This newer model is being continually trained on oodles of data from a huge base of customers; they have it do all sorts of tasks and it tries and sometimes fails and sometimes succeeds and is trained to succeed more often.
My sense is that this isn’t a big part of the story for how new models’ capabilities are being increased. Though I don’t think we know for sure either way.
Now many millions of people are basically treating it like a coworker and virtual assistant. People are giving it their passwords and such and letting it handle life admin tasks for them, help with shopping, etc. and of course quite a lot of code is being written by it.
This seems accurate for coders. Is it true for people who aren’t coders? It’s not really true for my job or life admin tasks (like I use the models a fair bit, but it’s more in chat-bot mode than in agent-mode / trusting the models to do a lot of stuff for me) but maybe it’s more true for others.
unfortunately i think the scenarios are vague enough that as a practical matter it will be tricky to adjudicate or decide whether they’ve happened or not
has come true (unlike many of my actual predictions about the rate of progress, which were consistently too bearish about the median world due to how I was hedging against the current paradigm of scaling running into a bottleneck).
Looking back at the dialogue, I can’t actually remember how I interpreted Daniel’s stage 4 when I offered my probability estimate of 30%. I don’t currently think there’s a 30% chance of “superintelligent AGI” over the next few years, which makes me think what I had in mind for stage 4 was something more impressive than what actually ended up happening.
This also matches my rough sense that the current world is more like my 75th − 80th percentile world from late 2023 and not the 90th+ percentile that would be needed to justify an update from 6% to 30%, which is the update I said I would make if I observed (4) happen.
tbf I myself think that it’s not clear we’re in stage 4, I think stage 3 is arguably where we’re at. But we seem closer to 4 than to 3 imo. Maybe we should interpolate between them?
My sense is that this isn’t a big part of the story for how new models’ capabilities are being increased. Though I don’t think we know for sure either way.
This seems accurate for coders. Is it true for people who aren’t coders? It’s not really true for my job or life admin tasks (like I use the models a fair bit, but it’s more in chat-bot mode than in agent-mode / trusting the models to do a lot of stuff for me) but maybe it’s more true for others.
I think my prediction about
has come true (unlike many of my actual predictions about the rate of progress, which were consistently too bearish about the median world due to how I was hedging against the current paradigm of scaling running into a bottleneck).
Looking back at the dialogue, I can’t actually remember how I interpreted Daniel’s stage 4 when I offered my probability estimate of 30%. I don’t currently think there’s a 30% chance of “superintelligent AGI” over the next few years, which makes me think what I had in mind for stage 4 was something more impressive than what actually ended up happening.
This also matches my rough sense that the current world is more like my 75th − 80th percentile world from late 2023 and not the 90th+ percentile that would be needed to justify an update from 6% to 30%, which is the update I said I would make if I observed (4) happen.
tbf I myself think that it’s not clear we’re in stage 4, I think stage 3 is arguably where we’re at. But we seem closer to 4 than to 3 imo. Maybe we should interpolate between them?