A few years ago the idea of something even like GPT 4 was decades or 100 years away. Then it dropped and everyone got used to it. And we all became very aware of its limitations especially hallucinations.
Over the next two years there was steady progress and improvement with specialized modes like 3o. These could seem downright magical at certain tasks.
GPT feels like another big leap to me. I have now become an almost constant user and it guides me in basically every moderate to big decision in work and personal life. And astonishingly it seems to not hallucinate (at least not nearly as much). In fact I haven’t caught it in a hallucination yet. Occasionally it will tel me something that I think is a hallucination and then when I dig into it I realize I’m the one who was mistaken.
If it had built in multi modality it would feel straight up agi me. For now when I need visual interaction I take pictures and send it to gpt5, and this works flawlessly across a wide range of tasks.
My job is now 60-70% interacting with AI and it would be pretty easy for a few engineers to cobble together a workflow using AI tools that completely replaced my job (which required a fancy post grad degree from an elite school). GPT-5 is basically my personal assistant now.
Ai is one of the most underhyped technologies I’ve ever experienced (and I lived through the rise of the internet and smart phones). It’s whatever the opposite of a bubble is.
[link] [comments]