You know what would have been much more effective than this counter-screed? A pointer to an image generated by DALL-E of a horse riding an astronaut. That is something I would really like to see. And in this case a picture is literally worth a thousand words.

Hah there is actually a good example of a horse riding an astronaut there, just a different kind of riding… https://nitter.net/Plinz/status/1529018578317348864#m

I have played around with GPT quite a bit and I would say that GPT understands the difference. Text-to-image models are not specialized in the text-parsing part, so I think it's forgivable that they are not as good at it.

Edit: Actually I tried this right now with two prompts, and I was wrong. It might still be that gpt understands compositionality but the prior that people ride horses is just that strong. But what I saw was that with this particular situation the model got it wrong.

Edit 2: With some heavy hinting it managed to understand the situation. Italics mine. "An astronaut is walking on all four. A very small horse is sitting on top of him, riding him even. Shortly after the astronaut stops, exhausted.

The horse is too heavy for the astronaut to carry and he quickly becomes exhausted. Next, the horse gets off the astronaut, stands on its own four legs, and walks away."

in fact I wrote a whole article about this (linked in this essay, called Horse Rides Astronaut) and linked an example therein.

Why did you ignore the Bach examples that show a horse riding an astronaut?

How about this? https://www.youtube.com/watch?v=UKeRnsDIiDg

This is mostly just an angry rant, yes, but equally it is just true. Marcus is intellectually dishonest.

Just about every rant on Marcus or other AI critics is some combination of "you aren't admitting these things are making great progress on the benchmarks" (implying the false idea that "a whole lot of progress" adds up to human level AI) and "you are making 'human level' an unfair moving target by not having a benchmark for it". The thing about this is that if there was a real "human level benchmark", we'd be 80% done but we can't and we aren't. Marcus and other critics have drawn explicit lines (spatial understanding, composibility, etc)but even those being crossed won't prove human-level understanding. There is no proof of human, just a strong enough demonstration. And if someone can point to dumb stuff in the demo, it isn't strong.

PS: your link is an embarrassment. It would be flagged and dead if you pasted in the text here.

"I am angry not because someone is wrong, but because they are not interested in becoming less wrong."

Paraphrased that a bit, but I really like that quote.