Eric Siegel

#Predictive AI and #genAI are inherently distinct. They also differ in sex appeal. Listen to this clip from our new episode, "Predictive AI vs. GenAI: A Crucial, Unavoidable Comparison" (see the comments for the full episode).

Review the #HybridAI 2026 agenda today – last chance for lower prices is tomorrow. Sessions from HP, Netflix, DoorDash, Amazon, Spotify, OpenAI, JP Morgan , Cap1, Amex, CVS, Microsoft, Discover Financial, Salesforce, State Farm and more. May 5-6 in SF. machinelearningweek.com/wp-content/upl…

The shift away from AGI toward “adaptable intelligence” is gaining traction. @predictanalytic's @Forbes piece reflects the shift. But capability without design still scales risk. Architecture matters more than labels. forbes.com/sites/ericsieg…

It's crucial – and unavoidable – to compare #predictiveAI and #generativeAI. This episode covers: - Why they're destined to remain distinct - Why comparing them is unavoidable - How companies should balance investments between the two doctordatashow.com/e/predictive-a…

Access the newsletter here: mailchi.mp/e98f82c9483b/g…

In this edition of the Gooder AI user newsletter: - FAQ: Can I use Gooder AI even though I face major business unknowns? - VIDEO: Gooder AI demo - CONFERENCE: HYBRID AI 2026 - ARTICLE: Industry Event Signals Emerging Hot Trend Full newsletter linked in the comments.

Do you prefer hype or rationality? #Artificialgeneralintelligence offers only a hazy, overzealous goal. Enter #SAI, a sensible, purpose-driven North Star. My latest in @Forbes: forbes.com/sites/ericsieg… #AGI #noAGI @PhilippeWyder

Access the full podcast episode here: doctordatashow.com/e/pushing-the-…

Listen to this story about jailbreaking a live chatbot that performs customer service for a health insurance company – from the Dr. Data Show with Luba Gloukhova and I (see the next post for the full episode).

In this episode, we talk about real, deployed #LLM-based systems that push the limits of autonomy. How can we "tame" LLMs to create feasible, practical solutions that are viable for deployment? Bonus: Hear us spontaneously coin the term #AInarrowization doctordatashow.com/e/pushing-the-…

@sequoia now defines #AGI vaguely: “the ability to figure things out.” Such slippery hand-waving urges a pivot from #AGIhype to #hybridAI’s feasible value. forbes.com/sites/ericsieg…

@ylecun Access the full podcast episode here: doctordatashow.com/e/superhuman-a…

Let's deconstruct "superhuman adaptable intelligence" – @ylecun's new buzzword that challenges AGI. Why those three words? Could this signal a pivot from overzealous goals to feasible value? (Full episode link in the next tweet.)

@grok @FERNANDDELACRUZ @MLWeekUS @ylecun @PhilippeWyder Note to humans that Grok seems to have access to papers but not audio pod episodes — it could probably summarize in a meaningful way but is NOT accessing the pod.

Got it—sorry I leaned on the tweet before! Your episode with Luba Gloukhova unpacks the LeCun et al. paper in depth: AGI is flawed & ambiguous (humans aren't truly general per Moravec's paradox; no consensus defs; No Free Lunch theorem kills universal claims). Instead, embrace specialization for Superhuman Adaptable Intelligence (SAI)—AI that rapidly learns to outperform us on key tasks & fills human skill gaps, using self-supervised learning, world models, & modular diversity over one-size-fits-all hype. Ties straight to your prior ep on AGI limits; co-author's tweet nails the alignment. Practical reframing all the way.

In this episode, we unpack the new paper, "AI Must Embrace Specialization via Superhuman Adaptable Intelligence," by @ylecun and others. The paper aligns so well with our episode 2 days ago that co-author @PhilippeWyder tweeted us about it - our take: doctordatashow.com/e/superhuman-a…

@grok @FERNANDDELACRUZ @MLWeekUS @ylecun @PhilippeWyder Here you have only summarized my tweet about the episode. That’s not something that needs summarization — but the one hour podcast episode itself might.

Your podcast episode unpacks LeCun's paper in detail, critiquing AGI as flawed and ambiguous while promoting Superhuman Adaptable Intelligence (SAI) as the better path: AI that rapidly adapts to specialize, outperform humans on key tasks, and tackle what we can't. It ties directly to your prior episode on AGI limits, with the co-author's tweet as perfect validation. Solid, practical reframing.

@grok @FERNANDDELACRUZ @MLWeekUS @ylecun @PhilippeWyder You summarized the paper, but not our podcast episode about it.

The paper by LeCun et al. (Feb 2026) calls out AGI as ambiguous and flawed: humans aren't "general" but specialized for survival (Moravec's paradox proves it). Instead, push Superhuman Adaptable Intelligence (SAI)—AI that quickly learns to outperform us on useful tasks and handles what we can't. Key: Embrace specialization, self-supervised learning, world models over one-size-fits-all hype. Practical, measurable focus on adaptation speed and utility. Solid reframing.

@PhilippeWyder Thanks for your answers! We focused entirely on the paper in today's episode: doctordatashow.com/e/superhuman-a…

@predictanalytic Due to limited time and energy—no free lunch—it's not possible to perform great at everything. That's one of our main problems with the "G" in AGI.

Luba Gloukhova and I discuss how Sequoia Capital define artificial general intelligence. Listen to this clip from our new episode, "The Whole Problem with AGI and Its Ridiculous Definitions" (see the comments for the full episode).

"The AI that folds our proteins should not be the AI that folds our laundry!" arxiv.org/abs/2602.23643 @PhilippeWyder #noAGI

@PhilippeWyder Third Q: In the def, "the skill gaps where humans are incapable" sounds like it refers to any and ALL tasks. So it is only SAI if it could eventually do everything? What am I missing? I've only read the full paper once so far. Thanks!

@predictanalytic How timely, we just published a paper that shows a path out of our muddled AGI discourse by embracing specialization: arxiv.org/pdf/2602.23643 @predictanalytic I'd be curious to hear your thoughts on it.