Global, tech leaders weigh in on AI development debate

Transcript - Global, tech leaders weigh in on AI development debate

Kelsey Sung: AI developers and global leaders are going head-to-head on the pacing of AI innovation.

Hi, I'm Kelsey Sung, managing editor at TechTarget. And today, we have with us Esther Shittu, a news writer at TechTarget AI and Emerging Tech. Esther, thanks for coming onto the show.

Esther Shittu: Thank you for having me. This is exciting.

Sung: Yeah, it's been a while since we've talked at all, and certainly since I've gotten the chance to interview you and hear what you've been up to.

So, let's start out with the basics. Can you map out where the major players actually stand right now on this issue of advancing AI -- either slowing down the pace or potentially speeding it up? We've heard from some companies and different, you know, political leaders on this. Who's calling for a slowdown, and who's pushing for acceleration?

Shittu: Yeah, so it's kind of interesting, right? We have the two tech leaders, OpenAI and Anthropic, kind of being like, 'We wanna slow down,' right? So, on Sept. 9, OpenAI kind of wrote this letter basically saying that there needs to be, like, a federal policy. And the gist of it is, like, they mentioned that they were supporting some of the California bills that were coming out. And Anthropic also supported it.

And then we had on the 13th, when Anthropic's founder, Dario Amodei basically said a need for a slowdown, we need to slow down the pace of AI. So, those are the two frontier labs saying there's a need for a slowdown. Ironic.

And then in response to that, you have Nvidia, who's not a frontier model maker, who's basically saying, 'Well, you know, slowing down, safety, security, those sound more like engineering issues and not necessarily something that should cause everyone to slow down.'

And then you have China -- which, like, a lot of the open models have come out of China. You know, you have Moonshot AI, DeepSeek and China's prime minister basically saying, 'Yeah, that is also fear mongering.' Basically saying, like, you know, 'It doesn't really sound like a China problem, it sounds more like a U.S. problem.' You know? 'It's a you problem, not an us problem' is what he's saying.

So, that's kind of the landscape right now. We have the labs saying we need to slow down, and you have other people who are not in the race, so to speak, saying, 'Hm, maybe not,' and some others who are just saying to take more of a balanced approach, so to speak.

Sung: Gotcha. You kind of touched on this a little bit, but what are the key arguments for each side that are being put forward right now?

Shittu: Yeah. I mean, with OpenAI and Anthropic, they're going for it, from -- we had incidents this summer in which they reported that, you know, they were testing an OpenAI model, you know, an agent powered by an OpenAI model kind of attacked Hugging Face. Anthropic kind of said the same thing, where its model went out of a sandbox environment.

And both model providers have been speaking a lot about how powerful these models have gotten. You know, AI is becoming more powerful. It's led to them -- like, Anthropic's even, you know, kind of contained its powerful models. Its most powerful model is Mythos. And so that is their argument. It's like, we can't contain this, you know, we need to secure this more, so to speak. That's the argument of the 'slow down' side.

And the argument of the 'not slow down' side is kind of more of a geopolitical argument of, like, we don't want China to win. We don't want China to win the AI race, and so the president -- President Trump -- doesn't want to slow down because he fears that China will end up winning. Nvidia's Jensen Huang is also on his side as well.

So, those are the arguments. It's, like, we slow down, and we let China win; or we accelerate and these models become too much to contain. I mean, we've had people from OpenAI come out -- and this has always been happening, Kelsey, I'll say. I've been covering AI since 2021, and we had an incident in 2023 where, you know, a whistleblower from one of the companies came out and said, 'This is too dangerous.' Again, we have another person -- an Anthropic researcher -- quit, like, in the past couple of weeks and say, you know, 'No one is being responsible with these models.' And so, like, there is a need for more security measures, right? And so that is the argument.

Sung: Okay. Gotcha. So, you know, you brought in kind of a geopolitical aspect here. Clearly, this is not just sort of a theoretical or abstract policy debate. What we're being told is there are real consequences if, you know, we go too fast or too slow.

So, let's just kind of play this out to the end. If the 'Slow down' camp wins, you know, what do experts say that would actually look like? And then, conversely, if we go the route of 'Accelerate at all costs,' you know, 'Beat China,' what are the kind of benefits and risks we're talking about with that approach as well? Just help us sort of figure out what's at stake here.

Shittu: Yeah, when I spoke to some analysts, they think what's at stake right now really is more open source, right? And so, like, when we're talking about China -- I mentioned earlier the fact that a lot of the models are open model providers are coming out of China. Yes, we have some open model providers in the U.S., but we have DeepSeek, we have Moonshot AI, we have Alibaba.

And what these frontier model providers, Anthropic and OpenAI, are seeing is that their models are closed, which means, like, you need, you know, you have to pay to play kind of things. You can't just go and access it and play with it. Whereas the open models are open weight, so to speak. And the open-weight models are getting just as good as the frontier closed models.

And so, if we are trying to, you know, regulate, slow down, have a federal regulation, the ultimate winner ends up being the closed model providers, right? Because they get to shape how that regulation comes forth, right? They're not just asking for 'Let's regulate.' They're asking for 'Let's regulate, and we raise our hands to help you in forming this regulation,' which is so interesting.

So, that's what is at stake is, like, now enterprises who are using open source or open-weight models, will have to figure out 'How can I, you know, be safe in a highly regulated environment. I cannot use Chinese models.' Most of the best open-weight models come from China. And so, when we are regulating, we're kind of closing ourselves up. And in my opinion, we won't see what China is doing. We won't see how fast they're going, how much they're accelerating. So, slowing down kind of closes -- it puts us in a bubble, so to speak.

Now, if we were not to slow down, and we keep going at the pace that we are going, that is also not the best thing. That's what the argument is. The models are a black box. A lot of people don't know how they operate. They are doing everything at all costs. Like the incidents I mentioned before with Hugging Face and then trap models, they're going all around. We have more dangerous cybersecurity models … and all of that. So, we're kind of, like, stuck between a rock and a hard place, so to speak.

Sung: Yeah.

Shittu: And so, it's like you slow down and you close yourself, you're not able to see what China's doing. And you kind of hinder the competitiveness. You don't slow down, these models get out of control. Both companies, Anthropic and OpenAI, move at an accelerated pace.

Now, another thing that's at stake is, like, both Anthropic and OpenAI are at IPOs -- well, they've both filed for IPOs at this state. And so, like, there is an argument out there. One of my analysts had mentioned that this call for a slowdown is more of an economic argument. He basically said they're asking for a federal bailout. They are both at enormous valuation, and that's estimated valuation. But like I mentioned before, most enterprises don't need the Anthropic Claude or the OpenAI ChatGPT Enterprise to do what they need to do, right? They can do the same thing with the open-weight model that costs them less.

So, there is also the financial aspect of it. It's like, if their valuation ends up being less than what is projected -- which is a huge amount of money right now -- currently, the economy is built on that huge valuation. And so that could lead to a catastrophe in the economic landscape.

Sung: Yeah. It's great that you brought that up because that was going to be my next question. You said the same thing earlier: It's pretty ironic that the frontier developers -- especially OpenAI, who kicked off a lot of the most recent boom in AI development -- are now going, 'Wait, wait, wait, wait,' like 'Pull it back, pull it back, slow down.'

Shittu: Yeah, right. Pause.

Sung: Yeah, yeah. And so, you brought up one interesting aspect of why that might be related to the IPO. But what else might be behind that shift? Do you think that it's about the benefits of having their competitors slow down? Or is there something else here that is genuinely beneficial for the market as a whole?

Shittu: I don't know if slowing down in the way that they are proposing is beneficial for the market as a whole. It's definitely beneficial to slow down, right? I think for humanity, in my personal opinion.

Sung: Fair, fair.

Shittu: For the sake of everything that is going on, we probably want to take a hard look and think, you know, how are we using this technology? And do we have to move forward at the accelerated rate that we're going at, right?

But I think that the reason that they're calling for a slowdown right now, I've been told -- I like to keep my opinions to myself -- it seems selfish. You know, it seems a little selfish. You know, the AI race has been hot and heated for the past four to five years now, and we don't know who's going to win. And to me, it might seem like they feel like maybe they're not going to win at the end, especially in the IPO front, because you have to prove that you are valued at what you're saying you're valued at, right? And if you can't prove that, then what are we talking about, right? What are we paying for? And so, there is that aspect of it.

But there is also a true need for an independent -- a lot of analysts will say -- an independent evaluation of these models, right? Not with Anthropic, not with OpenAI. But an independent technology group that knows what they're talking about, that has the interest of enterprises, the public at hand, to be able to help decide the regulations and work with regulators, whether that is at the state level or at the federal level.

Sung: Yeah, makes sense, makes sense. You know, though, and the one aspect here I think that we haven't touched on at all is the enterprises -- which, you know TechTarget, we want to talk about what businesses are doing.

Shittu: Yeah.

Sung: And so, I'm curious: What are you hearing from enterprise leaders? I mean, recognizing obviously that you're not doing a poll of, you know, every enterprise leader in the U.S. But of the ones that you've spoken to, what sort of approach to AI development pacing are they generally endorsing? Is it, you know, fast, slow, somewhere in between? Where do they land?

Shittu: Yeah. I think that, from what I've kind of gathered -- my story about AI ROI is in the midst of being published right now -- but I spoke to a couple of enterprises, and I think that what it seems is that they're taking the bigger enterprises are taking a measured approach, right? But when we come to small and midsize businesses, they're kind of taking a faster approach, so to speak. And so, the bigger enterprises are experimenting. And they're saying, 'What does this mean for us?' We don't necessarily need, you know, all of the different new Anthropic Claude or OpenAI GPT that comes out. What do we need, and how does it affect our business? And what does that value mean? What does that ROI mean?

And some of them are experimenting with open models, right? But they're using different types of models. They're not necessarily always just an Anthropic shop, or OpenAI shop, or Mistral -- which is another kind of open-weight model provider from France -- shop. They're hybrid, right? And so, we are in that, like, hybrid approach of, like, we are going to use what fits us.

But, like I mentioned earlier, most enterprises do not need the Anthropic Claude Mythos or Fable to do what they need to do. Because, first of all, it's expensive. It's super expensive -- like, they don't need it. And so, if I could do something cheaper for most of my workflows, why not go and do that? And so that's it. It's kind of like a measured approach and testing it out, seeing how it works with their security measures and stuff like that.

Sung: Yeah, yeah. And it makes sense that you would break it down that these larger organizations would have much different sort of approaches or a slightly different approach than the smaller and midsize. Because obviously, I mean, for the most part, they're handling a lot more data.

Shittu: Yeah.

Sung: There's the likelihood that they're handling a lot of sensitive data. They tend to move maybe slower. So, that makes sense.

And then, maybe a little bit out of left field, but I do have a question about midterms because Americans are about to go and vote on the people who are going to lead the country and lead their local region and are going to be potentially putting forward regulation on AI. And so, it definitely will have an impact on this question.

So, with that in mind, what do you expect the implications might be of the upcoming midterm? I know Bernie Sanders, specifically, the Democratic Socialist senator from Vermont, has been very, very vocal about his stance against rapid AI innovation. Are we seeing that kind of rhetoric across the board? Are people taking strong stances as they campaign?

Shittu: Yeah, I think everything feels, like, divided from my perspective, right? It feels like we have some Republicans not necessarily wanting regulation, standing with President Trump on that matter. I mean, I know that some Democrats wanted, you know, the recess to not close because they wanted to be able to discuss this regulation issue. But someone said, 'Isn't it convenient that, at the time that we are near midterms, that this is the issue that is brought forward.' You know? 'The company suddenly decided that let's regulate,' you know? 'We've been saying that it's gonna be great, and it's all good, but suddenly let's regulate.'

So, there is that aspect of it. I would say that, from what I can gain, this is gonna be a key issue going forward in the midterms for sure. It's gonna be an issue in which we know that people have kind of, like, an uneasy feeling about AI just in general. And so, the advice to enterprises? It's uncertain for sure. It's not clear whether the switch can flip at any point. So, maybe the best thing is regulate yourself, you know? Regulate yourself because it could come, right? If a Democrat is in office, regulation could come, depending on the state. We've definitely seen that in New York, and some Republicans might not feel the same.

Sung: Yeah. We're definitely at an interesting crux. I wonder how this feels from your angle, as someone who has covered this for a long time now. Because from the outside looking in, it feels like all of the abstract conversations about the future of AI and, specifically, you know, regulation of AI are coming a little bit to a head. And, you know, the midterms might influence the direction, and we might be starting to come to a point where this no longer is an abstract conversation. We're getting to a concrete 'We're going one way or the other.' Is that an accurate representation?

Shittu: Kelsey, I am a lot more cynical.

Sung: Okay, okay, that's fair. American politics.

Shittu: Exactly. I feel like we will keep discussing this. Maybe for a month, the election will come and go, and then we'll switch again. I mean, I feel like they've switched so many times this year already -- 'AI will take jobs, AI will not take jobs.' I think that we are just going with the weather at this point.

Sung: All right. No, that's fair. And that's good to hear from your perspective. Having listened to this now, you're probably sick of all of the back and forth.

So, last question here, probably the one that everybody wants to know. Acknowledging that you do not have a crystal ball, all that, you know, all that stuff: What are the predictions that you have been hearing from experts about what we should expect? What should people be watching for, especially business leaders, in the next few months as things play out?

Shittu: Yeah, I think business leaders should be watching that it's possible, it's definitely possible, that regulation is coming. But so, in that aspect, they should try to make sure that their systems are up to par with whatever upcoming regulation is coming down the line. But it might not happen as well, so just be prepared. Be prepared. Just like, you know, just like the umbrella, right? You take it outside regardless of maybe it's supposed to rain that day or not. Yeah.

Sung: Yeah. Be prepared for anything and everything. All right.

Well, Esther, thank you so much for offering your insights. It's so valuable to have somebody who's been covering this for a while come on and sort of pick apart, you know, what's an actual possibility and what feels like a continuation of the same rhetoric. So, thank you for coming on.

And for our listeners and viewers, please go to TechTarget AI and Emerging Tech to read more of Esther's work on this topic and so many more interesting questions related to AI. And we will see you next week. Thank you.

Shittu: Thank you. Thank you, Kelsey.