The Nonlinear Library

The Nonlinear Library

By The Nonlinear FundEducation
Download on the App Store

All The Nonlinear Library episodes

  • LW - I'm consistently overwhelmed by basic obligations. Are there any paradigm shifts or other rationality-based tips that would be helpful? by Benjamin Hendricks
    Welcome to The Nonlinear Library, where we use Text-to-Speech software to convert the best writing from the Rationalist and EA communities into audio. This is: I'm consistently overwhelmed by basic obligations. Are there any paradigm shifts or other rationality-based tips that would be helpful?, published by Benjamin Hendricks on July 21, 2023 on LessWrong.
    I often get panicky and stressed at the thought of the never-ending nature of simple tasks. Laundry and dishes will always pile up; food and other stocks will always need to be resupplied; bills, insurance, taxes, and other paperwork will always need to be redone; I will always need to work to get money; etc. All of these things seem to stress me out significantly more than they do other people, and the fact that I'll never truly be rid of them is almost terrifying - and has been since I realized it in my teens.
    When I've had difficulties with similar things in the past, I've been able to adapt by changing my perspective on the issue using "rationality" theories. For example, I used the theory of hyperbolic discounting and picoeconomics to change how I dealt with cravings and impulsive thoughts.
    Sometimes, rationality-based techniques can also help. Goal factoring (and aversion factoring) helped me to change my habits for the better.
    The problem is that I can't find anything similar to help with this issue - my difficulties with mundane responsibilities. I've heard of a few arguments/solutions that would help, but they seem insufficient:
    Everyone has these chores, so everyone has the tools to deal with it - I've heard this in a few forms, and it doesn't stand up to scrutiny. Just because an experience is common doesn't mean that everyone can do it. For an obvious example, walking is pretty common, but many disabled people are unable to do it. People with debilitating psychological issues may be unable to function without medicine - or at all. There's always a possibility that something about my brain makes my experience unusual, unpleasant, and - possibly - unfixable.
    Just trick yourself, fake it 'til you make it - A couple people have recommended that I just try to power through these chores, turning off my brain for a while. They suggest that if I do this enough times, I'll get desensitized to the stress and everything will be fine. Aside from the obvious silliness of "don't think about it," this hasn't worked so far. If I ignore the stressors, then I just end up getting stressed without noticing until I almost throw up.
    You're pushing yourself too hard, take time to relax - This makes a lot of sense, but doesn't seem to be the issue. I'm a full-time student with a lenient courseload. I have tons of free time, but the problems seem to persist. A logical counter-point is that it's not the quantity of free time that's the problem, but the quality. After several attempts at tweaking my recreation, there hasn't been any improvement. The problem seems to be with the stressors, not the relief.
    Change your thoughts, and everything will be alright - This one is an interesting one, because changing my thoughts is what I think will work. The problem is that the new thoughts people suggest aren't helpful. Changing "this is stressful" to "this isn't stressful" doesn't change the emotional response for me; I have to come up with a specific reason behind it. Hence my search for new paradigms.
    I'm eager to hear if anyone here has any thoughts - or better places I should ask. Thank you for your time.
    Thanks for listening. To help us out with The Nonlinear Library or to learn more, please visit nonlinear.org.
    4 min
  • EA - EA Survey 2022: How People Get Involved in EA by WillemSleegers
    Welcome to The Nonlinear Library, where we use Text-to-Speech software to convert the best writing from the Rationalist and EA communities into audio. This is: EA Survey 2022: How People Get Involved in EA, published by WillemSleegers on July 21, 2023 on The Effective Altruism Forum.
    Summary
    Personal contact (22.6%), 80,000 Hours (13.5%), and a book, article, or blog post (13.1%) are the most common sources where respondents first hear about EA.
    80,000 Hours, local or university EA groups, personal contacts, and podcasts have become more common as sources of where respondents first encounter EA.
    Facebook, Giving What We Can, LessWrong, Slate Star Codex / Astral Codex Ten, The Life You Can Save, and GiveWell have become less common.
    Respondents whose gender selection was 'woman', 'non-binary', or 'prefer to self-describe', were much more likely to have first heard of EA via a personal contact (30.2%) compared to respondents whose gender selection was 'man' (18.4%).
    80,000 Hours (58.0%), personal contact with EAs (44.0%), and EA groups (36.8%) are the most common factors important for getting involved in EA.
    80,000 Hours, EA Groups, and EAGx have been increasing in importance over the last years.
    EA Global, personal contact with EAs, and the online EA community saw a noticeable increase in importance for helping EAs get involved between 2020 and 2022.
    Personal contact with EAs, EA groups, the online EA community, EA Global, and EAGx stand out as being particularly important among highly engaged respondents for getting involved.
    Respondents who identified as non-white, as well as women, non-binary, and respondents who preferred to self-describe, were generally more likely to select factors involving social contact with EAs (e.g., EA group, EAGx) as important.
    Where do people first hear about EA?
    Personal contacts continue to be the most common place where people first hear about EA (22.6%), followed by 80,000 Hours (13.5%) and a book, article, or blog post (13.1%).
    Comparison across all years
    The plot below shows changes in where people report first hearing of EA across time (since we ran the first EA Survey in 2014).
    We generally observe that the following routes into EA have been increasing in importance over time:
    80,000 Hours
    Local or university EA groups
    Personal contacts
    Podcasts
    And the following sources have been decreasing in importance:
    Facebook
    Giving What We Can
    LessWrong
    Slate Star Codex / Astral Codex Ten
    The Life You Can Save
    GiveWell
    Comparison across cohorts
    Several of the patterns observed in the previous section are also observed when we look at where different cohorts of EA respondents first encountered EA. We see that more recent cohorts are more likely to have encountered EA via 80,000 Hours and podcasts, and are less likely to have encountered EA via Giving What We Can, LessWrong, and GiveWell. No clear cohort effects were observed for other sources. Note that the figure below omits categories with few observations (e.g., EA Global, EAGx).
    Further Details
    We asked respondents to provide further details about their responses, and provide a breakdown for some of the larger categories. Details of other categories are available on request.
    80,000 Hours
    The largest number of respondents who first heard of EA through 80,000 Hours reported doing so through an independent search, e.g., they were searching online for "ethical careers" and found 80,000 Hours. The second largest category was via the website (which is potentially closely related, i.e., contact with the website resulting from independent search).
    Relatively much smaller proportions mentioned reaching 80,000 Hours through other categories, including more active outreach (e.g., advertisements).
    Books, articles, and blogs
    A book was cited as the most common source of encountering EA when considering the category of books, articles, and blogs.
    Books
    Books by Peter Singer were by far the most frequently cited books, followed by Doing Good Better by Willia...
    12 min
  • LW - News : Biden-⁠Harris Administration Secures Voluntary Commitments from Leading Artificial Intelligence Companies to Manage the Risks Posed by AI by Jonathan Claybrough
    Welcome to The Nonlinear Library, where we use Text-to-Speech software to convert the best writing from the Rationalist and EA communities into audio. This is: News : Biden-⁠Harris Administration Secures Voluntary Commitments from Leading Artificial Intelligence Companies to Manage the Risks Posed by AI, published by Jonathan Claybrough on July 21, 2023 on LessWrong.
    Here are the extracts naming the companies and specific commitments :
    As part of this commitment, President Biden is convening seven leading AI companies at the White House today - Amazon, Anthropic, Google, Inflection, Meta, Microsoft, and OpenAI - to announce that the Biden-Harris Administration has secured voluntary commitments from these companies to help move toward safe, secure, and transparent development of AI technology.
    Today, these seven leading AI companies are committing to:Ensuring Products are Safe Before Introducing Them to the Public
    The companies commit to internal and external security testing of their AI systems before their release. This testing, which will be carried out in part by independent experts, guards against some of the most significant sources of AI risks, such as biosecurity and cybersecurity, as well as its broader societal effects.
    The companies commit to sharing information across the industry and with governments, civil society, and academia on managing AI risks. This includes best practices for safety, information on attempts to circumvent safeguards, and technical collaboration.
    Building Systems that Put Security First
    The companies commit to investing in cybersecurity and insider threat safeguards to protect proprietary and unreleased model weights. These model weights are the most essential part of an AI system, and the companies agree that it is vital that the model weights be released only when intended and when security risks are considered.
    The companies commit to facilitating third-party discovery and reporting of vulnerabilities in their AI systems. Some issues may persist even after an AI system is released and a robust reporting mechanism enables them to be found and fixed quickly.
    Earning the Public's Trust
    The companies commit to developing robust technical mechanisms to ensure that users know when content is AI generated, such as a watermarking system. This action enables creativity with AI to flourish but reduces the dangers of fraud and deception.
    The companies commit to publicly reporting their AI systems' capabilities, limitations, and areas of appropriate and inappropriate use. This report will cover both security risks and societal risks, such as the effects on fairness and bias.
    The companies commit to prioritizing research on the societal risks that AI systems can pose, including on avoiding harmful bias and discrimination, and protecting privacy. The track record of AI shows the insidiousness and prevalence of these dangers, and the companies commit to rolling out AI that mitigates them.
    The companies commit to develop and deploy advanced AI systems to help address society's greatest challenges. From cancer prevention to mitigating climate change to so much in between, AI - if properly managed - can contribute enormously to the prosperity, equality, and security of all.
    As we advance this agenda at home, the Administration will work with allies and partners to establish a strong international framework to govern the development and use of AI. It has already consulted on the voluntary commitments with Australia, Brazil, Canada, Chile, France, Germany, India, Israel, Italy, Japan, Kenya, Mexico, the Netherlands, New Zealand, Nigeria, the Philippines, Singapore, South Korea, the UAE, and the UK. The United States seeks to ensure that these commitments support and complement Japan's leadership of the G-7 Hiroshima Process - as a critical forum for developing shared principles for the governance of AI - as well as the United Kingdom's leadership in hosting a Summit on AI Safety, and India's leadership as C...
    4 min
  • EA - Linkpost: 7 A.I. Companies Agree to Safeguards After Pressure From the White House by MHR
    Welcome to The Nonlinear Library, where we use Text-to-Speech software to convert the best writing from the Rationalist and EA communities into audio. This is: Linkpost: 7 A.I. Companies Agree to Safeguards After Pressure From the White House, published by MHR on July 21, 2023 on The Effective Altruism Forum.
    The NYT just released a breaking news piece regarding an agreement on AI safeguards. It's hard to tell exactly how useful the proposed measures will be, but it seems like a promising step.
    Seven leading A.I. companies in the United States have agreed to voluntary safeguards on the technology's development, the White House announced on Friday, pledging to strive for safety, security and trust even as they compete over the potential of artificial intelligence.
    The seven companies - Amazon, Anthropic, Google, Inflection, Meta, Microsoft and OpenAI - will formally announce their commitment to the new standards at a meeting with President Biden at the White House on Friday afternoon.
    The voluntary safeguards announced on Friday are only an early step as Washington and governments across the world put in place legal and regulatory frameworks for the development of artificial intelligence. White House officials said the administration was working on an executive order that would go further than Friday's announcement and supported the development of bipartisan legislation.
    As part of the agreement, the companies agreed to:
    Security testing of their A.I. products, in part by independent experts and to share information about their products with governments and others who are attempting to manage the risks of the technology.
    Ensuring that consumers are able to spot A.I.-generated material by implementing watermarks or other means of identifying generated content.
    Publicly reporting the capabilities and limitations of their systems on a regular basis, including security risks and evidence of bias.
    Deploying advanced artificial intelligence tools to tackle society's biggest challenges, like curing cancer and combating climate change.
    Conducting research on the risks of bias, discrimination and invasion of privacy from the spread of A.I. tools.
    Thanks for listening. To help us out with The Nonlinear Library or to learn more, please visit nonlinear.org
    2 min
  • EA - Australians call for AI safety to be taken seriously by AlexanderSaeri
    Welcome to The Nonlinear Library, where we use Text-to-Speech software to convert the best writing from the Rationalist and EA communities into audio. This is: Australians call for AI safety to be taken seriously, published by AlexanderSaeri on July 21, 2023 on The Effective Altruism Forum.
    The Australian Government is considering how to regulate AI in Australia, has published a discussion paper ("Safe and Responsible AI"), and has invited feedback by 26 July 2023:
    "We want your views on how the Australian Government can mitigate any potential risks of AI and support safe and responsible AI practices."
    Good Ancestors Policy (goodancestors.org.au/policy), with the support of EA and AI Safety community organisers in Australia, have coordinated Australians' submissions to the feedback process.
    Today, the website Australians for AI Safety launched with a co-signed letter (media release). The letter called on the relevant Australian Federal Minister, Ed Husic, to take AI safety seriously by:
    recognising the catastrophic and existential risks
    addressing uncertain but catastrophic risks alongside other known risks
    working with the global community on international governance
    supporting research into AI safety
    Good Ancestors Policy have also held community workshops across Australia (e.g., Brisbane, Perth) to support members of the EA and AI Safety community in understanding the feedback process and preparing submissions, including access to some of the best evidence and arguments for acknowledging and addressing risks from AI. Policy ideas are drawn from the Global Catastrophic Risk Policy database (), the AI Policy Ideas database (aipolicyideas.com), and expert community input.
    So far, about 50 members of the community have attended a workshop, and feedback we've received is that the workshops have been very helpful, the majority (~75% people) are likely or very likely (>80% likelihood) to make a submission, and that most (~70% people) would be unlikely or very unlikely (<20% likelihood) to have made a submission without the workshop.
    If you're an Australian living in Australia or overseas, and you'd like to make a submission to this process, there is one more online community workshop on Saturday 22 July at 3pm AEST (UTC+10). Register here for the workshop!
    Contact Greg Sadler ([email protected]) or Alexander Saeri ([email protected]) if you'd like to stay involved.
    Thanks for listening. To help us out with The Nonlinear Library or to learn more, please visit nonlinear.org
    3 min
  • AF - Reward Hacking from a Causal Perspective by Tom Everitt
    Welcome to The Nonlinear Library, where we use Text-to-Speech software to convert the best writing from the Rationalist and EA communities into audio. This is: Reward Hacking from a Causal Perspective, published by Tom Everitt on July 21, 2023 on The AI Alignment Forum.
    Post 4 of Towards Causal Foundations of Safe AGI, preceded by Post 1: Introduction, Post 2: Causality, Post 3: Agency, and Post 4: Incentives.
    By Frances Rhys Ward, Tom Everitt, Sebastian Benthall, James Fox, Matt MacDermott, Milad Kazemi, Ryan Carey representing the Causal Incentives Working Group. Thanks also to Toby Shevlane and Aliya Ahmad.
    AI systems are typically trained to optimise an objective function, such as a loss or reward function. However, objective functions are sometimes misspecified in ways that allow them to be optimised without doing the intended task. This is called reward hacking. It can be contrasted with misgeneralisation, which occurs when the system extrapolates (potentially) correct feedback in unintended ways.
    This post will discuss why human-provided rewards can sometimes fail to reflect what the human really wants, and why this can lead to malign incentives. It also considers several proposed solutions, all from the perspective of causal influence diagrams.
    Why Humans might Reward the Wrong Behaviours
    In situations where a programmatic reward function is hard to specify, AI systems can often be trained from human feedback. For example, a content recommender may be optimising for likes, and a language model trained on feedback from human raters.
    Unfortunately, humans don't always reward the behaviour they actually want. For example, a human may give positive feedback for a credible-sounding summary, even though it actually misses key points:
    More concerningly, the system may covertly influence the human into providing positive feedback. For example, a recommender system with the goal of maximising engagement can do so by influencing the user's preferences and mood. This leads to a kind of reward misspecification, where the human provides positive feedback for situations that don't actually bring them utility.
    A causal model of the situation reveals the agent may have an instrumental control incentive (or similarly, an intention) to manipulate the user's preferences. This can be inferred directly from the graph. First, the human may be influenced by the agent's behaviour, as they must observe it before evaluating it. And, second, the agent can get better feedback by influencing the human:
    For example, we typically read a post before deciding whether to "like" it. By making the user more emotional, the system may be more likely to increase engagement. While this effect is stronger for longer interactions, the incentive is there even for "single time-step" interactions.
    Scalable Oversight
    One proposed solution to the reward misspecification problem is scalable oversight. It provides the human with a helper agent that advises them on what feedback to give. The helper agent observes the learning agent's behaviour, and may point out, for instance, an inaccuracy in a credible-looking summary, or warn against manipulation attempts. The extra assistance may make it harder for the learning agent to manipulate or deceive the human:
    Influential scalable oversight agendas include iterated distillation and amplification, AI safety via debate, recursive reward modelling, and constitutional AI.
    Unfortunately, the learning agent still has an incentive to deceive the human or manipulate their preferences, as the human preferences still satisfy the graphical criterion for an instrumental control incentive (it's on a directed causal path from behaviour to feedback). Additionally, the learning agent also has an incentive to deceive the helper agent:
    An important question for scalable oversight schemes is whether weaker agents can effectively help to supervise more capable agents (and whether this can be done recursively to supervise agents much smart...
    12 min
  • LW - Priorities for the UK Foundation Models Taskforce by Andrea Miotti
    Welcome to The Nonlinear Library, where we use Text-to-Speech software to convert the best writing from the Rationalist and EA communities into audio. This is: Priorities for the UK Foundation Models Taskforce, published by Andrea Miotti on July 21, 2023 on LessWrong.
    The UK government recently established the Foundation Models Taskforce, focused on AI safety and backed by £100M in funding. Founder, investor and AI expert Ian Hogarth leads the new organization.
    The establishment of the Taskforce shows the UK's intention to be a leading player in the greatest governance challenge of our times: keeping humanity in control of a future with increasingly powerful AIs. This is no small feat, and will require very ambitious policies that anticipate the rapid developments in the AI field, rather than just reacting to them.
    Here are some recommendations on what the Taskforce should do. The recommendations fall into three categories: Communication and Education about AI risk, International Coordination, and Regulation and Monitoring.
    Communication and Education about AI Risk
    The Taskforce is uniquely positioned to educate and communicate about AI development and risks. Here is how it could do it:
    Private education
    The Taskforce should organize private education sessions for UK Members of Parliament, Lords, and high-ranking civil servants, in the form of presentations, workshops, and closed-door Q&As with Taskforce experts. These would help bridge the information gap between policymakers and the fast-moving AI field.
    A new platform: ai.gov.uk
    The Taskforce should take a proactive role in disseminating knowledge about AI progress, the state of the AI field, and the Taskforce's own actions:
    The Taskforce should publish bi-weekly or monthly Bulletins and Reports on AI on an official government website. The Taskforce can start doing this right away by publishing its bi-weekly or monthly bulletins and reports on the state of AI progress and AI risk on the UK government's research and statistics portal.
    The Taskforce should set up ai.gov.uk, an online platform modeled after the UK's COVID-19 dashboard. The platform's main page should be a dashboard showing key information about AI progress and Taskforce progress in achieving its goals, that gets updated regularly. ai.gov.uk should have a progress bar trending towards 100% for all of the Task Force's key objectives.
    ai.gov.uk should also include a "Safety Plans of AI Companies" monthly report, with key insights visualized on the dashboard.
    The Taskforce should send an official questionnaire to each frontier AI company to compile this report. This questionnaire should contain questions about companies' estimated risk of human extinction caused by the development of their AIs, their timelines until the existence of powerful and autonomous AI systems, and their safety plans regarding development and deployment of frontier AI models.
    There is no need to make the questionnaire mandatory. For companies that don't respond or respond only to some questions, the relevant information on the dashboard should be left blank, or filled in with a "best guess" or "most relevant public information" curated by Taskforce experts.
    Public-facing communications
    Taskforce members should utilize press conferences, official posts on the Taskforce's website, and editorials in addition to ai.gov.uk to educate the public about AI development and risks. Key topics to cover in these public-facing communications include:
    Frontier AI development is focused on developing autonomous, superhuman, general agents, not just towards better chatbots or the automation of individual tasks. These are and will increasingly be AIs capable of making their own plans and taking action in the real world.
    No one fully understands how these systems function, their capabilities or limits, and how to control or restrict them. All of these remain unsolved technical challenges.
    Consensus on the societal-scale risk from AI is growing, and the gov...
    10 min
  • EA - FTX Foundation wanted to buy the nation of Nauru to save EAs' lives by Eugenics-Adjacent
    Welcome to The Nonlinear Library, where we use Text-to-Speech software to convert the best writing from the Rationalist and EA communities into audio. This is: FTX Foundation wanted to buy the nation of Nauru to save EAs' lives, published by Eugenics-Adjacent on July 21, 2023 on The Effective Altruism Forum.
    I thought this was interesting. I'd love to see a debate on whether this is a healthy line of thinking, and something we're glad the public knows about us now.
    The court filings in a federal bankruptcy court in Delaware, dated July 20, included a memo crafted by an FTX Foundation official and Sam Bankman-Fried's brother Gabriel Bankman-Fried. It outlined the future survival of FTX and Alameda Research employees and all those who subscribed to the effective altruism concept.
    The ultimate strategy, according to the memo, was "to purchase the sovereign nation of Nauru in order to construct a 'bunker / shelter' that would be used for some event where 50%-99.99% of people die [to] ensure that most EAs (effective altruists) survive."
    Thanks for listening. To help us out with The Nonlinear Library or to learn more, please visit nonlinear.org
    2 min
  • LW - Boundary Placement Rebellion by tailcalled
    Welcome to The Nonlinear Library, where we use Text-to-Speech software to convert the best writing from the Rationalist and EA communities into audio. This is: Boundary Placement Rebellion, published by tailcalled on July 20, 2023 on LessWrong.
    The idea for this post all started because I was confused about the concept of "narcissism". I learned about "narcissism" from reading differential psychology, where they often measure it by asking people whether they agree with statements such as:
    I have a natural talent for influencing people
    Modesty doesn't become me
    I think I am a special person
    I really like to be the center of attention
    In practice, empirically this correlates with being an assertive, confident person, which didn't match the discourse about narcissism, which typically seemed to more be about domestic abuse or people's ideologies, and it also doesn't AFAIK match the way "narcissism" gets used clinically, since clinical narcissists don't score higher than average on psychometric Narcissism scales used in personality psychology.
    Eventually, something clicked about what people were saying about narcissism. They were talking about a dynamic that occurs when someone feels morally entitled to violate a boundary. Such a person can exhibit signs of narcissism because they become focused on enforcing their own views on the boundary, and put a lot of effort into gaining power to continue violating it.
    Still, since I'm not sure whether what I'm talking about is actually "narcissism", I'm going to use a new, more descriptive term to refer to my concept: Boundary Placement Rebellion.
    Boundary Placement Rebellion comes up a lot in my experience. It's a key issue in AI safety, as well as in civil rights, psychology research, hierarchies, families and other areas of society. For good and for bad, I think Boundary Placement Rebellion sees far more symmetry between the sides than "narcissism" does.
    An Example: AI Safety vs Capabilities
    Common sense - at least among many right-wing techies - is that you have the right to work on whatever software projects you want. If people don't like your software, they can just not buy it. There may be exceptions when it comes to software whose purpose is to be used as a weapon, such as ransomware. If new issues are discovered, then maybe we can update this norm, but we should also be careful to not strangle the tech industry with paperwork.
    This is a boundary; you get sole decisive control over what you work on. AI capabilities research (as well as everything else in tech) makes a lot of use of this boundary, as it means that they can keep trying new things to push the state of the art. And of course earn lots of $$$ doing it.
    Now suddenly a bunch of people are coming in, arguing that AI will lead to the end of the world! Suddenly, the boundary for AI capabilities researchers is being directly threatened. And they can't just be dismissed by saying "don't worry, I'm not gonna destroy the world", instead they will dump huge arguments for why it will inevitably happen. Or maybe some of them will say "you may be right, but we cannot know for sure, so we still gotta stop you".
    The safety people's solution is that those who want to develop AI capabilities should either stop working on capabilities, or should pay the AI safety people engineer-level salaries to do philosophy, abstract mathematics, odd unprofitable ML experiments, premature safety tests, and similar.
    If the AI capabilities researchers try to come up with different frames to make the AI safety people go away, then the AI safety people will just keep on pushing back against those frames, constantly coming up with reasons for why we're all gonna die anyway or whatever. And if you don't give them what they want, they're going to complain about you, sometimes even accusing you of being the person in the history of the world who has caused the most damage.
    Broadening and Abstracting
    The above shouldn't be seen as an argument against AI safet...
    21 min
  • AF - Priorities for the UK Foundation Models Taskforce by Andrea Miotti
    Welcome to The Nonlinear Library, where we use Text-to-Speech software to convert the best writing from the Rationalist and EA communities into audio. This is: Priorities for the UK Foundation Models Taskforce, published by Andrea Miotti on July 21, 2023 on The AI Alignment Forum.
    The UK government recently established the Foundation Models Taskforce, focused on AI safety, modelled on the Vaccine Taskforce, and backed by £100M in funding. Founder, investor and AI expert Ian Hogarth leads the new organization.
    The establishment of the Taskforce shows the UK's intention to be a leading player in the greatest governance challenge of our times: keeping humanity in control of a future with increasingly powerful AIs. This is no small feat, and will require very ambitious policies that anticipate the rapid developments in the AI field, rather than just reacting to them.
    Here are some recommendations on what the Taskforce should do. The recommendations fall into three categories: Communication and Education about AI risk, International Coordination, and Regulation and Monitoring.
    Communication and Education about AI Risk
    The Taskforce is uniquely positioned to educate and communicate about AI development and risks. Here is how it could do it:
    Private education
    The Taskforce should organize private education sessions for UK Members of Parliament, Lords, and high-ranking civil servants, in the form of presentations, workshops, and closed-door Q&As with Taskforce experts. These would help bridge the information gap between policymakers and the fast-moving AI field.
    A new platform: ai.gov.uk
    The Taskforce should take a proactive role in disseminating knowledge about AI progress, the state of the AI field, and the Taskforce's own actions:
    The Taskforce should publish bi-weekly or monthly Bulletins and Reports on AI on an official government website. The Taskforce can start doing this right away by publishing its bi-weekly or monthly bulletins and reports on the state of AI progress and AI risk on the UK government's research and statistics portal.
    The Taskforce should set up ai.gov.uk, an online platform modeled after the UK's COVID-19 dashboard. The platform's main page should be a dashboard showing key information about AI progress and Taskforce progress in achieving its goals, that gets updated regularly. ai.gov.uk should have a progress bar trending towards 100% for all of the Task Force's key objectives.
    ai.gov.uk should also include a "Safety Plans of AI Companies" monthly report, with key insights visualized on the dashboard.
    The Taskforce should send an official questionnaire to each frontier AI company to compile this report. This questionnaire should contain questions about companies' estimated risk of human extinction caused by the development of their AIs, their timelines until the existence of powerful and autonomous AI systems, and their safety plans regarding development and deployment of frontier AI models.
    There is no need to make the questionnaire mandatory. For companies that don't respond or respond only to some questions, the relevant information on the dashboard should be left blank, or filled in with a "best guess" or "most relevant public information" curated by Taskforce experts.
    Public-facing communications
    Taskforce members should utilize press conferences, official posts on the Taskforce's website, and editorials in addition to ai.gov.uk to educate the public about AI development and risks. Key topics to cover in these public-facing communications include:
    Frontier AI development is focused on developing autonomous, superhuman, general agents, not just towards better chatbots or the automation of individual tasks. These are and will increasingly be AIs capable of making their own plans and taking action in the real world.
    No one fully understands how these systems function, their capabilities or limits, and how to control or restrict them. All of these remain unsolved technical challenges.
    Consensus on the so...
    10 min