Rob Wiblin
speaker
5,091 appearances
15 recordings
1 series
first heard Oct 2021
last heard 6 Aug
Rob Wiblin’s voice in public audio — every appearance, attributed to the second.
Trend
recordings per month · last 12 monthsRecordings per month over the last 12 months — 13 in all, peaking in Apr 2026 with 3.
Appearances
So that's another possible path.
It's possible that just through brute, I guess there's a bunch of areas where the AIs are weak, like structure, long-term planning, things that are hard to do, reinforcement learning from verifiable rewards.
But it's possible that just through like sheer force of effort, we could, or the AI companies could hire lots of people to basically read the outputs, read the work that the AIs do.
to like grade where they've been doing well and where they've been doing poorly and basically say whether they did a good job or not.
That would be expensive, but the companies are spending a lot of money.
I guess I've seen some people run numbers that maybe this would only kind of double the cost of a training run on this kind of thing.
And so that way you could just produce like just enough data to basically get them to be human level at these kinds of strategic planning tasks or long time horizon tasks that so far we've been struggling to crack.
There's probably a couple of others that I'm forgetting, but I guess I'm very worried about short timelines like other people are because there are a number of different like pathways to get there.
Is this like what...
10 years is a lot of time for China to change its posture and to potentially catch up.
I think even people who are relatively bearish on it.
Yeah, I guess they've been allowed to progress with it because nobody believed that it was actually going to happen.
Yeah, exactly.
But if they really did think they were right about to do it, they would object.
Yeah, I guess people who are already heavily involved in AI, I think they're very worried about missing the window, doing something that takes three years to pay off and then just everything is obviated by then.
Because you're pointing out that there are some opportunities that might be incredibly impactful that would work if things take eight years or 12 years to pay off.
That might be much more useful than the small margins that people are thinking about over the next couple of years.
Because-
If it takes that long, there could be really radical cultural change, opinion change, political change that opens up options that seem completely closed now.
It would be, I guess, a bit crazy to just have nobody working on thinking ahead about what those might be or positioning themselves to take advantage of it.
Showing 261–280 of 5,091 · page 14 of 255
← Previous
Next →