Listen To The Show

Transcript

Welcome back to The Prompt by Kuro House — your daily dive into the world of artificial intelligence and the people, politics, and power shaping its future. Today, we’ve got a lineup that spans ancient Latin, modern geopolitics, and the murky relationship between climate tech and fossil fuels. Let’s get into it.

First up, let’s talk about the Trump administration’s newly released AI Action Plan — and why MIT Technology Review is calling it a distraction. The plan positions itself as a bold step toward AI innovation, but the article argues it’s more about political posturing than policy substance. The backdrop here is America’s historic advantage in AI — an advantage largely built on immigration. Of the eight researchers behind the 2017 transformer paper that underpins models like ChatGPT, six were born outside the U.S. That’s not a fluke. According to the Institute for Progress, 60% of the top 50 AI startups in the U.S. have at least one immigrant cofounder. The founders of OpenAI, Anthropic, Nvidia, and Google? Many are immigrants or children of immigrants. But now, that talent pipeline is under threat. Anti-immigration policies and cuts to R&D funding are beginning to reverse the brain drain — away from the U.S. Instead of doubling down on what made America a leader in AI, the administration’s plan focuses on deregulation and nationalism. The article reminds us that innovation thrives not just on talent, but on freedom — the freedom to leave, to start anew, to challenge incumbents. That’s how Silicon Valley was built, from the “Traitorous Eight” to the PayPal Mafia. The fear is that this new plan forgets that legacy.

And that brings us to our second story, also from MIT Technology Review, which looks at the weakening of America’s AI watchdog — the Federal Trade Commission. The FTC under Biden took aggressive action against AI companies making false claims or harming consumers. It cracked down on Evolv, whose AI-powered security failed to detect a knife that was later used in a school stabbing. It fined startups for fake AI-generated product reviews and bogus AI legal services. These weren’t billion-dollar penalties, but they did stop deceptive practices and offered consumers a path to refunds. Now, the Trump administration is signaling that it wants to roll those actions back. In the same AI Action Plan, the White House says it will review all FTC actions from the past four years and cut federal AI funding to states with “burdensome” regulations. The message is clear: regulation is seen as a threat to innovation. But this isn’t just about red tape. It’s about whether there will be any meaningful accountability for AI systems that hurt people. The FTC is one of the few agencies that has actually tried to hold AI companies responsible. But with the firing of Democratic commissioners — one of which a court has already ruled was illegal — and promises to rein in independent agencies, the FTC’s future role in AI oversight is looking increasingly uncertain.

Now let’s shift from politics to archaeology, where Google DeepMind has unveiled a new AI tool that’s helping historians decode ancient Latin inscriptions. It’s called Aeneas, and it’s designed to analyze partial transcriptions and images of weathered stone engravings. Aeneas doesn’t just fill in missing words — it cross-references a database of nearly 150,000 inscriptions from across the Roman Empire to suggest when and where an inscription was likely made. For example, if it sees “…us populusque Romanus,” it might suggest “Senatus” comes before it, forming the classic phrase “Senatus Populusque Romanus” — the Senate and the People of Rome. It builds on DeepMind’s earlier tool, Ithaca, which helped reconstruct Greek inscriptions. But Aeneas goes a step further by offering historians hypotheses and parallels that can spark new research. In tests with 23 historians, Aeneas improved the accuracy of dating and locating inscriptions and even inspired new lines of inquiry in 90% of cases. It’s not meant to replace historians, but to work alongside them. That said, its real-world utility remains to be seen. It doesn’t interpret meaning, and its performance on lesser-known inscriptions is still untested. But the potential is huge — not just for understanding the past, but for how AI can augment human expertise in highly specialized fields.

And speaking of specialized fields, let’s talk about geothermal energy and the surprising alliances forming in climate tech. In a pair of stories from MIT Technology Review, we get a close look at Quaise, a geothermal startup that’s partnering with Nabors Industries — one of the world’s largest oil and gas drilling firms. Quaise is developing new drilling tech that could unlock deep geothermal energy, and Nabors has been involved since 2021, investing $12 million and now serving as a technical partner. “We are agnostic to what hole we’re drilling,” says Cameron Maresh from Nabors’ energy transition team. That might sound cynical, but it’s also practical. Oil and gas companies have the equipment, the expertise, and the infrastructure to scale geothermal quickly — if they choose to. According to the International Energy Agency, about 30% of a future net-zero energy system could rely on technologies where oil and gas know-how is useful: hydrogen, biofuels, carbon capture, and yes, geothermal. But here’s the catch: in 2022, oil and gas companies made up just 1% of global climate tech investment. That number’s rising, but slowly. These companies could be key players in the energy transition — or they could be hedging, investing just enough to stay relevant while continuing to profit from fossil fuels. The jury’s still out.

And finally, let’s return to the DeepMind story for a moment, because it’s not just about technology — it’s about method. Aeneas was trained on a relatively small dataset compared to models like Gemini or GPT. Just a few thousand images and 150,000 inscriptions. That’s tiny in AI terms. But it shows how targeted, domain-specific models can still be incredibly powerful when paired with expert workflows. The team behind Aeneas isn’t trying to automate epigraphy — they’re trying to enhance it. And that’s a lesson that applies far beyond ancient Latin: AI doesn’t have to replace humans to be transformative. Sometimes, the best use of AI is to make people better at what they already do.

That’s it for today’s episode of The Prompt. Whether it’s decoding ancient Rome or navigating modern regulation, AI is reshaping how we understand both the past and the future. Thanks for joining us — and as always, stay curious.