A bunch of people fine-tuning an off-the-shelf model on a proprietary task only to fail horrendously will never lead to any progress, let alone AGI.
So, nobody is trying AGI.
If all those people would actually collectively work on a large-scale research project, we’d see humanity advance. But that’s exactly my point.
“Nobody is trying AGI” is simply just not true. If you think what they’re doing will never lead to AGI, then that’s an opinion you’re free to have, but it’s still just that; an opinion. Our current LLM’s are by far the closest resemblance of AGI that we’ve ever seen. That route may very well be a dead end but it may also not be. You can’t know that.
Oh gosh, look, an AI believer.
No, LLM will not lead to AGI. But even if they did, applying existing tech to a new problem only to fail cuz you’re dumb at estimating the complexity does not, in fact, improve the underlying technology.
To paraphrase in a historical context: no matter how many people run around with shovels digging the ground for something, it will never lead to an invention of the excavator.
