Biden's AI Executive Order: Everything You Need to Know

0:00Today on the AI Breakdown, we're going in depth on the new Biden executive order on artificial artificial intelligence. The AI Breakdown is a daily podcast and video about the most important news and discussions in AI. Go to Breakdown.net for more information about our YouTube channel, our newsletter, and our Discord.

0:24Hello, friends. Just a quick note before we get into the show. I had been planning on doing a brief as normal today, but the executive order part of the show was so much longer than the normal main part of the episode that I decided to just run it in its entirety today. Obviously, having the White House have this comprehensive set of AI actions less than a year after ChatGPT, I think, makes very clear how significant this issue has become. But never fear, there is a lot to catch up on on the brief, from updates in the ChatGPT product itself to Anthropic getting new funding, and all of that will be back tomorrow.

0:57We will do a brief and a main episode as normally planned. But for now, let's talk executive order. Welcome back to the AI Breakdown. Today, as we were anticipating, we have gotten the Biden AI executive order. By way of introduction, Box CEO Aaron Levy tweeted, Biden's AI executive order is the gold standard for how governments should be regulating AI right now. Thoughtful but scoped oversight focused on practical risks, emphasis on privacy and security, focus on R &D across the ecosystem and encouraging use of AI in the government. Now, today I am not going to do very much of a primer or background because yesterday's episode was, of course, the state of play of USAI regulation.

1:39And instead, what we're going to do is actually dig into the fact sheet, which so far from what I've seen is all that has been released. Now, when you get an executive order like this, you have the actual text of the order. But the thing that most people reference is the fact sheet, which gives the high level summary. The fact sheet, which was published by the White House this morning, was called President Biden issues executive order on safe, secure, and trustworthy artificial intelligence. Now, the line the government is taking is very clear. In a statement, Deputy Chief of Staff Bruce Reed said, President Biden is rolling out the strongest set of actions any government in the world has ever taken on AI safety, security, and trust.

2:15It's the next step in an aggressive strategy to do everything on all fronts to harness the benefits of AI and mitigate the risks. So let's look first at the specific new actions that come from this executive order. Section 1 is called New Standards for AI Safety and Security. The White House writes, with this executive order, the president directs the most sweeping actions ever taken to protect Americans from the potential risks of AI systems. In many ways, the first bullet action here is the most significant and the most interesting, although it does leave some questions. That bullet is, require that developers of the most powerful AI systems share their safety test results and other critical information with the U.S.

2:52government. They write, in accordance with the Defense Production Act, the order will require that companies developing any foundation model that poses a serious risk to national security, national economic security, or national public health and safety must notify the federal government when training the model and must share the results of all red team safety tests. So there are two things that are really notable to me right away when I first read this. And again, this is the first bullet in the executive order fact sheet. The first is that this goes farther than what we had seen in some of the previews or reporting.

3:22A lot of the discussion from those preemptive reports was basically around the government using its own procurement process as a way to get some of this information versus just mandating they have access to it right away. Now, of course, this stops short of having some sort of certification program or licensing regime that stops companies before they release these models, but it still creates a more significant information flow than some were previously anticipating. Now, of course, the question here is how the Biden administration and any future White House occupants will define what constitutes serious risks to national security, national economic security, or national public health and safety?

3:57Is this a stand-in for anything more advanced than GPT-4? Would GPT-4 qualify here? I think obviously that is going to be one of the most important clarifications from this as we learn more about what the intention is. Now, speaking of testing, the next bullet is develop standards, tools, and tests to help ensure that AI systems are safe, secure, and trustworthy. Basically here, and a theme that you'll see throughout this, is the White House empowering the National Institute of Standards and Technology to not only get visibility into the safety tests that AI labs are running, but also to try to set standards for what those tests should actually look like.

4:29Now, importantly, in addition to the NIST helping develop these standards, the Biden administration is also directing the Department of Homeland Security to apply those standards to infrastructure risks, as well as to establish an AI safety and security board. And Homeland Security will also, with the Department of Energy, look at AI systems threats on the chemical, biological, radiological, nuclear, and cybersecurity front. Indeed, that brings us to bullet three. Protect against the risks of using AI to engineer dangerous biological materials. Basically, this is a mandate for agencies that fund life science projects to develop new standards for biological synthesis screening.

5:02After that, protect Americans from AI-enabled fraud and deception by establishing standards and best practices for detecting AI-generated content and authenticating official content. Obviously, this question around watermarking and other solutions to help people identify what is AI created and what is not has been a big conversation in the private sector as well, but the White House is now getting the Department of Commerce involved, actually developing guidance for content authentication, and then getting federal agencies to actually use those tools as well. The last two bullets in this section are, one, establish an advanced cybersecurity program to develop AI tools to find and fix vulnerabilities in critical software.

5:38Basically, this is in the category of using AI to make things better. And then finally, order the development of a national security memorandum that directs further actions on AI and security. Effectively, this creates a mandate for the National Security Council and the White House Chief of Staff to take some of these inspirations and bring them into the military and intelligence community. The next section of the EO is called Protecting Americans' Privacy. This section acknowledges that one of the risks of AI is the further deterioration of privacy around people's data. And so they write, The president calls on Congress to pass bipartisan data privacy legislation to protect all Americans, especially kids.

6:13Now, in addition to that call to Congress, the EO also identifies the following actions as well. The first is prioritizing federal support for accelerating the development and use of privacy-preserving techniques. They're interested here in how AI systems can be trained while retaining the privacy of the training data itself. A second action is to strengthen privacy-preserving research and technologies, such as cryptographic tools that preserve individuals' privacy, by funding a research coordination network to advance rapid breakthroughs and development. I can practically hear my zero-knowledge proof crypto kids screaming at their computers when they read that one.

6:45Next, the Biden administration is going to evaluate how agencies collect and use commercially available information, and strengthen privacy guidance for federal agencies to account for new AI risks. And lastly, they call to develop guidelines for federal agencies to evaluate the effectiveness of privacy-preserving techniques. The third bullet is advancing equity and civil rights. Now, one of the things you may have heard me talk about on here before is the tension not only between the accelerationists and the AI safety people, but the fracture within the AI safety space itself between those who are focused on longer-term existential risk-type questions and those who think that that conversation is distracting and leading people away from more urgent issues that are here and now.

7:23Now, I have a whole show-length rant on why I think the idea that focusing on one area of AI risk somehow crowds out space for other areas of risk. But that's neither here nor there, at least when it comes to this particular piece. Now, within this advancing equity and civil rights category, the White House is looking to, one, provide clear guidance to landlords, federal benefits programs, and federal contractors to keep AI algorithms from being used to exacerbate discrimination. 2. To address algorithmic discrimination through training, technical assistance, and coordination between the Department of Justice and federal civil rights offices on best practices for investigating and prosecuting civil rights violations related to AI and 3.

8:02Ensure fairness throughout the criminal justice system by developing best practices on the use of AI in sentencing, parole, and probation, pretrial release and detention, risk assessment, surveillance, crime forecasting, and predictive policing and forensic analysis. Now, I've seen a few people feel like this one is more generic than they would like, particularly those folks who are focused on this set of issues. I understand where they're coming from to some extent. The first bullet seems more clear-cut, providing guidance to these various categories of people to try to make sure that AI doesn't exacerbate discrimination.

8:29I also think that this second category, while a little vague, does have a clear purpose. Basically, they're trying to create an incentive structure where because people are getting investigated and prosecuted for civil rights violations related to AI, it creates an incentive for the companies who are training these models to try to address them before they become issues. This last one, though, is definitely a catch-all for this is something we need to address, but we're not going to go in-depth on it here. Obviously, when it comes to how AI is used in the prosecution of criminal justice, there are huge questions.

9:00Predictive policing, for example, is a total third-rail issue that comes with a ton of pitfalls. In fact, it's something that in the EU AI Act, they basically just say it's not allowed. So again, to the extent that you are really focused on this particular set of issues surrounding AI, I can see why this placeholder is not particularly satisfying. Alas, for now, that's all we have. And so we move on to the next section, standing up for consumers, patients and students. And now a word from today's sponsor. Are you interested in how two top of mind trends, AI and crypto, can work together? If so, I have the perfect podcast recommendation for you.

9:36Web 3 with A16Z Crypto, the chart-topping show brought to you by venture firm Andreessen Horowitz. Web 3 with A16Z Crypto is your definitive resource for the future of the internet, whether you're already building in these spaces or simply curious about what's next. If you need a place to start, they recently released an excellent episode with Stanford cryptography professor Dan Bonet and former Google Xer Ali Yahya in conversation with host Sonal Choksi about the intersection of AI and crypto. From fighting deepfakes and proving humanity to large language models like ChatGPT, they cover it all.

10:08I highly recommend checking it out, especially if you'd like to learn more about how AI and crypto will impact our everyday lives. Beyond Crypto and AI, this show is for creators seeking more ways to truly own their work, for business leaders trying to prepare for the future today, and for innovators exploring trending tech topics. So go ahead, listen to Web3 with A16Z Crypto wherever you get your podcasts. In this, there are two big categories that they are interested in harnessing the benefits of AI to promote. The first is in healthcare, and the second is in transforming education. On the former, the White House wants to advance the responsible use of AI in healthcare and the development of affordable and life-saving drugs, although they don't give any sort of program or clarity around how they're intending to do that.

10:50However, from a more practical standpoint, they are directing the Department of Health and Human Services to establish a safety program to receive reports of and act to remedy unsafe healthcare practices that involve AI. Education is similarly vague, with the intention to create resources to support educators deploying AI-enabled educational tools, which they give as an example personalized tutoring. The next category in this executive order is another one of the hottest button topics, which is, of course, what AI is going to do to people's jobs. The White House writes, AI is changing America's jobs and workplaces, offering both the promise of improved productivity, but also the dangers of increased workplace surveillance, bias, and job displacement.

11:29To mitigate these risks, support workers' ability to bargain collectively, and invest in workforce training and development that is accessible to all, these are the following actions. The first is pretty basic. Develop principles and best practices to mitigate the harms and maximize the benefits of AI for workers. Again, this one is a little bit vague, and whereas you saw specific departments called in to do specific things in some of the previous bullets, this action is just very generic. Now, the intention here is quite clear. They write, these principles and best practices will benefit workers by providing guidance to prevent employers from undercompensating workers, evaluating job applications unfairly, or impinging on workers' ability to organize.

12:05Basically, they're saying, AI could put way more power in the hands of the employers, and so we want to address and make sure that employees get some power back as well. The second bullet here is a little bit more action-oriented, although if in an intermediate kind of way and is to produce a report on AI's potential labor market impacts, as well as studying and identifying options for strengthening federal support for workers facing labor disruptions. Now, my personal bias and belief is that this is an area that is going to need a ton of attention. I mean, hell, Andrew Yang's entire campaign for the presidency was to some extent based on a zoom out look at what might happen in a world where AI was truly changing the nature of jobs and the social contract.

12:45And on top of that, a recent survey from edX showed that CEOs in America think that 49 % of the skills that their employees have are going to be irrelevant in just two years. That sort of disruption, in terms of both the scale and the speed of the disruption, is going to take a heck of a lot more than just a report on the potential labor market impacts. But as I think you're starting to see, a lot of what this executive order is about is planting a flag on issues that need to be addressed, where possible using the White House's influence or specific actions that it would otherwise take to start to adapt some systems.

13:21But as we'll discuss towards the end, also kind of laying out a mandate for what Congress should go and do as it thinks about comprehensive AI regulation. The next category is promoting innovation and competition. And this comes at an especially interesting time given that over the weekend, Facebook's chief AI scientist accused Sam Altman and the head of Anthropic and others as effectively doing corporate lobbying behind the scenes to try to close off AI innovation to anyone except the biggest AI labs. Now, from the White House's perspective, they argue that America currently leads AI innovation, with more AI startups raising capital in the U.S.

13:56last year than in the next seven countries combined. This section of the EU is about continuing that lead and indeed expanding it. To do so, they're looking at actions that include, one, catalyzing AI research across the United States. They're looking to pilot something they're calling the National AI Research Resource, which they describe as a tool that will provide AI researchers and students access to key AI resources and data, as well as expanding grants for AI research in areas like healthcare and climate change. Second, they want to promote a fair, open, and competitive AI ecosystem by providing small developers and entrepreneurs access to technical assistance and resources, helping small businesses commercialize AI breakthroughs, and encouraging the Federal Trade Commission to exercise its authorities.

14:34Now, this one is a small bullet that has significant implications. This one is very clearly written directly towards big tech. The FTC has, of course, been the U.S. body that has taken the most direct action against big tech over the last few years. And I think if you've been watching that, this bullet reads with a very clear, if subtle, anti-monopolous tone. Now, of course, right now, the White House is playing nice with all the big labs. They are bringing them into the fold, getting those voluntary commitments that we saw earlier in the year. And so they're not going to come right out and say, we're going to go after you if you get too big or if you block too many other smaller innovators.

15:10But that's certainly the implication here. Now, a last bullet in action point here, which I also think is hugely significant and shows that the White House really is not just looking at this in terms of mitigating risks, but also unseizing opportunities. This third bullet reads, use existing authorities to expand the ability of highly skilled immigrants and non-immigrants with expertise in critical areas to study, stay, and work in the United States. Basically, at a time where effectively across the board, immigration, even skilled immigration, is getting harder, not easier, when it comes to AI-related skills, the White House is very clearly trying to reverse that flow.

15:44Which brings us nicely to the penultimate section of the EO fact sheet, Advancing American Leadership Abroad. Basically, this is the section about playing nice with other nations and getting them to fall in line when it comes to aligned approaches to AI regulation. They discuss here, one, expanding bilateral, multilateral, and multi-stakeholder engagements to collaborate on AI. They mention Vice President Kamala Harris speaking at the UK Summit on AI Safety and direct the State Department in collaboration with the Commerce Department to lead an effort to, quote, establish robust international frameworks for harnessing AI's benefits and managing its risks and ensuring safety.

16:18They're also looking to accelerate the development and implementation of vital AI standards with international partners, as well as to work with those partners to promote the safe, responsible, and rights-affirming development and deployment of AI abroad to solve global challenges. The last section of the EO could sort of be summed up as doing our part. It's called Ensuring Responsible and Effective Government Use of AI. The White House writes, AI can help government deliver better results for the American people. It can expand agencies' capacity to regulate, govern, and disperse benefits, and it can cut costs and enhance the security of government systems.

16:50However, use of AI can pose risks, such as discrimination and unsafe decisions. To ensure the responsible government deployment of AI and modernize federal AI infrastructure, the president directs the following actions. Those include issuing guidance for agencies' use of AI, including standards to protect rights and safety, help agencies acquire specific AI products and services faster and more cheaply, and lastly, and this one is really interesting, accelerate the rapid hiring of AI professionals as part of a government-wide AI talent surge led by the Office of Personal Management, U.S. Digital Service, U.S.

17:21Digital Corps, and Presidential Innovation Fellowships. Agencies will provide AI training for employees at all levels of relevant fields. Now, the White House really leaned into this one. President Biden's tweet on this this morning was, artificial intelligence is moving quickly, and so is my administration. Visit AI.gov and see how we're harnessing the power of AI and keeping all Americans safe in the process. Now, AI.gov does have some information about the administrative actions around the risk mitigation side of this. It has the executive order, plus discussions of their blueprint for an AI Bill of Rights, the NIST AI Risk Management Framework, the voluntary company commitments that we got earlier this year, the National AI R &D Strategic Plan, and the National AI Research resource roadmap.

18:03But when it comes to the thrust of the page and what they're framing, it's definitely about recruitment and opportunity. The banner headline on the front page is, Making AI Work for the American People. Join the National AI Talent Surge, and the call to action is an Apply Now button. They write, The federal government is rapidly hiring talent to build and govern AI to fulfill the priorities set forth by the Biden-Harris administration. And if for some reason you are not ready to apply yet, well, then you can also go ahead and build your AI skills. They write, American students, workers, and educators are foundational to ensuring our nation's leadership in AI.

18:36Under that, they have a call to action, find resources to build your AI skills. They also have a whole section about bringing your AI skills to the US. And write, building the most talented AI workforce in the world includes welcoming the world's AI talent to our country. So again, I think the idea here that you're getting is that for as much focus as there is on concerns and risk issues, there is also definitely an emphasis on American leadership and frankly, extending the American leadership place when it comes to artificial intelligence. So subjectively, what do I think of this? First of all, while there are still a ton of implementation questions and detail questions, I'm glad and encouraged to see a mandate for advanced models to share their red teaming and basically just be more transparent about what they learn through the testing process.

19:22It's hard for me to see how that does anything but improve the likelihood of positive results and reduce the likelihood of bad results as we move farther and farther past where we are right now when it comes to the capabilities of these models. Second, I commiserate with folks who see a lot of these bullets as fairly incomplete and lacking in detail. And again, I have to kind of read this as almost placeholders in flag planting. And so I think that the optimistic take is that they didn't want to wait any longer on actually getting this executive order out, even if it meant that everything isn't quite figured out as it stands today.

19:54Although I'd like to hear more about exactly what they're thinking, I'm encouraged to see mention and reference to open competition when it comes to AI. And I am particularly glad to see all of this talk about how to help people get these skills and how to import these skills from other places. So far, I like the tone of this action a bit more even than what we've seen out of the UK. But then again, we're going to get a lot more information about how that government is viewing things later this week as well. In any case, tons here to digest, and I'm sure lots more to discuss later this week. But for now, that is the Biden executive order on artificial intelligence.

20:29It comes a little less than a year after the launch of ChatGPT and shows just how remarkably fast this issue has gone to the very center of the American economy and the American society. If you want to chat about this, come join the AI Breakdown Discord. You can go to bit.ly slash AI Breakdown. And until next time, peace.

20:57Thank you.