Judea Pearl on Why LLMs Won’t “Scale” to AGI — and What Comes Next (Causal World Models)

Judea Pearl on Why LLMs Won’t “Scale” to AGI — and What Comes Next (Causal World Models)

Judea Pearl argues that today’s transformer-based LLMs have mathematical limits that prevent them from internalizing causal world models, so “just scaling” them won’t deliver AGI. But does that mean LLMs are a dead end? In this episode, I break down what Pearl actually said, why the Internet misquotes it, and why the path to artificial general intelligence may come from a simpler breakthrough that produces “good-enough” causal reasoning, just like the transformer algorithm did for language.


Topics: AGI, causal AI, causal inference, Judea Pearl, transformers, LLM limitations, world models, scaling laws, reasoning, next-gen AI architectures.

Denne episoden er hentet fra en åpen RSS-feed og er ikke publisert av Podme. Den kan derfor inneholde annonser.

Episoder(33)

AI Adoption, Human Psychology, and the Road to AGI

AI Adoption, Human Psychology, and the Road to AGI

Is AI a new kind of intelligent species? This episode examines how humans are coping with the proliferation of AI, why AI adoption is now a competitive necessity, and what recent research reveals abou...

6 Apr 5min

AGI Is Here? Jensen Huang, Lex Fridman, and the Real Definition of AGI

AGI Is Here? Jensen Huang, Lex Fridman, and the Real Definition of AGI

Did Jensen Huang really say AGI is already here? Not quite. This episode dissects the viral claim, explains what Huang actually meant, and lays out a sharper definition of AGI based on one core capabi...

31 Mar 8min

Taste and Liability: Software Engineering After AGI

Taste and Liability: Software Engineering After AGI

Will artificial general intelligence wipe out software engineering jobs, or just transform them? In this episode of This Is AGI, I compare the future of software development to commercial aviation, sh...

23 Mar 10min

Why the Pentagon’s Anthropic Scandal Could Be Bigger Than It Looks

Why the Pentagon’s Anthropic Scandal Could Be Bigger Than It Looks

Did the Pentagon’s clash with Anthropic expose a deeper AI security crisis? This episode explores the Anthropic’s supply-chain risk designation by the Pentagon, what it could reveal about the state of...

16 Mar 6min

Software Engineering Taste Won’t Save Jobs from AI: Clean Code, AGI, and the Future of Programming

Software Engineering Taste Won’t Save Jobs from AI: Clean Code, AGI, and the Future of Programming

Will human “taste” in software engineering protect developers from AI job displacement? In this episode of ‘This Is AGI’, Alex Chadyuk argues that code elegance and architectural judgment are becoming...

9 Mar 7min

AGI Isn’t Needed to Replace Office Jobs: The Real Threat from LLM Agent Automation

AGI Isn’t Needed to Replace Office Jobs: The Real Threat from LLM Agent Automation

Everyone is waiting for AGI, but most office jobs don’t require general intelligence to automate them. In this episode, I explain why decades of hyper-specialization in white-collar work makes it uniq...

2 Mar 4min

Agentic AI in the Enterprise: The New Management Skillset for the Age of AGI

Agentic AI in the Enterprise: The New Management Skillset for the Age of AGI

As AGI moves from theory to deployment, the definition of “leadership” is shifting fast. In this episode of This Is AGI, Alex Chadyuk explains why the AI era is bringing back the classic general manag...

16 Feb 5min

Populært innen Teknologi

teknisk-sett
lydartikler-fra-aftenposten
energi-og-klima
rss-ki-praten
elektropodden
hans-petter-og-co
smart-forklart
rss-alt-som-gar-pa-strom
rss-snakk-om-sikkerhet
fornybaren
shifter
rss-ai-forklart
tomprat-med-gunnar-tjomlid
rss-teknologioptimistene-en-podkast-om-teknologi-og-mennesker
teknologi-og-mennesker
pedagogisk-intelligens
nasjonal-sikkerhetsmyndighet-nsm
rss-alt-vi-kan
rss-ki-til-kaffen
plattformpodden