Menu
Sign In Search Podcasts Charts People & Topics Add Podcast API Blog Pricing

Francois Chollet

👤 Speaker
See mentions of this person in podcasts
649 total appearances

Appearances Over Time

Podcast Appearances

Dwarkesh Podcast
Francois Chollet, Mike Knoop - LLMs won’t lead to AGI - $1,000,000 Prize to find true solution

And you could say this is reasoning.

Dwarkesh Podcast
Francois Chollet, Mike Knoop - LLMs won’t lead to AGI - $1,000,000 Prize to find true solution

And I say, yeah, sure, okay.

Dwarkesh Podcast
Francois Chollet, Mike Knoop - LLMs won’t lead to AGI - $1,000,000 Prize to find true solution

But another definition you can use is reasoning is the ability to, when you're faced with a puzzle, given that you don't have already a program in memory to solve it,

Dwarkesh Podcast
Francois Chollet, Mike Knoop - LLMs won’t lead to AGI - $1,000,000 Prize to find true solution

You must synthesize on the fly a new program based on bits of pieces of existing programs that you have.

Dwarkesh Podcast
Francois Chollet, Mike Knoop - LLMs won’t lead to AGI - $1,000,000 Prize to find true solution

You have to do on-the-fly program synthesis.

Dwarkesh Podcast
Francois Chollet, Mike Knoop - LLMs won’t lead to AGI - $1,000,000 Prize to find true solution

And it's actually dramatically harder than just fetching the right memorized program and replying it.

Dwarkesh Podcast
Francois Chollet, Mike Knoop - LLMs won’t lead to AGI - $1,000,000 Prize to find true solution

Sure.

Dwarkesh Podcast
Francois Chollet, Mike Knoop - LLMs won’t lead to AGI - $1,000,000 Prize to find true solution

I mean, in order to do on-the-fly program synthesis, you actually need building blocks to work from.

Dwarkesh Podcast
Francois Chollet, Mike Knoop - LLMs won’t lead to AGI - $1,000,000 Prize to find true solution

So knowledge and memory are actually tremendously important in the process.

Dwarkesh Podcast
Francois Chollet, Mike Knoop - LLMs won’t lead to AGI - $1,000,000 Prize to find true solution

I'm not saying it's memory versus reasoning.

Dwarkesh Podcast
Francois Chollet, Mike Knoop - LLMs won’t lead to AGI - $1,000,000 Prize to find true solution

In order to do effective reasoning, you need memory.

Dwarkesh Podcast
Francois Chollet, Mike Knoop - LLMs won’t lead to AGI - $1,000,000 Prize to find true solution

It's not really a size issue.

Dwarkesh Podcast
Francois Chollet, Mike Knoop - LLMs won’t lead to AGI - $1,000,000 Prize to find true solution

It's more like a train data issue in this case.

Dwarkesh Podcast
Francois Chollet, Mike Knoop - LLMs won’t lead to AGI - $1,000,000 Prize to find true solution

No, it's not.

Dwarkesh Podcast
Francois Chollet, Mike Knoop - LLMs won’t lead to AGI - $1,000,000 Prize to find true solution

If you scale up your database and you keep adding to it more knowledge, more program templates, then sure, it becomes more and more skillful.

Dwarkesh Podcast
Francois Chollet, Mike Knoop - LLMs won’t lead to AGI - $1,000,000 Prize to find true solution

You can apply it to more and more tasks.

Dwarkesh Podcast
Francois Chollet, Mike Knoop - LLMs won’t lead to AGI - $1,000,000 Prize to find true solution

But general intelligence is not task-specific skill scaled up to many skills.

Dwarkesh Podcast
Francois Chollet, Mike Knoop - LLMs won’t lead to AGI - $1,000,000 Prize to find true solution

Because there is an infinite space of possible skills.

Dwarkesh Podcast
Francois Chollet, Mike Knoop - LLMs won’t lead to AGI - $1,000,000 Prize to find true solution

General intelligence is the ability to approach any problem, any skill, and very quickly master it using very little data.

Dwarkesh Podcast
Francois Chollet, Mike Knoop - LLMs won’t lead to AGI - $1,000,000 Prize to find true solution

Because this is what makes you able to face anything you might ever encounter.