Remix.run Logo
▲ mbowcut2 a day ago

Not off-putting in the slightest. It's a well-deserved jab at the earliest LLM naysayers whose "criticism" was just ignorance wrapped in causeless activism. They claimed that LLMs were just randomly repeating locally coherent phrases back to us and that no amount of scaling could produce real intelligence. Well, the goal posts have moved far indeed if a proof of GRH isn't intelligence.

▲santadays a day ago | parent | next [-]

I think it's well deserved for all the people who made a name for themselves nay-saying llms/ai. I think it's less well deserved for people worried that their careers are shot and won't be able to provide for their families.

▲uptrails 20 hours ago | parent | prev | next [-]

I've not heard anyone seriously use "stochastic parrots" as a way to deny the utility of LLMs.

Even if you go back to the original "stochastic parrots" paper, it wasn't saying anything of this sort. It was a paper about costs and risks (chiefly: biases from training data, lack of accountability / explainability), not a criticism of AI hype. The concern was about probabilistically parroting bias.

▲dist-epoch a day ago | parent | prev [-]

> the goal posts have moved

Indeed, it's not fashionable anymore to say "stochastic parrot" or "glorified auto-complete". What is fashionable today is to talk about the "convex hull of human knowledge that LLMs interpolate in"