Should the US Gov't Have to Review and License AI Tools?

0:00The show you're about to hear originally came out as a YouTube video on Tuesday, April 11th. The discussion is all about governments trying to reclaim, well, something in the face of this insurgent movement towards AI. Enjoy.

0:19What's going on, guys? Welcome to the AI Breakdown. These are short videos that come out either daily or more than daily, I bet in some cases, and are also pushed to the AI Breakdown podcast feed, all about the most important news and discussions in AI. Sometimes that will be some new tool or some new use case, but often, especially right now, it will be about how AI is interfacing with regulations, with big economic questions, with societal concerns. And that's certainly been the story today. The Wall Street Journal and many other outlets ran a story about the Biden administration's new request for public comments around AI.

1:00So basically, what's happened here is that the Biden administration is getting deeper into the discussion of what rules there should be around new AI tools. Specifically, they've asked the Commerce Department to put out a formal request for comment about what it calls accountability measures. And that includes things like whether AI models should go through some sort of public certification process or review process before they're allowed to be pushed out into the public. So this is obviously very different or would be very different than other sorts of tech platforms where when a new Instagram clone or a new TikTok clone comes out, it just goes live into the app store.

1:42And the only real determinant of whether it gets to people is the owner of that app store or anyone who has a website can just download it directly. The idea here is that artificial intelligence may be not fit for that sort of paradigm of just test it out and see what works. there may be bigger implications that require or should mandate more government involvement. At least that's the discussion. This is a discussion that is happening not just in the U.S., but all over the world. I don't think the timing of this is an accident, given that also on Tuesday, China's top internet regulator proposed its own set of controls.

2:21So these are basically, instead of a public comment period, This is China saying that AI companies in their country would have to follow a set of rules. So in China, they're focused on things like making sure that those platforms don't generate content that could quote unquote disrupt social order, that could subvert state power. Those are the important things to them. Now, this is a really complex, thorny set of issues. One of the things that many have noted is that tools like MidJourney do allow people to make fakes or just, you know, representations of public figures in the U.S. like Biden, but do not allow for that with Chinese leaders like Xi Jinping.

3:04And whether that's just exporting basically Chinese censorship around the world so that these platforms can play globally is a real concern and a real point of discussion. But I think the broader point here is that we've gone from this sort of just this Cambrian explosion moment, this punctuated equilibrium moment, to now governments racing to catch up because they realize that something profound is happening. There are a lot of concerns about that, though. So we saw the Biden administration start to wade into this conversation about a week ago. The president's official Twitter account tweeted, and this was April 4th, When it comes to AI, we must both support responsible innovation and ensure appropriate guardrails to protect folks' rights and safety.

3:52Our administration is committed to that balance, from addressing bias in algorithms to protecting privacy and combating disinformation. Artificial intelligence has enormous potential to tackle some of our toughest challenges, but we must address its risks. That's why last year we proposed an AI Bill of Rights to ensure that important protections for the American people are built into AI systems from the start. Now, if you went into this particular tweet, these two tweets, the vast majority of the comments were basically things like these issues combating disinformation, bias and algorithms are very low on many people's lists of concerns with AI.

4:34It feels, in other words, like the Biden administration, which is almost to be expected, is bringing in their previous political biases about what the problems of this technology might be, rather than dealing with what the true challenges of this technology might be holistically from the ground up. Still, there are some folks who think, even if this isn't exactly the conversation they want, that this conversation has to start somewhere. Chamath Palihapitiya of Social Capital writes, If you invent a novel drug, you need the government to vet and approve it before you can commercialize it. If you invent a new mode of air travel, you need the government to vet and approve it before you can commercialize it.

5:12If you create a new security, you need the government to vet and approve it before you can commercialize it. More generally, when you create things with broad societal impact, positive and negative, the government creates a layer of review and approval. AI will need such an oversight body. The FDA approval process seems the most credible and adaptable into a framework to understand how a model behaves and it's counterfactual. Our political leaders need to get in front of this sooner rather than later and create some oversight before the eventual big avoidable mistakes happen and genies are let out of the bottle.

5:44This is the positive take that basically says we have to start this conversation. We're late to it, but we have to start it somewhere. Now, others are kind of frustrated that this conversation is coming after America has fallen out of love with tech. Luke Metro here says it's very unfortunate that the tech industry delivered the most self-evident revolutionary product in 15 years, right after it spent the last two years losing tons of credibility by shilling crypto and SPAC scams to the public. Effectively, this person is suggesting that AI has these profound implications, which is something I obviously agree with, which is why I spend time on it.

6:19but people are likely or more likely than they might have been previously to be skeptical of it, to want a more serious and severe response. I think it's an interesting take and certainly it is the case that America has fallen out of love with tech in a big way. On my other channel, the main breakdown channel, I did a show a few weeks ago about why America hates tech so much. It came out during the Silicon Valley bank affair and I think goes really deep into how our love affair with tech ended. Now, another take on this is that we're so likely to get it wrong because of it being reactionary. So Brian Rommel here says, instead of making self-governance rules in 1999 like it was advised, now we will get the hysterical reactionary politicians creating the Safety AI Act of 2023.

7:05You will not like it. I also think that this is a tremendously important point. It is highly possible that we see a massive overcorrection, a massive fear-based correction that just rips the pendulum to the other side because we feel, or politicians specifically, feel that they are behind and that they have to do something. and that if they don't do something, it's going to get very scary. I think that one of the things that we'll explore a lot on this channel is whether the safety advocates in AI are doing the space a service through their rhetoric or whether they actually create as many problems as they solve.

7:42I'm not making that claim, but I do think it's something that we have to discuss as this becomes a bigger national conversation. Now, there is another piece of this, which I have to say, coming from a space like the crypto space, is notable. The rhetoric that you see around some of this, and I think this goes to Brian's point from what we just saw before, is not unfamiliar to a crypto audience who has seen years and years of certain types of narratives be used to delegitimate the space. So you see in this Wall Street Journal article, fears grow over the potential use of AI to commit crimes. That's something akin to saying AI is for criminals.

8:22And that's, of course, something that we've seen over and over again in the crypto space. Now, you may think that that is the case, just the same way that you may think that Bitcoin mining uses too much energy and it shouldn't be allowed. What I would caution is that those types of arguments can be extracted and abstracted away from crypto specifically and used for more governmental control. Now, maybe that's what you want. There's a lot of different takes on what the right solutions to the challenges presented by AI are. But I think it's worth noting, let's say, that there is a similar rhetoric that we're seeing.

9:00So that's what is happening today in this story. I'm seeing a lot of discourse around this, and I think it just gets at so much of what's going on. This feeling of the government feeling like they're late and they have to catch up and potentially being overreactionary. The contrast between a potential U.S. approach to this and a Chinese approach and what that will look like. Then the larger arms race there and the brinksmanship that might come from it. Meanwhile, underneath all of this is the fact that these tools are out and clawing them back and walking them back once unleashed is going to be extremely difficult.

9:32So this is something that does need to happen fast. And I expect that this is going to rise from a thing that people didn't even imagine would be on their political agenda in early 2023 or heading into this year to something that is right at the very top. Anyways, guys, that's it for today's, or at least for this, AI breakdown. I don't know if there will be another one. But for now, I appreciate you hanging out. Let's keep focused. Let's keep talking. Until next time, peace.

10:03Thank you.