Cal Newport

speaker

22,298 appearances 39 recordings 5 series first heard Oct 2025 last heard 6d ago

Cal Newport’s voice in public audio — every appearance, attributed to the second.

Trend

recordings per month · last 12 months

Recordings per month over the last 12 months — 39 in all, peaking in Jun 2026 with 9.

Appearances

newest first · ▶ plays the moment

Deep Questions with Cal Newport · How Worrisome is GPT-6’s “Stealth Thinking”? | Tech Decoded · 10 Sep 2026 podcast

on this transformer box. podcast-host

Now, this is like super crude what I drew here, but it captures the main idea, which is when you use these recurrent depth techniques, what you're doing is internally in your large language model, as your table of numbers is moving from block to block, there might be a place in here where when one of the transformer blocks outputs its updated table, you loop back and put that table through the same transformer block again and let it take another stab at it, and then maybe you do it again. podcast-host

That would be called loop transformers. podcast-host

You can actually group together transformers to loop. podcast-host

So maybe it goes through a certain number of transformers and then you loop that output immediately back into the same transformers and have it go through again. podcast-host

The other variation of recurrent depth is where you are a little bit more selective about it. podcast-host

So some of the rows or the columns in that table podcast-host

You have a router that selects which of those you loop back through or have them go through extra layers and others you don't. podcast-host

But it all has this notion of instead of just going straight from block to block to block inside of the LLM, this table of values is we're going to let it go through some of these blocks, some of those values multiple times and then let it continue on until we finally get our token output out at the end. podcast-host

Now, the idea here, like, why would you do this? podcast-host

Well, the idea here is podcast-host

is that it can be relatively inefficient, perhaps, to have to take the results of the computation happening in these transformer blocks and transcribe them into English language tokens that you output, and then have those come all the way back in, be re-encoded, and go all the way back to the transformer block again. podcast-host

If it's, for example, this one block is the key block that's doing computation for your answer, podcast-host

they have it just have to encode its temporary work into some values that make their way all the way out of the LLM into a token, back into the bottom, re-encoded, go all the way back up again to that block before it can do future more computation using that as input as well. podcast-host

That's a lot of computation that's happening where if you just kept looping the input to that one block a few times, it could analyze storage results in that table of value, analyze it again, update that storage results. podcast-host

You could actually dynamically get podcast-host

much more computation, useful computation done right there internally before, you know, you actually went out and generated tokens, right? podcast-host

There are some advantages to this, right? podcast-host

So why would OpenAI do this? podcast-host

There's some advantages to this. podcast-host

Showing 201–220 of 22,298 · page 11 of 1115