It’s the same trap that execs fall into when thinking they can replace humans with AI
Gen AI doesn’t “think” for itself. All the models do is answer “given the text X in front of me what’s the most probable response to spit out”. It has no concept of memory or anything like that. Even chat convos are a bit of a hack as all that happens is that all the text in the convo up until that point is thrown in as X. It’s why chat window limits exist in LLMs, because there’s a limit of how much text they can throw in for X before the model shits itself.
That doesn’t mean they can’t be useful. They are semi decent at taking human input and translating it to programmatic calls (which up until that point you had to work with clunky NLP libraries to do). They are also okay at summarizing info.
But the chat bot and garbage hype around them has people convinced that these are things they’re not. And every company is starting to learn the hard way there’s hard limits to what these things can do.
It’s the same trap that execs fall into when thinking they can replace humans with AI
Gen AI doesn’t “think” for itself. All the models do is answer “given the text X in front of me what’s the most probable response to spit out”. It has no concept of memory or anything like that. Even chat convos are a bit of a hack as all that happens is that all the text in the convo up until that point is thrown in as X. It’s why chat window limits exist in LLMs, because there’s a limit of how much text they can throw in for X before the model shits itself.
That doesn’t mean they can’t be useful. They are semi decent at taking human input and translating it to programmatic calls (which up until that point you had to work with clunky NLP libraries to do). They are also okay at summarizing info.
But the chat bot and garbage hype around them has people convinced that these are things they’re not. And every company is starting to learn the hard way there’s hard limits to what these things can do.