Elon Musk: 1; Yann LeCun: 1; Humanity: 0

In case you haven’t heard, there‘s a brawl happening over at X, Musk vs LeCun. For days I tried to resist commenting, but so many people (friends, reporters, etc) keep asking me for my opinion I have decided to oblige. The first thing that I will tell you is that each of those not so-gentle men scored some points; neither kept things above the belt.

§

The second is that I probably should have listened to Gemini (courtesy Phil Libin) below, in one of its finer moments, and stayed out of this:

Alas, I am now succumbing to peer pressure :(

§

So, game on, here are some snippets from the brawl. The fight seems to have started when Elon tried to recruit and Yann rudely butted in rather aggressively, and without obvious provocation:

LeCun answered by flexing his number of recently published papers (80).

Musk missed a chance ask to whether any of them actually mattered. (I can’t offhand think of an important recent paper from LeCun that has been independently validated by others.) Merely publishing a paper is not necessarily a sign of genuine progress.

§

Skipping ahead, they got into another argument, about what even counts as science.: LeCun made a rather narrow-minded claim, and Musk called him on it. LeCun then protests too much and digs himself deeper. I basically agree with the first sentence of his monologue, but after that LeCun foolishly conflates the common currency of science (publication) with the thing (discovery) itself. And then he digs himself further to make grand but entirely untenable claims about happiness and impact on the world:

Let’s tally a score:

LeCun’s is right that Musk has been undermining his legacy with his politics, his conspiracy theories, and his hype. I give him a point for that.

But LeCun’s pettiness speaks for itself. Musk is not going to die bitter and forgotten; he’s not a leading scientist, or a scientist at all, but his electric cars, StarLink network, and reusable rockets won’t be forgotten any time soon, and they represent important substantive contributions to the world. So one point penalty, for LeCun’s cross between unsportsmanlike conduct and complete implausiblity. Back to zero. (Also, dude started the fight, and really ought to lose another point for that.)

LeCun regains one, charitably, for a point that he hinted at but didn’t quite spell out: we need more science if we are to get to AGI. Right now we have a lot of engineering and alchemy, but not enough basic principles and formal understanding. More science here might help. LeCun: 1. (I should dock him a bit for being a scientist on Meta’s payroll, often saying things that cannot easily be understood otherwise, but that’s a story for another day.)

On the other hand, Musk is right that “if it’s not published, it’s not science” is a ridiculous claim. Lots of science is done inside companies, as Colin Wright points out, and of course the implicit contrapositive is wrong, too: lots of stuff that is published isn’t very good science. More generally, LeCun has emphasized his own narrow-mindedness in his allegation that the only the thing that matter is science; Musk is right that products matter. And AI is a funny hybrid between science and engineering. AI needs more science, but it needs engineering, too. I am giving Musk one point for taking the broader view and calling out the narrow view. That leaves us with a tie, 1-1.

§

On the other hand, the chatbots of both X.ai and Meta still seem to me to be wildly inadequate, hallucinatory machines that merely hint at what AI could be, but with precious little understanding of the concepts underlying the words they so fluently use.

Meta’s AI has been dissected before; it’s not great. By coincidence I played around with Grok, Musk’s latest AI toy, yesterday, and although it has its moments, it can be pretty dismal, too.

Um, cynoacrylate is Krazy Glue, most definitely not designed to be “non-toxic and safe for direct contact with food items”

Oops. Tell that to Jackson, Mississippi, Juneau Alaska, and two others.

Um. Let’s try something different.

1861? Why? Because Abraham (BKA Abe) Lincoln had a Jewish sounding first name??

And oops, Grok’s basic understanding of time is shaky, too:

Frequently wrong (and sometimes right; asking the same question often yields different results), never in doubt.

Any resemblance to AI leaders, past or present, is purely coincidental.

§

Both Musk and LeCun (via Meta) command massive budgets to explore AI; I am not convinced that either is making real progress. Musk is just building bigger and bigger LLMs, and Meta is mostly (despite LeCun’s reservations) mostly doing the same.

Musk and LeCun may have the soapboxes and the money, but what we really need is a breakthrough. I hope that ego won’t preclude that.

Gary Marcus is taking a break from social media but still plans to write here occasionally.