You know, OpenAI published a paper in 2020 modelling how far they were from human language error rate and it correctly predicted the accuracy of GPT 4. Deepmind also published a study in 2023 with the same metrics and discovered that even with infinite training and power it would still never break 1.69% error rate.
These companies knew that their basic model was failing and that overfitying trashed their models.
Sam Altman and all these other fuckers knew, they’ve always known, that their LLMs would never function perfectly. They’re convincing all the idiots on earth that they’re selling an AGI prototype while they already know that it’s a dead-end.
As far as I know, the Deepmind paper was actually a challenge of the OpenAI paper, suggesting that
models are undertrained and underperform while using too much compute due to this. They tested
a model with 70B params and were able to outperform much larger models while using less compute by
introducing more training. I don’t think there can be any general conclusion about some hard
ceiling for LLM performance drawn from this.
However, this does not change the fact that there are areas (ones that rely on correctness)
that simply cannot be replaced by this kind of model, and it is a foolish pursuit.
You know, OpenAI published a paper in 2020 modelling how far they were from human language error rate and it correctly predicted the accuracy of GPT 4. Deepmind also published a study in 2023 with the same metrics and discovered that even with infinite training and power it would still never break 1.69% error rate.
These companies knew that their basic model was failing and that overfitying trashed their models.
Sam Altman and all these other fuckers knew, they’ve always known, that their LLMs would never function perfectly. They’re convincing all the idiots on earth that they’re selling an AGI prototype while they already know that it’s a dead-end.
As far as I know, the Deepmind paper was actually a challenge of the OpenAI paper, suggesting that models are undertrained and underperform while using too much compute due to this. They tested a model with 70B params and were able to outperform much larger models while using less compute by introducing more training. I don’t think there can be any general conclusion about some hard ceiling for LLM performance drawn from this.
However, this does not change the fact that there are areas (ones that rely on correctness) that simply cannot be replaced by this kind of model, and it is a foolish pursuit.