Reading NLP papers from the age of Symbolic AI is the most interesting way to spend a weekend. It looks like everyone in the 60s and 70s quotes Chomsky, if only to disagree with something :D
I feel like the optimism of that age still exceeds what we have now, even though many systems were rather like Potemkin’s villages, however interesting from the engineering standpoint… See the Georgetown-IBM experiment, SHRDLU. Nowadays, everyone is running benchmarks.
Apparently, Joseph Weizenbaum was quite puzzled that many treated ELIZA as something “intelligent”. And it’s much, much less impressive than #chatgpt .