Ben Lorica

Things Ben Says on Podcasts

Data scientist covering AI infrastructure, foundation models and enterprise AI adoption. Founder of Gradient Flow and former Chief Data Scientist at O'Reilly Media, where he was also program chair for the Strata Data Conference, the O'Reilly Artificial Intelligence Conference and TensorFlow World. He holds a PhD in mathematics from UC Santa Barbara and was previously an assistant professor at UC Davis. Hosts The Data Exchange podcast and writes the Gradient Flow newsletter, and has appeared as a guest on Practical AI, the TWIML AI Podcast and the AI in Business Podcast.

Where to Find Them

Ben Lorica hosts and writes The Data Exchange with Ben Lorica and writes Gradient Flow . They have also been a guest on Practical AI , The AI in Business Podcast and The TWIML AI Podcast .

Recently: “The AI Boom Is Impressive. Can It Last?” on The Data Exchange with Ben Lorica (October 2026); “When AI Writes the Code, Who Proves It Works?” on The Data Exchange with Ben Lorica (October 2026); “Let Domain Experts Build the Agents” on The Data Exchange with Ben Lorica (October 2026); “AI Is Coming for Tasks Before It Comes for Jobs” on The Data Exchange with Ben Lorica (October 2026); “Why Build Your Own AI Model When Frontier Models Already Exist?” on The Data Exchange with Ben Lorica (October 2026); “Why AI Makes Games Worse—and How to Make Them Better” on The Data Exchange with Ben Lorica (October 2026).

What They Said

Listing his bear-case worries from the Anthropic S-1, including falling prices as open-weights models improve, the host offers this analogy for how open models relate to the frontier labs.

The Data Exchange with Ben Lorica · 2026-10-31 Permalink → Listen →
The Data Exchange with Ben Lorica Around 15:23 into the episode
Ben Lorica

…have to incur all of these expensive costs around training, right? Yes. Yes. And by the way, before you continue there, but if you talk to anyone in enterprise, they'll tell you, quite frankly, look, if they don't release another model for the next six to 12 months, we have more than enough to do.

Evangelos Simoudis

…this type of model switching that even that has been assumed because there is going to be updates, obviously. Every application does it, but it's not going to be at the pace that it is anticipated. So I think that will slow down. And the question then becomes, what's the mode for these companies?

Ben Lorica

Right, right. And so again, the third item I don't like is that prices are heading down. And you can see this through all sorts of third-party data, right? So the Chinese labs are producing tokens at an industrial scale, releasing their models as open weights. And I think in the next six to 12 months, we'll... Have respectable Western open weights models, right? So NVIDIA is starting to buy Neolabs to beef up Nematron. Reflection has an open weights model. Thinking machines will probably continue to release open waste models. And then I know some initiatives that are going to go do kind of a full open source model initiative, right? So I think the pressure from some notion of open is real. And people always say, well, the open weights are just distilling the proprietary models, right? But then as long as the proprietary models keep improving, the open waste models are going to improve. So the analogy that I've come to like is the water skiing, right? So a water skier is being pulled by a speedboat. So the water skier can go as fast as the speedboat or maybe even faster, but can never go get ahead of the speedboat.

Evangelos Simoudis

That's right. That's right.

Ben Lorica

So, you know, prices are heading down. I think that's going to continue, right?

Evangelos Simoudis

Yeah.

Speaker names come from our own diarization. The position is estimated from where the line sits in the episode.

Collections They Appear In