , which sells a variety of educational technology products. The company claims to serve more than 26 million students across 2,700 institutions. also acquired a co-ownership stake in Ellucian That consolidation of data and power has triggered a backlash from privacy-minded parents, some of whom have been trying, unsuccessfully, to find out what the deals mean for their children’s sensitive data. Piercing the veil of secrecy can be difficult, even when parents turn to privacy laws designed to increase transparency. Illinois, for example, has a state law that requires school districts to post specific information about the ed-tech vendors they use, including all written agreements with vendors and lists of the data elements shared with those vendors. Despite that, districts like Chicago Public Schools have pertaining to PowerSchool and Naviance. CPS has, however, posted data use disclosures for other vendors. Across Illinois, 5,800 schools use PowerSchool software, . yet to post any of the required material according to the company FERPA has also proven of little use for some parents. Cheri Kiesecker, a Colorado parent of two, said that she requested her children’s records under the law from PowerSchool earlier this year after it completed the Naviance deal. “Each school district owns and controls access to its students’ data,” Flagg, from PowerSchool, wrote in an email to The Markup. “Any requests from parents for access to their children’s data must be managed through their respective school districts. '“Our children aren’t products.” Deborah Simmons, parent PowerSchool instructed Kiesecker to request the records through the school, which she did. When PowerSchool did not comply with the school’s subsequent request by the statutory 45-day deadline, her school’s attorneys sent a legal demand to the company, which The Markup reviewed. To date, Kiesecker said, she has still not received her children’s complete records, although PowerSchool has provided partial documentation. Deborah Simmons, a Texas parent, said she began looking into the Vista-owned companies after discovering that her school had automatically uploaded her child’s data into Naviance. She filed public records requests and grievances with her school but still doesn’t know the full extent of the data the companies hold or who else it’s been shared with. “These tech companies want to eliminate the data silos and merge and streamline all of this stuff, but no, our children aren’t products,” Simmons said. That’s what they do, they treat our children like products. They’re human beings and they deserve privacy and freedom.” Written by: Todd Feathers Photo by on Jonas Jacobsson Unsplash Also published here

Your Children's Data Is Being Collected by Educational Technology Companies  

Astounding Stories of Super-Science August 1931: VOL. VII, NO. 2 - The Moon Weed

200 Black Women In Tech To Follow on Twitter {2021 List}

Introduction Project Xanadu® is the original hypertext project — the brainchild of Ted Nelson, and the result of careful and passionate work by innumerable clever people over the course of nearly sixty years. Because of the length of its development period, the project has spawned and used many ideas of varying importance, and particularly important ideas have had many names. Because its history spans several eras of computing, ideas spawned by the project that were once considered radical have become commonplace and other ideas that were once commonplace have become forgotten and become radical again. Over this time, most documentation available to the public has been written by Ted, and intended for a non-technical or semi-technical audience. I had the opportunity to work on Xanadu for five or six years, after spending years as a fan, trying to piece together a general idea of the project from scattered documentation. This gave me the privilege of being able to ask Ted for clarification on both technical and philosophical points, and it also gave me a better view of how different ideas and terms fit into different eras. Xanadu has a poor reputation in many technical communities. Part of this is due to the popularity of a misleading and in parts factually incorrect 1995 Wired article about the project. However, I consider a larger issue to be the project’s tendency (normal through the 1980s but now very strange) to default to secrecy (even with regard to ostensibly non-secret ideas) and to consider any public release of information in terms of PR. This ultimately meant that most available information was lacking in technical detail or had its technical detail hidden behind a general-audience-friendly front. With a few exceptions (such as ), there are no complete, well-organized, publicly-released explanations of Xanadu concepts aimed at engineers. Engineers have, as a side effect, failed to understand how certain ideas fit together and have filled in gaps in the explanation with ideas already familiar to them. Udanax Green / xu88’s FeBe manual Xanadu concepts are, to a great extent, an alternate (and alien) universe. They were developed earlier than their rough general-computing counterparts in many cases, and are not widely understood outside of the community of former and current Xanadu developers. I aim to write the guide I would have liked to have been able to read, back before Ted found me and invited me to the project. With any luck, this will clear up some misconceptions, introduce some unfamiliar ideas, and prepare people who are interested in joining the project with enough background to understand the internal documentation quickly. This document is not intended as an introduction to Xanadu as a whole. In particular, I cannot & will not review the long chronology of the project, although when appropriate I will mention the time period in which something was developed in order to provide context to particular technical decisions. Explaining the historical importance of Xanadu and providing a gentle or gradual introduction to these ideas is also beyond the scope of this document: there is much to be explained and little time, so I will focus on the technical background and assume that readers already have a passing familiarity with Xanadu. For a non-technical introduction and brief chronology, see Ted’s video series . Xanadu Basics I will be covering only projects I have intimate knowledge of — either from working on them, consulting on them, or immersing myself in their documentation. This leaves large gaps in the chronology (such as xu92 / Udanax Gold), and it means that I cannot adequately cover projects started after 2017 (including the current web-based demo). I will describe a few projects and ideas in passing that I don’t have detailed knowledge of, due to their importance — including OSMIC, xu92, and specialized xu88 enfilade types like the POOMfilade; however, my brief description should not be taken as complete or accurate. I will try to mark these cases clearly. In some cases, I have independently written open source implementations of some data structures or algorithms developed under Project Xanadu. I will provide links to these when I have them. They are intended as didactic aids — clear examples of implementations — and are often missing optimizations or integrations with other features; they should therefore not be taken as complete or production-ready implementations. Legal notes Xanadu®, ZigZag®, and the flaming X logo are registered trademarks. When possible, I will be using the following generic terms: ‘xanalogical’ to refer to a system that incorporates ideas from Xanadu; ‘translit’ or ‘transliterature’ to refer to a xanalogical system for representing hypertext or hypermedia; ‘ZZStructure’ to refer to the underlying data structure and facilities within a ZigZag implementation in the absence of a UI layer. While I was at one time privy to trade secrets belonging to Project Xanadu, to my knowledge nothing I describe here is subject to currently-live trade secret protection. Furthermore, while I have written code and documentation whose copyright is ceded to Project Xanadu, I am not writing this based on that code and documentation, but instead based on my memories of it. To my knowledge, nothing explained here is secret. There are several live patents associated with ideas I outline here. , which was granted in 2001 & so should expire in a few years (while non-interactive ZigZag interfaces are not patent-encumbered at all). , and is not covered in this document because at the time I worked on the project it was a trade secret; it was granted in 2014. There are two abandoned applications for patents on transpointing windows: and ; the existence of these applications represents prior art that should ensure that no later patent filed for this technology will be granted. There is from 2017. And, finally, . In my experience, the patent status of Xanadu ideas have been no impediment to independent implementation: Ted is generally happy to see his ideas carefully and seriously implemented. Interactive ZigZag interfaces are covered by US262736B1 Slice intercalation (a means of federating multiple independent zzstructures) is covered by US8682884B2 one in 2008 another in 2012 an application for a patent on navigable transpointing windows there is an expired patent on online payment systems This document is not officially blessed by Project Xanadu, and any inaccuracies are mine. I do not represent Project Xanadu or Ted Nelson. The aerial overview: key concepts Work on Xanadu can be divided into two general realms: transliterature, which involves hypertext and hypermedia, and ZigZag, which involves much smaller pieces of data. These realms very occasionally interact (see my section on ZZOGL / FloatingWorld), but for the most part the technologies involved remain separate, and they have distinct and almost contradictory concerns and philosophies. If we wanted to find a very rough equivalent in the realm of normal computing, transliterature would be word processing while ZigZag would be the use of spreadsheets and databases. Transliterature, in all its incarnations, involves several concepts missing from or alien to mainstream computing: unique permanent addressing for immutable documents and source data, indirect document delivery, visible connections, and external markup. All of these ideas stem from a single basic idea: the manipulation of immutable sequences of bytes with permanent addresses. ZigZag’s universe is based around a particular data structure — a generalization of the spreadsheet, wherein a cell can have connections along arbitrarily many named dimensions. I will focus on transliterature in this document because it has a longer history and more concepts associated with it. It is also what most people associate with the ‘Project Xanadu’ name. Some effort has been made to develop programming languages that use ZigZag natively, the way spreadsheet formula languages take advantage of spreadsheet structure. There is some discussion , and . here here is my contribution to the problem Transcopyright Transcopyright is a concept that gets a lot of push-back from people who don’t really understand what it’s trying to do. I’ll do my best to explain it, with special emphasis on the confusions I have seen. Transcopyright is a system for monetizing or controlling the distribution of one’s material within a transliterature system. It is not DRM, because it applies to data the same ownership rules that are applied to objects: what one has purchased, one owns in perpetuity, including the right to remix and release remixes. It is specifically intended to simplify rights negotiations on derivative works, and it uses the same model as RiffTrax: remixers are providing overlay or re-arrangement rules for material already owned by the people downloading the remix. It relies upon permanent addressing for this. On a technical level, how this works is: when a work subject to transcopyright is published, a one-time pad is generated that is the same length as all the newly-published material in the work. The material is encrypted with that one-time pad, and it is the cyphertext that is published. Anyone can view any span of the cyphertext. Anyone who wants to view a portion of the plaintext must request that portion of the one-time pad from the author or some trusted oracle. A remix or edit can be downloaded by anyone, but those pieces the viewer doesn’t yet have permission to view will be cyphertext (subject to an ODL entry that blacks it out and provides information about how to purchase the rights to it). One may decide to pay for only the bytes present in the remix. Remixes of things the viewer already owns will not require a second copy. Once you have a copy of the plaintext, there are no technical mechanisms to prevent you from doing what you like with it. However, etiquette dictates that you do not distribute the plaintext or encryption key beyond a reasonable level, and normal copyright law kicks in at that point. The software itself does not provide the facility to distribute either on your behalf, for material you have not created. ZZOGL / FloatingWorld XanaduSpace and XanaSpace use a unique system for handling display, wherein a ZZStructure controls the placement, shape, color, and interaction profile of objects in 3d space. This is called ZZOGL (for ZigZag OpenGL) or FloatingWorld (FW for short). FW is an extreme form of the ‘property orientation’ model mentioned in the ZigZag section. In it, cells correspond either to object heads (around which properties coalesce) or property values. Starting from a cell representing the start of the structure (the zzogl head), one dimension represents the draw order. We iterate over objects along this dimension twice. During the first pass, we calculate sizes and locations; during the second pass, we draw using our calculated sizes and locations. Each object calculates its location with respect to some other object within a location-offset DAG. Its parent’s size in three dimensions is multiplied by a vector of between 0 and 1 (the ‘ ’ — the point on the parent with respect to which we move), we add the , and then position it on a point on ourselves (size * center). parentcenter delta pos = parent.pos+(parent.size*(parent.center-parentcenter))+delta-(size*center) This kind of dynamic relative positioning made it possible to represent relatively complex relationships between objects. We used it to display both transliterature and ZigZag systems. There were a couple kinds of objects: ‘groups’: invisible dummy objects for moving whole groups of objects at the same time ‘slabs’: rectangular prisms ‘beams’: tetragonal two-dimensional shapes with a solid color ‘tetroids’: textured two-dimensional shapes that are shaped like tetris-blocks or paragraphs of text (in other words, a rectangle with at most two smaller rectangles abutting it at top or bottom) Unfortunately, getting OpenGL to render large quantities of text on top of 3d objects while being responsive enough to edit that text in real time was a difficult problem, and one that we were never able to solve. As a result, XanaduSpace and XanaSpace never had a release. OSMIC is an alternative document versioning system — a general-purpose journal-based system for tree-based versioning of any byte sequence. Each version is an address in a list of steps. A step can be one of: OSMIC insert bytes at a particular point delete bytes from a particular point A version’s address is, like a tumbler address, a period-separated sequence of numbers. Each time a new version is created, the last number increments. Each time the version tree forks, a new number is added at the end, starting with zero. While it’s possible to reconstruct a version from scratch by running the journal, it’s also possible to rewind to a particular version by treating every delete as an insert and every insert as a delete. An OSMIC journal can be much larger than the byte string it represents, so a compact representation is important. I am not aware of any work on a compact binary representation of OSMIC journals within Xanadu. I haven’t worked with OSMIC, but at one time it was suggested that XanaduSpace should be retrofitted with an OSMIC-based system for storing the state and history of the ZZOGL system (and thus, all internal information necessary to display and edit documents). I have written . It is not compatible with earlier Xanadu-internal implementations. I have also proposed . an open source proof-of-concept implementation of OSMIC an alternative high-performance storage system for it Transliterature editing Several systems exist for transliterature editing, some of them released. None of them are, to my mind, satisfying. All are based around selecting and re-ordering spans from existing documents. and is intended for use with OpenXanadu. XanaduCambridge has . My implementation is called SPANG a less elaborate, more manual system We had pitched a more natural system, wherein selected text could be dragged out of a window to create a new document, or dragged to snap to the end of a document (or to the middle of a document as an insertion) to produce transclusions. We suggested dragging a selected span of text on top of another selected span of text to produce a link between them. Such a system would require an environment where documents floated and text could be dragged between widgets — in other words, something that the cross-platform graphics libraries we were working with didn’t support. Formats On-disk formats (for ZZStructures, EDLs, and ODLs) were constantly in flux while I was at Xanadu. Some were, from my perspective as a professional programmer, abysmal. Most were either based on CSV or key-value pairs. In some cases, they used CSV for mandatory fields and key-value notation for optional fields. I prefer a subset of YAML or JSON for ZZStructures, because ZZStructures fit that model well. However, since these structures have explicit hierarchy, they were controversial. Generally speaking, EDL formats have been one line per span, with ordered triples (address, start, end). Whether or not keys are used differs from spec to spec, as does how (or whether) one can specify a whole document without specifying its length. Generally speaking, ODLs have been lists of addresses pointing to “link” files, and those “link” files contain a type followed by sections called ‘endsets’ that contain one or more spans each. Generally, a link has one or two endsets. If a single endset has non-contiguous spans, the resulting beam resembles a starfish or a hand. I argued in favor of arbitrarily many endsets, each with arbitrary many spans. I also argued in favor of zero-length single-sided links, which I considered to be like bookmarks. I implemented page breaks in XanaSpace as zero-length single-sided links. There have been arguments over whether or not it is appropriate to support spans that are with respect to the concatext of some document (and if so, how to specify that). Ted has generally been against this feature, and I have generally been in favor of it — both in EDLs and ODLs. If you’re planning to implement a xanalogical system, compatibility with an existing format will not be a problem: there are no genuine standards, even within the project. However, you should consider the problems mentioned above.

An Engineer’s Guide to the Docuverse

There is no denying the fact that Artificial Intelligence is the breakthrough technology of recent times. The machines have come a long way from assisting humans in mechanical operations to performing smarter tasks using cognitive intelligence. Every day, we are coming across interesting applications of AI. The ability of Deep Learning algorithms to learn and predict efficiently has opened the doors of possibilities. Nowadays, AI is impacting many other areas as well. In this blog post, we will discuss some niche applications of AI. Let’s discuss such application in detail. Applications of AI in self-driving cars, image processing (Prisma, Google Photos), and Chatbots are something you must be experiencing in your daily life or hearing about it a lot on the internet. However, there are many other niche areas where AI is creating a significant impact. These niche areas of application don’t get as much attention in the press but they actively feature in AI research publications. With this post, let’s explore these niche domains and develop a deeper and well-rounded understanding of applications of AI. Astronomy Astronomers deal with a lot of data in the form of signals and images. Analyzing and categorizing this data is a humongous task. The astronomers are now leveraging the power of unsupervised machine learning to automate this task, which was previously done by thousands of volunteers. Alex Hocking, a masters student from the University of Hertfordshire, who led the work on one such system, commented: “The important thing about our algorithm is that we have not told the machine what to look for in the images, but instead taught it how to ‘see’.” AI is helping astronomers make the blurry and noisy images better. Researchers have used neural networks to enhance the already existing data as well. The deployment of such machine learning algorithms have certainly reduced the human effort and at times, have enhanced the images better than their original source. What did Alex Hocking say about teaching a machine to SEE? The future of AI in astronomy looks brighter as more and more astronomers are shifting to advanced machine learning techniques to analyze the astronomical data. During this era of extensive space exploration, maybe AI can help identify a habitable planet by deeply analyzing the images and signals. Construction and Planning Machine learning algorithms have been used to model the best building layout to minimize energy consumption. It is important to adapt to energy efficient construction practices. The machine learning powered energy can provide the best solutions by combining the analysis from historical and current data. models Certain buildings in Manhattan, NY are already being constructed using machine learning powered energy modeling. Talking about energy conservation, AI is being deployed in optimizing electricity grids for entire cities. is one of the companies providing predictive analytics solutions for smart grids. Machine learning algorithms can analyze the data from millions of smart meters. Clubbing this data to the consumer behavior gives a holistic solution for smart delivery of electricity. This method can learn from user behavior and appliance usage. This will help suppliers understand the peak usage time and the downtime at the granular level. Thus, optimizing the overall supply by the smart electricity grid. Grid4c AI in Agriculture AI is entering even one of the most traditional verticals, Agriculture. The idea is to produce more with same resources. This is where AI is helping farmers. The AI research is exploring various aspects of farming, most notable areas being automated intelligent irrigation and detecting crop diseases. Intelligent machine learning algorithms are helping farmers automate and optimize irrigation. It reduces water wastage and optimizes water supply to the crops. AI-driven system can analyze soil, weather and crop type to deliver optimized water supply. Input and Output: Detecting bacterial disease using image processing in beans leaf. An app-based startup is helping farmers detect plant diseases using image recognition technology. It leverages the power of crowd-sourced data of images for identification, prevention, and treatment. Currently, it can identify 60 plant diseases with more than 90% accuracy. Both these numbers should increase with more crowd-sourced data getting fed to the app. Plantix Sports Analytics Nowadays, players and coaches heavily analyze their performance. The analysis of competitors is another important area coaches and managers deal with. There is a lot of data available to capture, analyze and predict the performance. This is where machine learning is making a mark in sports analytics. The huge amount of data sets likes fitness, health, strength, speed, ability, skills, match-day conditions et cetera makes machine learning algorithm an ideal fit. The algorithms can help stakeholders predict the performance metrics of their players and competitors at an individual level. This is getting achieved by learning from historical data and clubbing it with real-time analysis of the game which was not possible by mere human cognitive abilities. is one such company which is helping ice hockey teams analyze data by leveraging AI. They claim to give real-time insights accurate to 0.1 seconds and 500 player stats. Who knows, in future, AI might help us predict outcomes and results of a match. It will be interesting to see how deeply can AI go in analyzing and predicting stats for a sports team/athlete. Iceberg Entertainment On 9th of June 2016, a short movie named was released. It was a movie written by Benjamin, an intelligent automatic screenwriter. Benjamin was trained using an . Apart from screenwriting, AI is being used in Direction, Cinematography and other aspects of movie production. For example, check out this trailer of the AI thriller movie named Morgan, . Notably, AI boiled down the time taken to complete the overall process of trailer creation from approx 30 days to 24 hours. Sunspring LSTM RNN created by IBM’s Watson Morgan’s trailer: Created by IBM Watson The applications of AI in entertainment industries are not limited to movies. Art and music verticals are also getting inspired by AI. , is a music album completely composed by Melomics109, a computer cluster at Universidad de Málaga, Spain. 0music With computational and cognitive abilities of machine learning algorithms increasing day by day, we will see much more intervention of AI in entertainment industries. Maybe in upcoming future years from now, AI could replace a musician or cameraman with average skills, but it would still be no match for the best artists in the business. After all, generating original thoughts in a subjective domain is not something AI has been proven to achieve. Education Educational Data Mining is among the new and emerging applications of AI in education. Generally, schools opt a single approach to teach students. This approach has its downsides as not every student has similar learning capabilities. Also, it is difficult for teachers to carefully assess the performance of every individual student. Nowadays, algorithms can assess and analyze the performance of students at an individual level. Students can learn through visual content and every activity can be tracked to analyze the learning performance of an individual student. and Learning are two such companies who are using advanced analytics in education. Carnegie Learning Thinkster Math Education can also be benefited by advanced machine algorithm by taking learning out of the limitations of classrooms. AI can assist learners with self-direction, self-assessment, teamwork and more. Also, AI driven collaborative learning platforms are coming up. , a collaborative learning platform, is exploring the power of AI to enhance social learning. Brainly Protecting Oceans AI-powered robots are detecting pollution levels and tracking changes in temperature and pH of the oceans much better than humans. These autonomous robots can collect a huge amount of data from multiple geographies. This data is analyzed and predictions are made. Action plans to curb such changes in the ecosystem can then be executed more accurately. Another important application of AI in protecting oceans is to control invasive marine animals. AI is helping robots become more precise, effective, and constant by using Machine Learning and Computer Vision to track invasive species and eliminate them. Notably, such operations are already taking place to eliminate Lionfish from Atlantic ocean and Crown-of-thorns Starfish from Great Barrier Reef. Invasive species are capable of causing extinctions of native plants and animals, reducing biodiversity, competing with native organisms for limited resources, and altering habitats. When Lionfish expanded to Atlantic from their native Indo-pacific region, they flourished due to the lack of a natural predator. Thus causing harm to local Atlantic species. There are more emerging applications of AI in areas never explored before. Healthcare is getting better diagnosis and prevention solutions powered by AI. Also, logistics industries are eyeing fully automated warehouses using intelligent robots. There are tons of industries that can leverage the power of AI. People consider AI’s impact on industries to be similar to what electricity did when it was invented in the early 20th century. Many tech veterans have said that Machine Intelligence is going to be the biggest disruptive force in the future, but I will leave you with a piece of advice given by our favorite visionary, Bill Gates, to the young college crowd: AI, energy, and biosciences are promising fields where you can make a huge impact. It’s what I would do if starting out today. – Bill Gates, Founder of Microsoft As a member of an AI research group, I often come across talented people working on niche and interesting problems of our world. If you know of any other interesting and niche applications, free to comment below and I will be happy to append the list. Note: Photo Credits for most of the images used in the blog: Pexels I work for , an applied AI research group. We develop AI powered solutions for real world problems. Paralleldots If you liked reading the article, please clap it, share it or comment below for further discussions.

STARFISH SPACE

starfishspace.com

-- employees

Since n.d.

COMPANY RANKING

STARFISH SPACE

EVERGREEN INDEX #602

Representing this company? Click here to claim and customize this page!

Starfish Space WIKI

Apply to get your Tech Company News Page live on HackerNoon.com today!

STARFISH SPACE

starfishspace.com

-- employees

Since n.d.

COMPANY RANKING

STARFISH SPACE

EVERGREEN INDEX #602

RELATED COMPANIESRANK

STORIES ABOUT STARFISH SPACE

Representing this company? Click here to claim and customize this page!

Starfish Space WIKI

Apply to get your Tech Company News Page live on HackerNoon.com today!