Transcription
These companies, which have enormous amounts of debt and really no viable business model right now, they die and other companies appear, right? The Anthropics, the Open AIs are betting this doesn't happen and they're running as fast as they can with that singular bet.
Hello and welcome to the Tech Report. I'm Louis Sykes. Today we are speaking to Bruce Schneider, technology and cyber security expert and best-selling author of Rewiring Democracy, who lectures in public policy at the Harvard Kennedy School. Bruce, thank you very much for joining me.
>> Thanks for having me.
>> Well, Bruce, the UK competition regulator says news businesses can now block their content from appearing in AI searches. What's that going to mean for Google and their AI search system?
>> You know, Google is incredibly powerful in search. They basically have a monopoly and they control what people see. The ability to tell Google, don't include me, puts some power back to the content creators, which is a general a good thing. Specifically with AI, I mean, if you think about what AI search does is you never go to the place. So, it used to be with search you'd get results and then you go visit the news site or my blog or whatever. But the AI summarizes it. People never see what other people write. And that's actually bad for the web in general. Like why am I going to write if only AIs will read it? And forcing people or at least making people still go back to the source material, I think makes a lot of sense, especially in a world where the AI gets it wrong so often.
And is there also a level of concern around these AI companies having such a significant power in being able to almost influence the way you read something by almost reinterpreting these news stories as part of an AI summary?
>> I mean, yes, there's extraordinary power and the power still exists, right? Because the first page of the AI search results is all you ever look at. So they have the power to determine where you go for your information. But to actually take it and synthesize it so you don't have to go anywhere gives them even more power. And I think we should rightly worry about the web as it exists. It's not going to exist if no one visits it anymore. If only AI visit it. Which means the training data that the AI needs to be smart would no longer be produced. In a sense, Google is eating its own seed corn doing this. And I really wonder what they're thinking and why it's only short term.
Yes. Because there have been a lot of concerns raised in the past about AI summaries sort of cannibalizing themselves in that there is such a sudden and rapid proliferation of AI information on the internet that's that some of it is becoming unreliable because it's relying on other unreliable AI information. And it's actually even worse than that. There's good research that shows that AI trained on AI output actually does worse. So you can imagine if you train on an AI in the world's music, you get an AI that's Mozart. You train it on its output, you get Satie. You train on its output and you get gibberish. So, it's not just AI search. It's all of these AI content, whether it's web pages or articles or commentary on Reddit that is in effect making the AIs of tomorrow worse.
>> And there is of course this ongoing debate around what data AI companies should be allowed to access and how they access it. You know, this big battle over so-called data scraping. Is it fair to say that with decisions like this by regulatory bodies the data scraping battle is one that AI companies might be losing? You know, they're certainly losing it in the court of public opinion, right? Stealing everybody's copyrighted work, you know, my books have been stolen by the AI companies, anything you've written has been almost certainly. So yes, giving power back to content creators feels like a good thing. It's hard to know, right? I'm in the US and there's really not a lot of movement in that direction, although there are lawsuits that are going through the courts. In the EU, we're seeing movement to give rights back to the content creators and here we see in the UK a similar thing happening and this feels like a good, this feels like a good thing in general.
How difficult is it for regulatory bodies though to lay down the law on that issue? Because at some point you're essentially just trying to stop access to the internet for these for this data scraping to happen presumably.
>> Well, but you know, it's access by machines. I mean, so I mean, we could take a human parallel. I want you as an artist to be able to go into a museum and look at everybody's noses on paintings and learn how to paint a nose, right? I mean, I want you as a human to be able to look at all the previous work. It's different when a machine does it. And you're not going to prevent it through tech means. You're going to make it illegal. You're going to make it right. You're going to make the rules such that you can't do that. Companies do follow the law largely. So I think that will have an effect. The problem is only very few countries can do this effectively, right? The US, the EU, you know, large countries, large bodies can have this effect. You can imagine a country, a smaller country on the planet, AI companies going to just ignore them because they have just too much power right now. We saw that with social media. So I do like it that large countries like the UK, you can imagine India doing this or Brazil, you know, countries with a population that matters taking these steps.
>> And who is leading the charge on this? Is it is it the European Union that is still a sort of bastion of regulation in this in the tech space or or do we think these lawsuits might actually be the thing that really gets the attention of AI companies?
>> It's both. I mean, right now the EU is the regulatory superpower on the planet. I mean, they are the ones who've regulated social media and digital markets, digital services and now AI. In the US, it really is the court system and it's hard and I mean, there's a lot of settlements, the numbers aren't that great. Like better you go back to the EU and look at some of the fines against companies under GDPR for privacy violations. And right now the EU AI Act has not come into force yet. A lot of pressure by the US not to enforce it. So we'll see how that goes. I do look to Europe more than the US, although things are happening in the US courts that make a difference.
>> The we mentioned the court of public opinion on this earlier. There does seem to be a sort of general rejection of AI search results. Duck.go Go, the search engine that's made a point of not showing AI results, reports a threefold increase in users since Google began showing AI results. Is that likely to worry Google and other AI companies that have really tried to integrate the idea of AI almost as a replacement for the search engine?
>> I don't know. I mean, it's still, you know, three times increase is still a rounding error for Google's monopoly status. Most of the AI we use is thrust upon us. It's not stuff we choose, right? When we open up our turn-by-turn directions on our phone or we get our social media feed, I mean, those are AI systems. We're not really using AI. AI is being used for the things that we do use. So, while I think there is some rejection, certainly by younger people, of using chatbots, it's hard to know how much that is affecting AI adoption. AI as a programmer has gone through some tumultuous months. A few months ago, I mean, six months ago, it was not good enough yet. Three months ago, it was great. We're all using it. Now, it's, whoa, it's way too expensive. We're back to hiring humans. And that really speaks to how fast this is all changing, right? No policy lasts more than a few months because this technology moves so fast.
And the decision by the UK's CMA, the regulator, the market regulator came after the UK recognized Google as having a strategic market status, which I mean, anyone could recognize that Google holds what is in effect an almost monopoly over the search engine.
>> The US, we call it a monopoly, but that is a fine, multi-syllable way of saying the same thing.
>> Yeah. And you know, this wider recognition though of Google's new monopoly position as it uses AI to shape data, could this really catch people's attention, catch more regulators' attention on Google's position where they may earlier have just seen it as being a less influential position? Because with AI, there is perhaps more shaping of data that goes on rather than just provision of data.
>> It's interesting and I don't know, it's always hard for me to guess on public opinion or future policy. I mean, things happen so almost randomly it seems, but certainly people are looking at AI more critically. These companies are extraordinarily large and powerful. In the US, of course, money and politics means they have an outsized influence on policy. Just watch what happened a few weeks ago. Trump said that the US government's going to have to review new models to make sure they're not dangerous. That got watered down this week to an executive order where we asked, pretty please, can the companies show the government the models, but of course, no regulation under any circumstances. So it's hard to know how this goes politically. This is really why I look towards the EU, the UK, other countries to lead here rather than the US. Although in the states, we are seeing AI regulation, most notably California. Trump administration wants to preempt all those state regulations, that hasn't happened yet.
And how are companies like Google, that may before not have had to worry about regulator challenge or just a change in public opinion, how worried are they going to be about losing maybe their market dominance by pushing too hard on these more untested AI systems?
>> I mean, I don't know. The big companies are betting that the big models will always be better. It's an open question in the research, these smaller, cheaper models that are coming from China from elsewhere are almost as good. And if they get just as good, it's all over, right? Suddenly there are hundreds of AI companies who are doing just as good and the price goes down, they're more energy efficient, they can run on your computer. I know someone who runs models on their phone. They run slow, but it's on their freaking phone. And that's the kind of thing Apple is looking to do because they are more privacy protecting than some of these more surveillance-oriented companies. So there's a lot of technology unknowns that go into that. The big AI companies are making a bet on big AI and we don't know if that's going to be the right bet.
>> Is it inevitable then that this market will become a more competitive one? Because of course at the moment, with chips companies like Nvidia, it does seem to be very much led by or even controlled by a small group of largely American companies. Is it inevitable that there will be some element at least of more competition that comes in?
>> I mean, nothing is inevitable, but it does seem that the future is turning towards these smaller, more efficient models. I mean, when you think about it, well, you don't need a model that does everything. You need a travel agent. You need a college guidance counselor. You need a home shopping assistant, right? You need a specialized task, just like you use specialized humans. You don't hire someone who knows everything, you hire someone who knows the thing you want. And if that's the way the AI models go, I think we see a lot more competition and a lot less dominance. But if in fact lighting fire to huge piles of money means you have a better AI, then the big companies, and you're right, right, the big American, Californian, you know, white guy companies will dominate.
And presumably then, that is a question that must be keeping a lot of people in Silicon Valley up at night. This idea that maybe the big AI bet won't pay off. What happens if it doesn't?
>> You know, I mean, we've seen this play out in the dot-com bubble. And if you remember, like in the early 2000s, the internet becomes a big thing, the web becomes big, and all these really dumb companies are formed and they go out of business because they didn't make any sense. But the internet was valuable and new companies arose that understood how to use the technology. I think we see the same thing here, like these companies which have enormous amounts of debt and really no viable business model right now, they die and other companies appear, right? The Anthropics, the Open AIs are betting this doesn't happen and they're running as fast as they can with that singular bet. I don't know if they're losing sleep because I think they're just running as fast as they can.
>> Well, Bruce Schneider, technology and cyber security expert and best-selling author of Rewiring Democracy, who lectures in public policy at the Harvard Kennedy School. Thank you very much for joining me today on the Tech Report.
>> Thanks for having me. This is a fascinating conversation and we'll do it again in a month and the answers will all be different.