Also not saying it’s useless completely. I’m a ChatGPT power user for all sorts of stuff. It’s a great tool for becoming acquainted with a new topic and provides all sorts of benefits at the margins.

But for generative AI, I’ve hit a tipping point. Recently I’ve become frustrated at its blind spots. It’s not exactly hallucination, but rather a confident answer that misses a hugely important piece of information. I had always thought they would iron this out but ChatGPT 5 is as bad or worse as a model that came out a year ago, even when I direct it to the “thinking” model.

If the error rate is persistently 10 percent , which is pretty good, it is still not good enough for people to use it confidently for the types of low hanging tasks that should be automated. The promise of AI was to automate relatively simple or repetitive work, but by nature these tasks require a very high degree of accuracy and reliability. They are the infrastructure of “higher order” work that requires human insight and judgment. Yet, the AI is seemingly incapable of hitting the reliability threshold , and worse, individual users cannot teach the AI.

Asking myself - what is the reason for all this spend on Gen AI? Three years ago the tech was nascent and we could all imagine how it would be used. Today it seems we are stuck on having cool, useful, but ultimately unreliable products with error rates too high for actual commercial deployment. At some point this reality is going to set in and people are going to panic sell.

(http://www.autoadmit.com/thread.php?thread_id=5775089&forum_id=2#49270484)

So if you think the internet was a good longterm investment, AI will be even more so.

(http://www.autoadmit.com/thread.php?thread_id=5775089&forum_id=2#49270557)

(http://www.autoadmit.com/thread.php?thread_id=5775089&forum_id=2#49270583)