Here's the algorithm to get you appropriately scared.
- Measure the amount of time it took to go from GPT 2 to GPT 3
- Measure the amount of time it took to go from GPT 3 to GPT 4
- Measure the code quality improvement for the same spans. Go ahead. Pick a metric.
You'll notice that, not only is GPT-4 really fscking good, it represents a year-over-year improvement that is accelerating.
What you are likely looking at is a classic hockey-stick exponential graph. And there is no ceiling yet in sight.
Its a pretty ironclad rule that all growth charts that look exponential are really closer to logistic and seem like they are going to go up more rapidly forever until they suddenly flatten out.
For now, it approximates exponential. That's bad.
We are approaching the time when we find out if the human brain uses language encoding as its core mechanism of thought, or if it has any tricks left up its damp, fleshy sleeve.
It's breathtaking for a philosopher to live through this era. I just wish I wasn't so frightened by the likely practical implications of handing out such a powerful tool-that-can-be-a-weapon (a TTCBAW) to anyone who can pull down a torrent.
1 year.
> - Measure the amount of time it took to go from GPT 3 to GPT 4
3 years.
> it represents a year-over-year improvement that is accelerating.
I don't see how the above shows it is accelerating.
Please evaluate the code quality metric, as I originally asked. (Your pick.)
Go look at GPT2's code.
