Future of Tech

There are a handful of genuinely useful claims - task-specific small models outperforming large general-purpose models, the gap between mass experimentation and mass deployment, and the limits of LLM arithmetic - but they are buried in extended biographical preamble, AI-history recap, and filler. The ratio of novel claims to total airtime is low for a 26-minute episode.

“what we've pioneered is what we call task-specific models. These are models that, by the way, they also tend to be small, but they're optimized for certain tasks like summarizing a text, answering questions in a rag style”

“we've moved from sporadic experimentation to mass experimentation, everybody experimenting, not mass deployment”

The episode largely recycles well-worn AI narratives - the AI winter, data-and-compute as the core driver, hallucination risks - without offering a genuinely contrarian or first-principles frame. The 'false intimacy' observation is interesting but is explicitly attributed to Harari, and the dislike of the term 'generative AI' is a mildly novel framing that goes nowhere substantive.

“There's no such thing called generative AI. There are various terms that I dislike.”

“Yuval Noah Harari, our compatriot, spoke about the danger of false intimacy. That's actually a very interesting observation”

Shoham is genuinely high-caliber - 28 years directing Stanford's AI Lab, co-founder of AI21 Labs, originator of the AI Index, and a serial practitioner entrepreneur - but the conversation extracts only a fraction of what his depth could offer, keeping him largely in explainer mode rather than expert mode.

“I studied computer science by mistake just because I needed to study something”

“We have a very close collaboration with both GCP and AWS. NVIDIA is an investor in us.”

A few concrete anchors appear - WordTune's 10 million users, training scale moving from 300 billion to multi-trillion tokens, and a named financial-institution summarization bake-off - but most enterprise examples are anonymized and lack performance metrics, dollar figures, or timelines that would make them actionable for a B2B operator.

“well beyond 10 million users”

“In the past, we trained on 300 billion tokens... Today, we don't get out of bed for that. It's multi-trillion tokens.”

The host's questions are consistently surface-level - 'give me a glossary,' 'tell me a few words about precision,' 'how do you balance personal life' - and there is no meaningful pushback, probing follow-up, or challenge to any of Shoham's claims throughout the episode. The interview functions as a PR-friendly profile rather than a substantive dialogue.

“Now, tell me a few words about precision. How important is it?”

“You have one personal question. You being an entrepreneur for many years, how do you balance your personal life”