ChatGPT’s capabilities have shifted dramatically over time, according to a new study by researchers from Stanford University and the University of California, Berkeley. The findings provide more evidence for recent user complaints about the chatbot’s declining performance in some areas.
All the doomsayers were like “look how much ChatGPT has improved in the past 6 months, imagine how much more it’s going to improve e in the next 6!”
As I wrote just a few months ago:
I deeply believe that’s a mistake and they’ll realize it soon, if they haven’t already. Sure, maybe they can technically do it, but as the kids say, what they produce is going to be very mid at best, and embarrassing or illegal at worst.
I think LLMs have real world uses that we will enjoy for many years to come, but I also think they’re much dumber and less impactful than some of the folks out there spreading fear about them.