[jarpschop]

A year ago he said that language models would never be able to explain why if you move a table, the bottle on top of it would not fall. He definitely had no idea what was coming with LLMs. And now he is arrogant enough to say "we all knew what capabilities LLMs would have". He has no intelectual humility or honesty at all. I at least admit openly that language models blew my mind.

Lost respect for his opinions massively after seeing how he lied about the topic.

[agnosticmantis]

He’s said it’s easy to generate this kind of questions that trick LLMs because of the lack of physical grounding in models trained solely on text.

And that’s true now as ever. I also heard him say that training multimodal models on text+image/video would mitigate the grounding issue, and that’s proven to be true too.

So I’m not sure exactly what your objection is.

[blueblisters]

Can you cite a reference for that quote? I can’t find it. He has definitely underestimated language model capabilities but he has been mostly right on the “language models can’t plan” stuff. He has been consistent and correct about his criticism about auto regressive model as well.