My current view is that synthetic data and/or RL is probably sufficient to get us to AGI, but I can respect people who think otherwise. If the gains of synthetic data and/or RL become so overwhelming at any point this year that I start to think data-naysayers are dumb, I will resolve this YES. Otherwise, I will wait until 2026 to resolve this NO. I won't trade on this market.
1,000🏅 Top traders
| # | Trader | Total profit |
|---|---|---|
| 1 | Ṁ5,277 | |
| 2 | Ṁ1,819 | |
| 3 | Ṁ1,349 | |
| 4 | Ṁ438 | |
| 5 | Ṁ410 |
People are also trading
If the companies are currently somewhat bottlenecked on data, so they spend a lot of resources making their training procedure more sample efficient til the point where data is clearly no longer a bottleneck, am I right in thinking this would indicate the data-naysayers are not dumb? Ig there’s the subtlety where the people that remain data naysayers past the point where data stops being a bottleneck, are selected for for being dumb, so maybe they are currently not dumb but the problem is solved and the ones that remain end up being dumb.. uh
I think that there is no bottleneck as there's a lot of non-public data that will find its way into training through partnerships, etc.
The idea of a bottleneck is partly based on a wrong assumption about data resources and their utilization to date.
Thinking of proprietary technical standards, etc.