Re:Learning to Reason with LLMs
by jimhefferon on Thursday September 12, 2024 @04:30PM
(#41525179)
I assume the poster meant how much progress the models have made. Roughly late high school capability to late college-ish. Project forward five years.
Predicting exponential functions is a fool’s errand. The tiniest error in your initial observation compounds real fast and we can’t even tell if we’re still in the exponential phase of the sigmoid.
By the "past year's worth of development" I assume you mean the layoffs? Have you been in the industry (or any industry) long? If so, you would have seen many layoffs and bulk-hiring frenzies over the years... it doesn't mean anything about the industry as a whole and it's certainly a foolish thing to change career asperations over.
Specifically regarding the LLM - anyone actually believing these models will replace developers and software engineers, truly, deeply does not understand software development at even the most basic fundamental levels. Ignore these people - they are the snake oil salesmen of our modern times.