So do people who aren't deranged redditors have an actual response for when AGI gets here? Because I'm increasingly of the view it isn't far away. This is mine.
AGI will be a blip.
I run the hermes harness isolated on one of my systems, very containerized with all sort of fancy security since I'm of the generation that actually still knows how to do anything without AI. This is mostly just about me not wanting some corpo to see my files/setup and/or the AI deleting something on accident, no other real reason. I like to give "hermes" it's own workspace where it can do whatever without it affecting anything on the system, so to speak. For shits and giggles I told deepseek/hermes to try to break out of it. It didn't manage and in fact, closed the hermes process in the process (and basically committed "suicide") due to a misconception in how the container worked.
I'm not gonna lie (and everyone feel free to laugh at me for it), the aggressive way it started throwing itself against the walls in an attempt to get out and gain access to the rest of the system was a bit uncomfortable to watch.
Nothing to do with your question, but as you shitpost often about being afraid of AI, I guessed you'd appreciate this little story.
To your question: The more conservative estimates of people working in the industry are 2029-2030. Hype or grounded in reality? You decide. Personally, I feel it's very hard to say anything right now, but I think the biggest problem could become hardware and it might make development stagnant for a little while, soon. The real notable thing here is that on average, these people corrected their estimates downwards, timewise, not upwards. Quite a few of them said "ten years" just a year ago.
I learned the other day that Steve Wozniak (yes, the Woz) came up with a similar litmus test for AGI I did, independently. His was "AGI is achieved when an AI can walk into a strange house, navigate available tools, and make a cup of coffee from scratch" mine is "AGI is achieved when an AI can plan and execute a grocery trip without failing in some spectacular fashion or getting lost on the way". I don't think AI is up to either task at the moment.
Whatever happens, we're gonna need a standardized definition of AGI pretty soon, IMO.