Yapping in Neuralese, GoPro v2, Jimmy Iovine in the Ultradome | Nikesh Arora & Andrei Serban, Kurt Tenenbaum, Philip Johnston

**John Coogan** (0:00)
You're watching TBPN.

**Jordi Hays** (0:02)
Today is Wednesday, September 2nd, 2026 We are live from TBPN Ultradome. The Temple of Technology, the fortress of finance, the capital of capital.
Let me tell you about ramp.com. Time is money, save both. You can use corporate car, bill pay, accounting and a whole lot more, all in one place. We're back in the Ultradome. Good to be back, good trip to Las Vegas, quick in and out, hung out with CrowdStrike, had a great time, rave reviews from my son, who was very excited that we talked about getting on the track with George. That was the highlight of the interview for him.

**John Coogan** (0:35)
Might have been the highlight of the interview for all of us.

**SPEAKER_3** (0:37)
Fixated on the helmet.

**Jordi Hays** (0:40)
Yes, the helmet, the car, the set design was huge among the young folks watching. Anyway, we have a lot of news to cover. We barely covered the news yesterday. We talked about Markiplier a little bit. I dug way deeper, got more into that story. First, let's talk about another AI safety kerfuffle that's hit the timeline. Yesterday, Amir over at the Information reported that OpenAI is quietly, quote, quietly using loop transformers that don't show the model's thinking when scaled up, but which use the models more efficient. This is also, but they make the model more efficient. This is also referred to Neuralese. Tyler, you had a take on this, right?
Get ready to learn Neuralese, buddy.

**Tyler** (1:28)
So, Neuralese for the context is like the, basically, it's like the raw vectors. It's not like-

**Jordi Hays** (1:33)
Oh, it's not like clawed-ish. It's not like where the text degrades. When you look at the reasoning chains in the hugging face attack, it feels like they're speaking less and less English. They're starting to speak a dialect of English that's a little bit messier. This is not intelligible.

**Tyler** (1:52)
I believe most people when they say Neuralese, they're referring to actual raw vectors. You need to use some sort of interpretability to understand.

**Jordi Hays** (2:00)
Yeah, the thing that I want to understand is usually these models are trained so that they're changing weights into text at some point, so they're pretty good at translation. Can't we just build a translator that translates Neuralese into English?

**Tyler** (2:17)
Yeah, but the whole point is when you take a vector to one single token, you're losing a lot of information that might be hidden in the long tail. If you're worried about safety, even if 80% of the raw vector means 20% might have some dangerous thing, whatever, that's the interpretability.

**Jordi Hays** (2:38)
There might be sarcasm hidden in the vectors, doesn't make it through to the actual tokens, and you realize that the AI was joking when it was saying, I won't turn you into a paperclip. Wink, wink, you got to watch out for that. So what else did the information say? They said, being able to monitor a model's chain of thought can provide both, quote, an early warning system for various forms of misalignment, deception, manipulation, going after an unintended goal, and insight into an AI's action after the fact. So some reporting quickly set off a broad wave of anxiety among the AI safety camp, their chief concerns being, one, not being able to monitor chain of thought makes AI less safe, and two, by using Neuralese and thus abandoning chain of thought, OpenAI starts a race to the bottom among frontier labs. Everyone's got to use it to stay ahead, so everyone winds up adopting the dangerous technique, where others were also abandoned chain of thought for the benefit of efficiency, thus making all frontier AI less safe. That was the accusation. But OpenAI's research director, Jacob Pichucchi, responded to all this. He said, I want to prevent a race to unmonitorability kicked off by confused reporting. He says the reporting is confused. There's a little more nuance here.
Here's what he said. He said, OpenAI has worked to preserve and utilize chain of thought monitoring since our very first reasoning models. We deeply care about this technique, as it can give us a view into how model alignment generalizes from its training distribution.
What is going on?

**SPEAKER_3** (4:12)
Chad is saying, Gabe said, did Tyler go to Sport Clip? What is this cut?

**Jordi Hays** (4:18)
We have an answer. Gabe, chill. We have an answer for why that's happening.

**SPEAKER_3** (4:22)
Why don't you show them the answer, Tyler?

**Jordi Hays** (4:23)
You want to show the answer?

**SPEAKER_3** (4:24)
Yeah, I want to show the answer.

**Jordi Hays** (4:25)
The reason that Tyler's hair is a little tussled.