https://astralcodexten.substack.com/p/somewhat-contra-marcus-on-ai-scaling

I.

Previously: I predicted that DALL-E’s many flaws would be fixed quickly in future updates. As evidence, I cited Gary Marcus’ lists of GPT’s flaws, most of which got fixed quickly in future updates.

Marcus responded with a post on his own Substack, arguing . . . well, arguing enough things that I’m nervous quoting one part as the thesis, and you should read the whole post, but if I had to do it, it would be:

Now it is true that GPT-3 is genuinely better than GPT-2, and maybe (but maybe not, see footnote 1) true that InstructGPT is genuinely better than GPT-3. I do think that for any given example, the probability of a correct answer has gone up. [Scott] is quite right about that, at least for GPT-2 to GPT-3.

But I see no reason whatsoever to think that the underlying problem — a lack of cognitive models of the world —have been remedied. The improvements, such as they are, come, primarily because the newer models have larger and larger sets of data about how human beings use word sequences, and bigger word sequences are certainly helpful for pattern matching machines. But they still don’t convey genuine comprehension, and so they are still very easy for Ernie and me (or anyone else who cares to try) to break.