It feels like there’s an “uncanny valley” style gap with this. People who aren’t tech savvy talk to AI and think they’re like people. Then the more tech savvy people get, the more they understand that it’s just code generating responses and it feels stupider and stupider to talk to it like a person. Then there’s the other side of the dip where tech bros who have fully drunk the coolaid know it’s just code but they still talk to it like that, believing it’s sentient.
Yep. I use LLMs and have always had a mental approach of testing a model to find what it can be reliably useful for.
It is at a point now where the hype has gone for so long, explaining the basics is much harder, as “the magic” has gone for so long day-to-day that it’s instilled. Like being faced to understand the fundamentals forces someone to have to question the last year or two of their understanding and behaviours. And in there comes an increased chance of putting heads and sand. People may have to confront themselves and perspectives up to the point, things hard to let go the longer they’re held onto.
I saw Cory Doctorow recently comparing it in some sense to phonograph records. Before the masses were playing records, for most people, if there was audible music there was always a live performer. It had just always been that way. It broke a lot of brains back then.
I don’t really understand that analogy. Records are pretty faithful in their reproduction of the artists music, while the “AI companion” is as far away from a companion as can be.
It feels like there’s an “uncanny valley” style gap with this. People who aren’t tech savvy talk to AI and think they’re like people. Then the more tech savvy people get, the more they understand that it’s just code generating responses and it feels stupider and stupider to talk to it like a person. Then there’s the other side of the dip where tech bros who have fully drunk the coolaid know it’s just code but they still talk to it like that, believing it’s sentient.
Yep. I use LLMs and have always had a mental approach of testing a model to find what it can be reliably useful for.
It is at a point now where the hype has gone for so long, explaining the basics is much harder, as “the magic” has gone for so long day-to-day that it’s instilled. Like being faced to understand the fundamentals forces someone to have to question the last year or two of their understanding and behaviours. And in there comes an increased chance of putting heads and sand. People may have to confront themselves and perspectives up to the point, things hard to let go the longer they’re held onto.
i mean we’ve had lots of practice talking to machines
I saw Cory Doctorow recently comparing it in some sense to phonograph records. Before the masses were playing records, for most people, if there was audible music there was always a live performer. It had just always been that way. It broke a lot of brains back then.
I don’t really understand that analogy. Records are pretty faithful in their reproduction of the artists music, while the “AI companion” is as far away from a companion as can be.