Deedy Das
speaker
881 appearances
1 recordings
1 series
first heard Nov 2025
last heard 14 Nov
Deedy Das’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 — 1 in all, peaking in Nov 2025 with 1.
Appearances
It's not really top-down, but the way I frame it is if I fast forward 10 years from the future, what do I think is very likely to exist?
And what are the ways I can get there?
If I do believe strongly that there's something like that, and I believe there's a team very strongly headed towards that direction, I can sort of draw a dotted line and go like, okay, maybe we can see something here.
So that's how I broadly think about it.
The way I think about the company is right now, almost all frontier and some many non-frontier AI models are complete black boxes.
You don't understand why they produce the outputs they produce.
All of the eval and studies on them are empirical studies, not intrinsic to the model.
So it's like, hey, here's the outputs we saw, and therefore this is the benchmark score, or this is how we think it did.
If we believe as a society that
five and 10 years later in the future, these models are going to be critically important for making pretty heavy decisions, whether it's, I call it, anything from whether somebody should get a loan or insurance or a legal decision, then
I don't think that the black box approach is long-term scalable.
It's just not how society can function, where you throw your hands up and say, well, this is what the model said, and then I asked it, explain yourself, and it said this other stuff.
Great.
That's kind of what we have today.
That's the best thing that we have.
Mechanistic interpretability is really going into the weights of the model and trying to figure out why did the model do what it did.
And one of the more concrete and relatable examples of this that you guys may be aware of is GPT-4.0 had this phase of sycophancy that a lot of users really liked, but it's kind of one of those things that's not as easily detectable in an eval.
Unless you know you're specifically maybe testing for it.
Even then, it's quite hard.
It's very personalized.
Showing 521–540 of 881 · page 27 of 45
← Previous
Next →