Gary Marcus

Rank 19 of 47

Score 64

@thieme @engmlubbad Sur, that (“no better model”) wasn’t my prediction; that’s a strawman.

these were, and all 7 were correct. we did not see GPT-5 today, much less get it to actually use.

we saw GPT 4.15, not GPT-5. and even that wasn’t actually released.

12/20/2024, 9:19:53 PM

In reply to:

thieme

@thieme

238d

@GaryMarcus @engmlubbad Earn it.

Prove that you have more knowledge about training Transformes than Hinton or I.

Be a man and admit that you lost your "no better model than GPT-4 in 2024" prediction/bet.

Then request people to spend more time with your texts. That’s the order.

Gary Marcus

@GaryMarcus

241d

@engmlubbad Read my substack and my books.

Mohammed Lubbad 🇵🇸

@engmlubbad

241d

@GaryMarcus The evolution of Generative AI certainly raises intriguing questions about its limits and future potential. What specific aspects lead you to this conclusion?

Gary Marcus

@GaryMarcus

241d

Important food for thought, if you care about Generative AI:

Six months ago I predicted that LLMs were beginning to reach a point of diminishing returns. 𝗔𝗹𝗹 𝘀𝗶𝗴𝗻𝘀 𝗻𝗼𝘄 𝘀𝘂𝗽𝗽𝗼𝗿𝘁 𝘁𝗵𝗮𝘁 𝗰𝗼𝗻𝗷𝗲𝗰𝘁𝘂𝗿𝗲.

But it is not the only prediction I made: 𝗜…

The statement is part of a public discourse on the development and future of Generative AI models. It addresses predictions about AI advancements and clarifies a misunderstanding about a previous prediction. The tone is corrective and aims to clarify the speaker's stance on AI development.

Principle 1:
I will strive to do no harm with my words and actions.
The statement strives to do no harm by clarifying a misunderstanding and correcting a misrepresentation of the speaker's predictions. [+1]
Principle 4:
I will engage in constructive criticism and dialogue with those in disagreement and will not engage in personal attacks or ad hominem arguments.
The statement engages in constructive criticism by addressing a strawman argument and providing factual corrections without resorting to personal attacks. [+1]