Podcasts are exploding in popularity, and we are on a mission to make all of this audio both readable and searchable!\n\nWe rely on proprietary transcription tech, the latest in machine learning, and scalable data pipelines (serverless, containers, etc) to convert and analyze thousands of hours of audio content released each day

Podscribe

Our definition of 'goal setting' isn't what it used to be. Somewhere along the line, we got it twisted – and now, kids who have barely got their driver's license are expected to plan their entire future. We carry this pressure into adulthood, too. I know plenty of people who have nailed down exactly where they're (allegedly) going to be in five years' time. Despite the supposed benefits of future planning, they're very stressed individuals. Here's what I think happens. Job titles, cities, and companies are all tangible things we can grasp with our minds. It's easy to say, "I plan to be a Senior Vice President of something or other at this company in five years" – so we hold on tight to those specifics. The problem is that, in prescribing a future for yourself, you leave little room for unexpected opportunities or a change of heart. You box yourself in when there was never a need to use boxes in the first place. Let me dig into this a little deeper – you'll see what I mean. Why We Prescribe I am forever haunted by the statistic that end up dropping out. The actual dropping out part doesn't scare me; it's the implication. 40 percent of US college students We feel so pressured to line up our ducks, to prescribe our futures, that the detours and roadblocks we encounter later on are seen as failures. Even as adults, why are we so hellbent on having a rigid five-year or ten-year plan? How on earth can we predict what we'll want, who we'll be, and where we'll be in a decade? If I made decisions now based on what I wanted ten years ago... well, things would be different. Let's put it that way. So, why do we do this? We're Conditioned This Way It's easy to see how rigid planning stems from childhood; we're raised with the ideas of career progression, degrees, and qualifications, and having defined career paths that lead logically from one place to the next. There's little talk of flexibility. At my school, you were either going to college, or you were lost. It stands to reason that we'd come away from this conditioning with a set-in-stone mindset: you choose your career now, and you plan it out to the nth degree. We Want A Sense of Security of Dr. Thema Bryant-Davis, a psych professor at Pepperdine University, "It’s a way of dealing with worry and can give people a false sense of control." In the words I can understand that. I've been there myself. When all else feels chaotic, there's a temptation to retreat into our comfort zone and make rigid plans for the future. We don't want any surprises – we just want what we expect. Of course, we know deep down that life has other ideas. Planning our lives seems like the best way to combat unexpected changes... but in reality, it's like locking our knees before hitting the ground. We make ourselves so inflexible that even a short fall leaves us with a broken bone. Family and Societal Expectations It's easy to start prescribing your life as a way to explain or justify yourself to others. If you're not in a stable job at the moment, for instance, maybe you'll create a plan for the future to soothe family members, friends, and peers. But again, this comes with an expiration date. We may be able to fool others for a little while – but we can't keep up that façade forever. It's human nature to change our minds and switch paths; trying to stick it out despite feeling like you're going in the wrong direction is counterintuitive at best, and damaging at worst. But Isn't Planning Helpful, Usually? Great point – yes, it is. Planning is both helpful and essential to getting the life you want. But there is a difference between planning and prescribing yourself to a certain path. Planning involves setting goals, making budgets, and taking stock of what you already have. It's like mapping out the route to your destination, but knowing that you can change course if needed. I'd also say that mindful planning comes from your values and passions. Instead of looking at a career progression chart for your area of study, you should plan your future around the things you know bring you the most joy. Why? Because typically, your values and passions are the things that don't change. The Psychology of Open-Ended Planning I frequently think about this quote by one of the best philosophers of all time, Ludwig Wittgenstein: “A man will be imprisoned in a room with a door that’s unlocked and opens inwards; as long as it does not occur to him to pull rather than push." It's hard to know exactly what he's referring to here. The way I interpret it, though, is that doors to new opportunities will remain closed if you don't realize they're able to be opened. This is what I believe happens to us when we plan our lives out too rigidly. We end up clueless as to the possibilities lying just outside our peripheral – because why would we bother looking when everything is already decided? Open-ended planning sidesteps this problem by actively putting questions in the air. What could you be doing in five years' time? What could your career path look like? What new opportunities can you take advantage of in the meantime to explore your passions? You acknowledge the infinite options available to you, and therefore, it does occur to you to "pull rather than push." Values-Based Planning 101 Rounding up everything I've said so far, here's my hypothesis: we need to be engaging in values-based planning rather than prescriptive planning. There's a clear distinction between the two. With prescriptive planning, you are largely focused on the way your future looks and sounds. What's your job title? What do you earn? In values-based planning, you're focusing on how your future feels. Knowing what you respond to emotionally, and knowing where your passions and values lie, you're able to construct a picture of the future that will gel with your personality regardless of where you end up. A rigid, prescriptive, plan might be, "I want to finish this degree, get a job in X field and work my way up the ladder within Y company." A flexible, open-ended five-year plan might be, "I want to end up in a position where I'm working with people on an individual level and helping them to solve problems." That could mean a lot of different things – so you can take the time to explore and find out what that looks like in practice. You can see how one plan is much more open to unexpected opportunities and changes, while the other has a much narrower path. Yes, the more ambiguous option feels a little intimidating – but that's the beauty of it. A little uncertainty is a small price to pay for flexibility and uncapped potential. When Prescriptive Planning Helps By all means, there are certain situations where making a rigid timeline can be your saving grace. For instance, you might be working toward a 10-mile run in six months' time. You might have a specific career path in mind that requires certain qualifications, or you may be working toward a promotion at your current job – in which case you'll need to plan out the steps to get there and make sure they're achievable within a certain timeframe. In these cases, planning a rigid timeline is your roadmap – it keeps you on the path and provides a sense of security in the face of uncertainty. As soon as you start talking about five or ten years down the track, though, you're essentially making plans for someone you don't even know. That person could have completely different wants and needs than you do now. By all means, keep your goals in sight – but make sure you're open to the unexpected and embrace opportunities as they come. That's how real growth happens. Wrap Up It's easy to get caught up in what feels like a logical plan; life rarely works out the way we expect it to. (In fact, it usually deviates in very drastic ways). That doesn't mean you should abandon all hope for progress – just that you should always be ready to adapt when necessary. Center yourself around the things that ignite your passions, and don't put a low ceiling over your head as a self-inflicted cap on your potential. I'd love to hear from you if anything here resonated, so feel free to respond by email or add a comment below. Until next time! If you enjoyed this article, I’d love to hear from you. Reply to this or at me and I'll do my best to get back to everyone! email tweet @ScottDClary

It's Important to Plan for the Future, But Don’t Prescribe It

Since the wave of digitization is sweeping businesses, managing an enterprise has become a much easier task. Enterprises now utilize applications or software like Customer Relationship Management or CRM, project management applications, ERP, and more to carry out several business-critical tasks. However, the widespread use of these applications may also come at a cost as data breaches have been becoming more common since 2020. For this reason, Identity and access management or IAM architects have shed light on the importance of enterprise application security. Vulnerabilities in an Enterprise Application Enterprises are always on the lookout for opportunities to drive digital transformation. They do so by introducing certain technological innovations. For example, the creation of enterprise IoT and Cloud computing, and more. Although these innovations have significantly contributed to the digitization of an enterprise and its subsequent growth, they can also be an entry point for hackers and malware. These entry points are known as vulnerabilities in an application. Let’s take a look at a few of the most common ones found in applications. Lack of Access Control A surprising number of data breaches take place internally. That is, the employees working within the company may consciously or unconsciously break access protocol. It can also occur when user restrictions are not implemented properly. Therefore, a lack of control over who accesses what in the application is considered to be a vulnerability. Exposure of Sensitive Data This can be a big problem for the enterprises that collect extensive information about their employees and their customers. This involves information like addresses, , account details, etc. The databases that house this information are among the most common targets for cyberattacks and may result in the unauthorized distribution of sensitive data. passwords Non-functioning Authentication Authentication is a measure that is present in every enterprise application. This security measure is essential as it plays a vital role in the aforementioned user restriction. Applications usually utilize the security along with the session management application functions to ensure security. security two-factor authentication Security Misconfigurations Finally, applications could also suffer from security misconfigurations. These can be due to various reasons like default configurations that are not secure, misconfigured HTTP headers, or incomplete configurations. Tools that Promote Enterprise Application Security At present, the tools that an enterprise uses to ensure the security of their applications can be categorized into two types. Firstly, there are the tools that test the existing security measures put in place in the application. The other type of tool is the one that fortifies or adds to the security measures. 1. Types of testing tools The testing is carried out using the following tools: Static application testing: This tool essentially reads the code that the application is created from to ensure no mistakes that could lead to security gaps. Dynamic application testing: These tools detect gaps in the code while it is running. To a developer, this kind of testing mimics a cyberattack so that the developers can work towards reinforcing the code. Hence, the name dynamic. 2.  Variations of Fortifying Tools Under the fortifying tools, one will find the following: RASP: The Runtime Application Self-protection tool acts as a tester and a fortifier simultaneously. It does so by monitoring the environments present on the desktop and even mobile devices. Anti-Tampering software: This software functions to prevent cybercriminals from obtaining access to the code of an application. Threat detector: This tool is usually used by professionals to ensure that the environment in which the application is running is safe. Therefore, aspects like the state of the environment, potential threats, and compromised devices can be found. Benefits of Implementing Enterprise Application Security From the aforementioned points, it becomes clear that an enterprise application with the required security measures offers an unprecedented sense of safety to the users of the applications. In addition to this, other benefits make application security desirable to IAM architects and enterprises alike. These benefits include: Real-time detection: Cyberattacks are usually detected quite quickly, given the number of security measures that are put into place. Some measures also detect high-risk activity through behavior analysis. The detection can also be automated in some cases. Versatile integration: The security measures are compatible with multiple applications, irrespective of whether it is a cloud platform or a SaaS platform. Compliance monitoring: The testing tools mentioned before will be used to ensure that the application complies with the safe practices and regulations of data security. These tools can also monitor the critical infrastructure and prevent possible data loss. Enterprise application security (EAS) is transforming the information technology landscape. As the pace of digital business accelerates and the complexity of enterprise IT grows, regulatory requirements are demanding more serious attention. No longer a “nice-to-have”, many boardrooms now acknowledge that is essential to securing infrastructure, data, and applications in today’s digital economy. enterprise application security

Why IAM Architects Prescribe Enterprise Application Security

Testing the Extents of AI Empathy: A Nightmare Scenario

এআই সহানুভূতির পরিধি পরীক্ষা করা: একটি দুঃস্বপ্নের দৃশ্য

Testen des Ausmaßes der KI-Empathie: Ein Albtraumszenario

Probando el alcance de la empatía con la IA: un escenario de pesadilla

Tester l’étendue de l’empathie de l’IA : un scénario de cauchemar

एआई सहानुभूति के विस्तार का परीक्षण: एक दुःस्वप्न परिदृश्य

AI 共感の程度をテストする: 悪夢のシナリオ

AI 공감의 범위 테스트: 악몽 시나리오

Testando a extensão da empatia com IA: um cenário de pesadelo

Проверка степени эмпатии ИИ: кошмарный сценарий

Yapay Zeka Empatisinin Kapsamlarını Test Etmek: Bir Kabus Senaryosu

Kiểm tra mức độ đồng cảm của AI: Kịch bản ác mộng

测试人工智能同理心的程度：噩梦场景

Metaphysics and Mathematics 2: The Complex Space of Prime Numbers

United States v. Apple INC Court Filing, retrieved on March 21, 2024 is part of . You can jump to any part in this filing . This part is 5 of 25. HackerNoon’s Legal PDF Series here B. Apple invited third-party investment on the iPhone and then imposed tight controls on app creation and app distribution 41. Apple controls how developers distribute and create apps for iPhone users. For example, developers can only distribute native iPhone apps through Apple’s App Store, which is the only way for users to download native iOS apps. Limiting distribution to the Apple App Store enables Apple to exert monopoly power over developers by imposing contractual restrictions and rules that limit the behavior of non-Apple apps and services. Specifically, Apple sets the conditions for apps it allows on the Apple App Store through its App Store Review Guidelines. Under these guidelines, Apple has sole discretion to review and approve all apps and app updates. Apple selectively exercises that discretion to its own benefit, deviating from or changing its guidelines when it suits Apple’s interests and allowing Apple executives to control app reviews and decide whether to approve individual apps or updates. Apple often enforces its App Store rules arbitrarily. And it frequently uses App Store rules and restrictions to penalize and restrict developers that take advantage of technologies that threaten to disrupt, disintermediate, compete with, or erode Apple’s monopoly power. 42. Apple also controls app creation by deciding which APIs are available to developers when they make third-party apps. For example, developers cannot provide native apps on the iPhone unless they enter into Apple’s non-negotiable Developer Program License Agreement (DPLA). That agreement requires developers to use public APIs only “in the manner prescribed by Apple.” It also prohibits third-party apps from using APIs that Apple designates as “private.” Apple selectively designates APIs as public or private to benefit Apple, limiting the functionality developers can offer to iPhone users even when the same functionality is available in Apple’s own apps, or even select third-party apps. Similar to Apple’s App Store restrictions, Apple uses its DPLA to impose restrictions that penalize and restrict developers that take advantage of technologies that threaten to disrupt, disintermediate, compete with, or erode Apple’s monopoly power. 43. Developers cannot avoid Apple’s control of app distribution and app creation by making web apps—apps created using standard programming languages for web-based content and available over the internet—as an alternative to native apps. Many iPhone users do not look for or know how to find web apps, causing web apps to constitute only a small fraction of app usage. Apple recognizes that web apps are not a good alternative to native apps for developers. As one Apple executive acknowledged, “[d]evelopers can’t make much money on the web.” Regardless, Apple can still control the functionality of web apps because Apple requires all web browsers on the iPhone to use WebKit, Apple’s browser engine—the key software components that third-party browsers use to display web content. 44. Nor can developers rely on alternative app stores even though this would benefit developers and users. For example, developers cannot offer iPhone users an app store that only offers apps curated for use by children, which would provide opportunities to improve privacy, security, and child safety. By contrast, Apple allows certain enterprise and public sector customers to offer versions of app stores with more curated apps to better protect privacy and security. 45. Apple’s control over both app distribution and app creation gives Apple tremendous power. For example, Apple designates as “private” the APIs needed to send Short Message Service, or SMS, text messages, which is a protocol used by mobile carriers since the early 1990s to allow users to send basic text messages to other mobile phone numbers using their own mobile phone numbers. Developers have no technical means to access these private APIs, but even if they did, doing so would breach their developer agreement with Apple, and therefore put the developer at risk of losing the ability to distribute apps through the App Store. For example, Apple prohibits third-party iPhone apps from sending or receiving SMS text messages even though this functionality is available through Apple Messages. Likewise, Apple can control the functionality of third-party apps and accessories through its control of app distribution because if an app includes functionality that Apple does not like, Apple can and does exercise its discretion to simply block the app from the App Store. 46. Apple’s dominance is such that neither app developers nor iPhone users can benefit from lower cost or higher quality means of distributing apps or purchasing and providing digital products and services. Instead, Apple guarantees that it continues to benefit from the contributions of third-party developers and other platform participants while also protecting itself from the competitive threats and pressure those participants pose to Apple’s smartphone monopoly. 47. This complaint focuses on Apple’s use of its dominance to impose contracts and rules that restrict the behavior and design decisions of companies other than Apple. Continue Reading . Here About HackerNoon Legal PDF Series: We bring you the most important technical and insightful public domain court case filings. This court case retrieved on March 21, 2024, from is part of the public domain. The court-created documents are works of the federal government, and under copyright law, are automatically placed in the public domain and may be shared without legal restriction. justice.gov

Inside Apple's Stranglehold on App Development and How it Shapes Consumer Choices

USA v. Samuel Bankman-Fried Court Filing, retrieved on March 15, 2024 is part of . You can jump to any part in this filing . This part is 26 of 33. HackerNoon’s Legal PDF Series here III. A Very Significant Sentence Is Necessary for General Deterrence A very significant sentence of imprisonment is also necessary in this case to afford adequate general deterrence to criminal conduct. See 18 U.S.C. § 3553(a)(2)(B). “Considerations of (general) deterrence argue for punishing more heavily those offenses that either are lucrative or are difficult to detect and punish, since both attributes go to increase the expedited benefits of a crime and hence the punishment required to deter it.” United States v. Heffernan, 43 F.3d 1144, 1149 (7th Cir. 1994); see also United States v. Martin, 455 F.3d 1227, 1240 (11th Cir. 2006) (“Because economic and fraud-based crimes are more rational, cool, and calculated than sudden crimes of passion or opportunity, these crimes are prime candidates for general deterrence.” (quotation marks and citation omitted)); United States v. Mueffelman, 470 F.3d 33, 40 (1st Cir. 2006) (deterrence of white-collar crime is “of central concern to Congress”). Here, a sentence of 40 to 50 years is necessary to serve that purpose. The enormity and audacity of Bankman-Fried’s crimes has understandably captured the attention of both the public, generally, and those involved in cryptocurrency or finance specifically. Thus, the deterrent message and effect of the sentence imposed by the Court in this case will resonate significantly with any individual or institution tempted to engage in conduct similar to that of the defendant. This need for deterrence is especially important in the area of cryptocurrency, where some individuals have operated under the misimpression that they are unregulated, not subject to criminal laws, or can avoid scrutiny or significant jail time. Bankman-Fried contends that the loss of his reputation and finances is sufficient to deter him (and others who have seen what has happened) from further misconduct. (Def. Mem. at 66- 69). But that is simply not the case. As Judge Swain cogently explained in sentencing Peter Madoff: The Court is also very much aware that [the defendant] and his family have lost the assets, reputation, financial security and social position that they had enjoyed.… That punishment while indisputably severe is not alone sufficient to effect a punishment that is just, given the nature, magnitude and context of the crimes for which [the defendant] is being sentenced today. Many people and institutions have suffered the same fall from wealth, privilege and in many cases even basic levels of financial security as a result of the massive fraud. Ordinary wage-earning people who had planned for ordinary comfortable retirements are among the victims here too…. Proper punishment also requires a lengthy term of imprisonment, the punitive loss of [the defendant’s] personal freedom. Such a sentence is also necessary to serve the deterrence function of sentencing…. the world in which [the defendant’s] conduct took place is full of many temptations and opportunities to breach trust relationships for personal gain. The consequences of such behavior must be harsh to help deter others from taking the path of dishonesty and theft. United States v. Peter Madoff, No. 10 Cr. 228 (LTS) (S.D.N.Y. Dec. 20, 2012) (Sent. Tr. at 28- 29). That reasoning applies with greater force to this defendant. The studies cited by the defendant, on the other hand, are wrong, or at least do not apply here. (Def. Mem. at 66). The prospect of punishment, and not the severity, may be enough in some cases to deter. But not here. This case presents an appropriate opportunity for the Court to send a strong signal to market participants. See, e.g., United States v. Ulbricht, 858 F.3d 71, 94 (2d Cir. 2017) (affirming district court’s sentence which took into account, among other things, general deterrence and the fact that the sentence imposed “could have a powerful general deterrent effect because the case had attracted an unusually large amount of publicity”). A sentence of the length proposed by the defendant would have the perversely opposite effect: it would send the message that “the defendant’s crime is not so serious,” “the punishment is not so severe,” “it is not like the major frauds of the past.” Especially in light of the publicity to which the defendant points, an unjustifiably lenient sentence would encourage others who might be tempted to engage in fraud that the rewards of such wrongful activity may be worth the price. Bankman-Fried also argues that the negative publicity from the prosecution is enough of a deterrent. But simply because a case has been widely reported on and drawn notoriety does not suggest that imprisonment is not required to promote deterrence. As the Second Circuit has explained: [T]he circumstances referred to by the district court do not constitute punishment. The public nature of criminal prosecutions is part of our constitutional fabric; the public humiliation suffered by one prosecuted and convicted of a crime is an ordinary consequence of his conduct, not a condition imposed by the criminal codes or the judicial process. These circumstances, though adverse, are not what § 3553(a)(2)(A) means by “punishment.” Hence they cannot properly be viewed as fulfilling the need for the imposition of just punishment. And given that the more massive a fraud, the more likely it is that the prosecution will generate publicity, the logical extension of the district court's view—i.e., that Freedman’s public humiliation and the public nature of his prosecution were punishment enough—would mean that the more flagrant the crime, the less actual statutorily prescribed “punishment” it would require. And of course, the less punishment that is meted out, the less deterrent effect the sentence will have on others contemplating similar crimes. United States v. Cutler, 520 F.3d 136, 171 (2d Cir. 2008). Here, general deterrence is of the utmost important in this case, and that is why a lengthy term of imprisonment in line with other defendants who embezzled hundreds of millions or billions in customer funds is necessary. Continue Reading . Here About HackerNoon Legal PDF Series: We bring you the most important technical and insightful public domain court case filings. This court case retrieved on March 15, 2024, from is part of the public domain. The court-created documents are works of the federal government, and under copyright law, are automatically placed in the public domain and may be shared without legal restriction. storage.courtlistener

SBF lengthy Jail Time to Serve as Financial Crime Deterrent 

This paper is available on arxiv under CC 4.0 license. Authors: (1) Ali Ghanbari, Dept. of Computer Science, Iowa State University; (2) Deepak-George Thomas, Dept. of Computer Science, Iowa State University; (3) Muhammad Arbab Arshad, Dept. of Computer Science, Iowa State University; (4) Hridesh Rajan, Dept. of Computer Science, Iowa State University. Table of Links Abstract & Introduction Background Motivating Example Proposed Approach Supported DNN Bugs Evaluation Threats to Validity Related Work Conclusion, Acknowledgments & References VI. EVALUATION We evaluate deepmufl and compare it to state-of-the-art static and dynamic DNN fault localization techniques, by investigating the following research questions (RQs). • RQ1 (Effectiveness): How does deepmufl compare to state-of-the-art tools in terms of the number of bugs detected? How many bugs does deepmufl detect from each subcategory of model bugs in our dataset and how does that compare to state-of-the-art tools? What are the overlap of detected bugs among deepmufl and other fault localization techniques? • RQ2 (Efficiency): What is the impact of mutation selection on the effectiveness and efficiency of deepmufl? How does deepmufl compare to state-of-the-art tools in terms of end-to-end fault localization time? A. Dataset of DNN Bugs To evaluate deepmufl and compare it to state-of-the-art DNN fault localization techniques, we queried StackOverflow Q&A website for posts about Keras that had at least one accepted answer. Details about the SQL query used to obtain the initial list of posts is available online [36]. The query resulted in 8,412 posts that we manually sieved through to find the programs with model bugs. Specifically, we kept the bugs that satisfied the following conditions. • Implemented using Sequential API of Keras, • The bug in the program was a model bug supported by deepmufl as described in §V, and • The bug either had training datatset available in the post in some form (e.g., hard-coded, clearly described in the body of the post, or a link to the actual data was provided) or we could see the described error using synthetic data obtained from scikit-learn’s dataset generation API. This resulted in 102 bugs and we paired each bug with a fix obtained from the accepted answer to the question. We further added 7 bugs from DeepLocalize dataset [11] that are also coming from StackOverflow and we paired these bugs also with their fixes that are obtained from the most up-voted answer. Thus, we ended up with 109 bugs in total. To the best of our knowledge, this is the largest dataset of model bugs obtained from StackOverflow and it overlaps with the existing DNN bug datasets from previous research [12], [56]. Our bug dataset contains 85 classifiers (45 fully-connected DNNs, 29 CNNs, and 11 RNNs) and 24 regression models (19 fullyconnected DNNs, 3 CNNs, and 2 RNNs). And each category has at least one example of model bugs. Therefore, we believe that our dataset is greatly representative of model bugs, i.e., the bugs supported by deepmufl (and other tools that support this type of bugs), as we have examples of each sub-category of bug in various locations of the models for various regression and classification tasks. After loading the training dataset for the bugs, we fitted the buggy models three times and stored them in .h5 file format separately. The repetition was conducted to take randomness in training into account. Randomness in data generation was mitigated by using deterministic random seeds. For fault localization purposes, we used the test dataset, and if it was not available, we used the training dataset itself. When we had to use synthesized data points, we deterministically splitted the generated set of data into training and testing datasets. B. Baseline Approaches and Measures of Effectiveness In RQ1 and RQ2, we compare five different configurations of deepmufl to recent static and dynamic DNN fault localization tools. The five configurations of deepmufl are as follows. • [30]: In this setting, we use the Metallaxis formula to calculate suspiciousness values of model elements. Metallaxis, by default, uses SBI [46] to calculate suspiciousness values for individual mutants. A recent study [57] provides empirical evidence on the superiority of Ochiai [47] over SBI when used within Metallaxis formula. Thus, we considered the following four combinations: type 1 impact: (1) SBI formula (i.e., Eq. 1); (2) Ochiai formula (i.e., Eq. 2), and type 2 impact: (3) SBI formula (i.e., Eq. 1); (4) Ochiai formula (i.e., Eq. 2). Metallaxis • [17]: We used the default formula of MUSE to calculate the suspiciousness of model elements. For this, only type 1 impact is considered, as the heuristics behind MUSE are defined based on type 1 impact. MUSE Our technique follows a more traditional way of reporting root causes for the bugs [30], [17], [57], [19], [22], [21], [15], in that it reports a list of potential root causes ranked based on the likelihood of being responsible for the bug. This allows the users find the bugs faster and spend less time reading through the fault localization report, which in turn increases practicality of the technique [58]. We have used top-N, with N = 1, metric to measure the effectiveness of deepmufl in RQ1 and RQ2. Specifically, if the numbers of any of the buggy layers of the bug appeared in the first place in the output of deepmufl, we reported it as detected, otherwise we marked the bug as not-detected. We emphasize that top-1 metric gives a strong evidence on the effectiveness of deepmufl, as the developers usually only inspect top-ranked elements, e.g., over 70% of the developers only check top-5 ranked elements [59]. Our selection criteria for the studied fault localization techniques are: (1) availability; (2) reproducibility of the results reported in the original papers, so as to have a level of confidence on the correctness of the results reported here; and (3) support for model bugs in our dataset, so that we can make a meaningful comparison to deepmufl. Below we give a brief description of each of the selected tools, why we believe they support model bugs, and how we have interpreted their outputs in our experiments, i.e., when we regard a bug being detected by the tool. A static fault localization tool that uses 23 rules to detect faults and design inefficiencies in the model. Each rule is associated with a set of rules of thumb to fix the bug that are shown to the user in case the precondition for any of the rules are satisfied. The five rules described in Section 4.2.1 of the paper target model bugs. Neuralint produces outputs of the form [Layer L ==> MSG] ∗ , where L is the suspicious layer number, and MSG is a description of the detected issue and/or suggestion on how to fix the problem. A bug is deemed detected by this tool if it is located in the layer mentioned in the output message or the messages describe any of the root causes of the bug. 1) Neuralint: A dynamic fault localization technique that detects numerical errors during model training. One of three rules described in Section III.D of the paper checks model bugs related to wrong activation function. DeepLocalize produces a single message of the form Batch B Layer L : MSG, where B is the batch number wherein the symptom is detected and L and MSG are the same as we described for Neuralint. A bug is deemed detected if it is located in the layer mentioned in the output message or the message describes any of the root causes of the bug. 2) DeepLocalize: A tool similar to DeepLocalize, but with more bug pattern rules and a decision procedure to give actionable fix suggestions to the users based on the observations. All 8 rules in Table 2 of the paper monitor the symptoms of model bugs. Similar to DeepLocalize, DeepDiagnosis produces a single message of the form Batch B Layer L : MSG1 [OR MSG2], where B and L are the same as described in DeepLocalize and MSG1 and MSG2 are two alternative solutions that the tool might suggest to fix the detected problem. A bug is deemed detected if it is located in the layer mentioned in the output message or the message describes any of the root causes of the bug. 3) DeepDiagnosis: : A hybrid, i.e., a combination of static and dynamic, technique that works by applying heuristic static checks on, and injecting dynamic checks in, the program, parameters, model structure, and model behavior. Violated checks raise error flags which are propagated to a web-based interface that uses visualizations, tutorial explanations, and code snippets to help users find and fix detected errors in their code. All three rules described in Section 5.2 of the paper target model bugs. The tool generates outputs of the form [< MSG1 > · · · < MSGm >] ∗ , where m > 0 and MSGi is a description of the problem detected by the tool. A bug is deemed detected if any of the messages match the fix prescribed by the ground-truth. 4) UMLAUT C. Results To answer RQ1, we ran deepmufl (using its five configurations) and four other tools on the 109 bugs in our benchmark. We refer the reader to the repository [36] for the raw data about which bug is detected by which tool, and here we describe the summaries and provide insights. At top-1, deepmufl detects 42, 47, 26, 37, and 53 bugs using its Metallaxis SBI + Type 1, Metallaxis Ochiai + Type 1, Metallaxis SBI + Type 2, Metallaxis Ochiai + Type 2, and MUSE, respectively, configurations. Meanwhile Neuralint, DeepLocalize, DeepDiagnosis, and UMLAUT detect 21, 26, 30, and 25, respectively, bugs. Therefore, as far as the number of bugs detected by each technique is concerned, MUSE configuration of deepmufl is the most effective configuration of deepmufl, significantly outperforming studied techniques, and Metallaxis Ochiai + Type 2 is the least effective one, outperformed by DeepDiagnosis. An empirical study [57], which uses a specific dataset of traditional buggy programs, concludes that Metallaxis Ochiai + Type 2 is the most effective configuration for MBFL. Meanwhile, our results for DNNs corroborates the theoretical results by Shin and Bae [60], i.e., we provide empirical evidence that in the context of DNNs MUSE is the most effective MBFL approach. Table 3 reports more details and insights on the numbers discussed above. Specifically, it reports the number of bugs detected by each configuration of deepmufl an four other studied tools from each sub-category of model bugs present in our dataset of bugs. As we can see from the upper half of the table, MUSE is most effective in detecting bugs related to activation function (SC1), bugs related to model type/properties (SC2), and wrong/redundant/missing layer (SC4), while Metallaxis Ochiai + Type 1 configuration outperforms other configurations in detecting bugs related to layer properties (SC3). Similarly, from bottom half of the table, we can see that other tools are also quite effective in detecting bugs related to activation function, with DeepDiagnosis being the most effective one among others. We can also observe that UMLAUT has been the most effective tool in detecting bugs related to layer properties. As we can see, MUSE configuration of deepmufl is consistently more effective than other tools across all bug sub-categories. Table 4 provides further insights on the overlap of bugs detected by each variant of deepmufl and those detected by the other four tools. Each value in row r and column c of this table, where 2 ≤ r ≤ 5 and 2 ≤ c ≤ 6, denotes the percentage of bugs detected by the deepmufl variant corresponding to row r and tool corresponding to column c. The values inside the parenthesis are the actual number of bugs. For example, 8 out of 42, i.e., 19.05%, of the bugs detected by Metallaxis SBI + Type 1 configuration of deepmufl are also detected by DeepLocalize. The last column of the table reports same statistics, except for all four of the studied tools combined. As we can see from the table, 60.38% of the bugs detected by MUSE configuration of deepmufl are already detected by one of the four tools, yet it detects 21 (=53-32) bugs that are not detected by any other tools. This is because deepmufl approaches fault localization problem from a fundamentally different aspect giving it more flexibility. Specifically, instead of looking for conditions that trigger a set of hard-coded rules, indicating bug patterns, deepmufl breaks the model using a set of mutators to observe how different mutation impact the model behavior. Then by leveraging the heuristics underlying traditional MBFL techniques, it performs fault localization using the observed impacts on the model behavior. Listing 2 shows an example of a model bug that only deepmufl can detect. The problem with this regression model is that it does not output negative values, and given the fact that the dataset contains negative target values, the model achieves high MSE. Since this model does not result in any numerical errors during training, DeepLocalize does not issue any warning messages, and since MSE decreases and the model does not show any sign of erratic behavior DeepDiagnosis does not detect the bug. UMLAUT’s messages instruct adding softmax layer and checking validation accuracy which is clearly not related to the problem, because the bug is fixed by changing the activation function of the last layer to tanh and normalizing the output values. Lastly, Neuralint issues an error message regarding incorrect loss function which also seems to be a false positive. To answer RQ2, we ran deepmufl and the other four tools on a Dell workstation with Intel(R) Xeon(R) Gold 6138 CPU at 2.00 GHz, 330 GB RAM, 128 GB RAM disk, and Ubuntu 18.04.1 LTS and measured the time needed for model training as well as the MBFL process to complete. We repeated this process four times, and in each round of deepmufl’s execution, we randomly selected 100% (i.e., no selection), 75%, 50%, and 25% of the generated mutants for testing. Random mutation selection is a common method for reducing the overhead of mutation analysis [61], [35]. During random selection, we made sure that each layer receives at least one mutants, so that we do not mask any bug. The last row in Table 5 reports the average timing (of 3 runs) of MBFL in each round of mutation selection. The table also reports the impact of mutation selection on the number of bugs detected by each configuration of deepmufl. As we can see, in MUSE configuration of deepmufl, by using 50% of the mutants, one can halve the execution time and still detect 92.45% of the previously detected bugs. Therefore, mutation selection can be used as an effective way for curtailing MBFL time in DNNs. For a fair comparison of deepmufl to state-of-the-art fault localization tools in terms of efficiency, we need to take into account the fact that deepmufl requires a pre-trained model as its input. Thus, as far as the end-to-end fault localization time from an end-user’s perspective is concerned, we want to take into consideration the time needed to train the input model in addition the time needed to run deepmufl. With training time taken into account, deepmufl takes, on average, 1492.48, 1714.63, 1958.35, and 2192.4 seconds when we select 25%, 50%, 75%, and 100% of the generated mutants, respectively. We also emphasize that the time for DeepLocalize and DeepDiagnosis varied based on whether or not they found the bug. Given the fact that a user could terminate the fault localization process after a few epochs when they lose hope in finding bugs with these two tools, we report two average measurements for DeepLocalize and DeepDiagnosis: (1) average time irrespective of the fact that the tools succeed in finding the bug; (2) average time if the tools successfully finds the bug. Unlike these two tools, the time for Neuralint and UMLAUT does not change based on the fact that they detect a bug or not. DeepLocalize takes on average 1244.09 seconds and it takes on average 57.29 seconds when the tool successfully finds the bug. These numbers for DeepDiagnosis are 1510.71 and 11.05 seconds, respectively. Meanwhile, Neuralint and UMLAUT take on average 2.87 seconds and 1302.61 seconds to perform fault localization. D. Discussion It is important to note that while deepmufl outperforms state-of-the-art techniques in terms of the number of bugs detected in our dataset, it is not meant to replace them. Our dataset only covers a specific type of bugs, i.e., model bugs, while other studied techniques push the envelope by detecting bugs related to factors like learning rate and training data normalization, which are currently outside of deepmufl’s reach. We observed that combining all the techniques results in detecting 87 of the bugs in our dataset; exploring ways to combine various fault localization approaches by picking the right tool based on the characteristics of the bug is an interesting topic for future research. Moreover, depending on the applications and resource constraints, a user might prefer one tool over another. For example, although Neuralint might be limited by its static nature, e.g., it might not be able analyze models that use complex computed values and objects in their construction, it takes only few seconds for the tool to conduct fault localization. Thus, in some applications, e.g., online integration with IDEs, approaches like that of Neuralint might be the best choice A major source of overhead in an MBFL technique is related to the sheer number of mutants that the technique generates and tests [62], [61]. Sufficient mutator selection [63] is referred to the process of selecting a subset of mutators that achieve the same (or similar) effect, i.e., same or similar mutation score and same or similar number of detected bugs, but with smaller number of mutants generated and tested. For the mutators of Table 2, so far, we have not conducted any analysis on which mutators might be redundant, as a reliable mutator selection requires a larger dataset that we currently lack. We postpone this study as a future work. Combining fault localization tools can be conducted with the goal of improving efficiency. We see the opportunity in building faster, yet more effective, fault localization tools by predicting the likely right tool upfront for a given model or running tools one by one and moving on to the next tool if we have a level of confidence that the tool will not find the bug. We postpone this study for a future work. Lastly, we would like to emphasize that comparisons to the above-mentioned techniques in a dataset of bugs that deepmufl supports is fair, as the other tools are also designed to detect bugs in the category of model bugs. However, making these tools to perform better than this, would require augmenting their current rule-base with numerous new rules, yet adding new rules comes with the obligation of justifying the generality and rationale behind them, which might be a quite difficult undertaking. deepmufl, on the other hand, approaches the fault localization problem differently, allowing for more flexibility without the need for hard-coded rules.

Mutation-based Fault Localization of Deep Neural Networks: Evaluation

There is no doubt that is the most successful economic system to date among all those that have existed. Thanks to capitalism, we owe all the scientific and technological advancements that have allowed humanity to live at a much higher standard of living today than a few hundred years ago. capitalism The system owes its success to a simple rule (as mentioned by Harari in " "): be selfish. A fundamental principle of modern capitalist economics is that every participant is "rational". Their sole aim is to maximize profits and minimize costs. Sellers and buyers meet in the free market, guaranteeing that the system will function optimally. For instance, if a product costs $10, but could actually be produced for $5, a market gap is created, which someone will sooner or later fill to generate profit for themselves, thus the price of the product will eventually drop to $5. Capitalism, therefore, does not need to be regulated. It is best left alone, as it will regulate itself. This is what we call the of capitalism. A simple and elegant theory, right? Sapiens invisible hand Too bad it doesn’t work… It's important to recognize that capitalism is a self-organizing optimization system that, although not perfect, functions quite well. However, it's crucial to establish the system's boundaries for this to be the case. A simple example is slave labor. Since capitalism lacks morality, its sole aim is to maximize profit and optimize the system. Therefore, if a country allows slave labor and it results in the highest profit, then according to capitalist logic, slaves should be employed. There's nothing inherently wrong with this; capitalism itself doesn't become evil by doing so. It merely selects the most optimal solution within the given constraints. Thus, we should not expect the amoral invisible hand of capitalism to abolish slavery. This task falls to us, the people. We must set the boundaries within which capitalism can operate. For instance, employing slaves must be prohibited, as it is morally unacceptable. Fortunately, slavery is now prohibited in most places, but other challenges have taken its place. Currently, one of the biggest challenges is climate change and sustainability. As long as it is much cheaper to produce goods in an unsustainable manner, capitalist logic will utilize these methods. There is nothing strange about this. Capitalism isn't evil; this is simply its logic. If we don't want to overheat, drown in waste, or face wars due to some areas becoming uninhabitable, then we must set boundaries just as we did with slavery. We must prohibit unsustainable production methods and the use of unsustainable energy sources. Perhaps in 100 years, unsustainable production will seem as barbaric to people as slavery does to us. They will ask: How could humanity have allowed all this? But accurately delineating boundaries is just one of the problems; the free market doesn't operate entirely as prescribed in theory. In the market, we might observe the system working well between two egg vendors, but the situation is far from ideal when looking at large corporations. For instance, in the tech market, instead of many competitors, we see only a few dominant ones. In the realm of social networks, Facebook prevails; in search, it's Google, and so on. Although the market is theoretically open, in reality, it's nearly impossible for any new players to emerge. The reason for this phenomenon is quite simple: the "gravity" of large companies. Once a company grows sufficiently large, it can easily absorb the competition and smaller ones. Just consider the numerous acquisitions by or . Additionally, the services themselves have their gravity. People use Facebook because their friends are there. This is the , which becomes very hard to counteract beyond a certain size. Even if a much better social network were introduced, it would be very difficult to lure people away from Facebook. And if it ever did become a significant player, Facebook could easily acquire the company. Google Facebook network effect But we can see the same trend on a smaller scale, at the individual level as well. Along with the companies, the wealth of investors also grows. The rich become richer, the poor become poorer, and the gap between the two social strata widens. If our goal as humanity is to increase happiness on a global scale (for every individual), then in its current form, capitalism has failed to achieve this goal. Of course, let's not throw the baby out with the bathwater. Capitalism isn't a bad system, but it needs bug fixes, or rather, a significant upgrade. First and foremost, we need to reevaluate the boundaries of capitalism, which can only be done on a moral basis. Just as we abolished slavery on moral grounds, we must also prohibit the use of our descendants' resources in advance on moral grounds. We have reached a point where the living conditions of future generations are at risk. Consider the cruelty of forcing our children and grandchildren to live in an uninhabitable world, which is at least as cruel as slavery, if not worse. Yet, we don't feel the weight of this issue. The second issue, the problem of inequality, can only be addressed by a massive redistribution system. Although one of the foundational ideas of capitalism is that it should not be interfered with, and the free market along with the "invisible hand" will solve everything, it appears this doesn't work in practice. Due to the "gravitational" force of capital, the rich get richer and the poor get poorer, and the situation will only get worse if artificial intelligence takes away many people's jobs. What can help with this is a massive redistribution system. But what would such a system look like? A suitable tax system can be used to collect money, but who decides who gets the money collected in this way? There is a very simple answer to this, which is essentially the typical capitalist response as well: let the market decide. Simply distribute the money equally among the people, and they will then support the actors who are important to them with their spending. Meanwhile, no participant can grow too large because of the redistribution system. The basic logic of such a system is as simple as the capitalist "be selfish" logic. It doesn't require a complex apparatus, the whole process could be automated. This is the logic behind Philp Rosedale's (creator of SecondLife) system called . On the project's website, there's a simulation that demonstrates how social inequalities grow over time in a free market, and how the process can be regulated by a simple redistribution system. This system, where people receive an income on a universal basis, is called a . The system is so simple that it could even be automated. In a digital currency system (for example, based on blockchain), it would be possible for taxes to be automatically paid from every transaction. There would be no need for separate accounting and bookkeeping systems, as the tax system would be "built into" the money itself. The distribution of the basic income would also be part of the system, occurring automatically without human intervention. Just imagine how much human resources such an automated system would free up. FairShare universal basic income Rosedale's system is ingenious. The only point where I disagree with him is in his attempt to implement this money system as a form of community currency. I believe such a system cannot function as a community currency because only those who would gain more from the basic income than they lose through taxes would find it worthwhile to connect to such a system. Without net contributors, it would not be possible to cover the basic income, leading to the system's demise. As a community currency, I could better envision " ," which is reputation-based credit money. karma money This brings us to the final important conclusion: capitalism cannot be fixed with local bug fixes. The problem is systemic and thus requires a systemic solution. Convincing people to cycle to work or to separate their waste selectively won't significantly make the economy more sustainable. There needs to be substantial regulation of industries and companies, something only states are capable of. Similarly, it's impossible to locally implement a basic income system like the one Rosedale envisioned. This can be forced only by the state, and yes, it will hurt many wealthy individuals, but at the same time, it will reduce social inequalities, make the market function better, and last but not least, increase global well-being. But if this is the case, why don't states act? To fix the bugs in capitalism, we first need to fix the bugs in democracy. In current representative democracies, the state typically represents the interests of corporations. I'm not suggesting that elections are merely an illusion, as we genuinely can choose between value systems. However, it's typical for all sides to primarily serve corporate interests, with value choices coming secondary. This could only change if there were a much greater level of political awareness among people. It's important to separate your waste selectively, but it's even more important whom you vote for in the election. Of course, we would also need representatives who primarily represent the voters (as they should by default), and it would be also important to include people to participate more directly in decision-making. For instance, there are already technologies available that could organize on important issues, allowing everyone to comfortably vote from home via the Internet. Such voting could be initiated by anyone, and the representatives would then need to act according to the voting results. If people are dissatisfied with a representative's work, they could be recalled at any time. If we could fix the bugs in our democracy, we could fix the entire economic system through it. completely secure, anonymous voting As I mentioned at the beginning of the article, capitalism is still the best among all economic systems. However, the flaws that are encoded into the system are becoming increasingly visible. If we want to avoid collapse, it is essential that we upgrade the system, and hopefully, with Capitalism 2.0, we can make it through a few more centuries.

PODSCRIBE

podscribe.com

-- employees

Since n.d.

COMPANY RANKING

PODSCRIBE

EVERGREEN INDEX #8693

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

Podscribe WIKI

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

PODSCRIBE

podscribe.com

-- employees

Since n.d.

COMPANY RANKING

PODSCRIBE

EVERGREEN INDEX #8693

RELATED COMPANIESRANK

STORIES ABOUT PODSCRIBE

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

Podscribe WIKI

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