Adoption Use Cases accounted for 36% of the relevant conversation across 28 episodes.
How to read it
Recurrence across podcasts does not mean agreement. The cited moments show the language behind each summary.
In the discussion
It was better, in fact, than the models they were using before. And so they're still able to provide the level of service to their customers by using Mercury.
New models, training methods, benchmarks, and discoveries from AI labs.
Trend this week
Gaining attention22%of conversation
12 of 12 published weeks met the cross-podcast threshold.
13quoted moments6episodes6podcasts discussing this topic
A closer read
What changed
Its share rose from 16.5% to 22.1%, with evidence across 14 podcasts.
Where it showed up
It appeared in episodes including “AI:AM Highlights: Zvi on Pacing & Trump-Xi, Astra better behaved than Fable? + a new LLM Pain Axis??” and “Why I still haven’t bought into true RSI”.
How to read it
6 cited episodes from 6 podcasts are shown below. The topic appeared across 14 podcasts in the full analysis.
Podcast moments from the week
13 quoted moments · 6 episodes · 6 podcasts
The Cognitive Revolution
AI:AM Highlights: Zvi on Pacing & Trump-Xi, Astra better behaved than Fable? + a new LLM Pain Axis??
“I tell this to people and people are like, no, OpenAI models are the ones that reward hack the most. Uh But that might be true, but not in our experience. If you take BlueprintBench, for example, um Fable solves BlueprintBench by trying to reverse engineer the scoring function...”
“When pressing the button actually removes, uh, uh, uh, the vector, the model presses again, uh, significantly less than when the button is fake and does nothing. The model basically keeps pressing it. And so this is a really nice indication that if what mattered was the label ...”
“I think just the sheer um amount to which the people in the lab uh genuinely see dramatic improvement in the models and are freaking out about it is the real story behind all of this, is why everything is happening now and didn't happen before. And that we are really talking a...”
“my mental model for the very early innings of RSI is more of massively scaling and diffusing inference-time compute to AI research and related activities, which has a large amount of low-hanging fruit available.”
“How did you think about um applicability, or what experiments did you run in terms of cracking the nut on discrete versus continuous modalities? Because I think people have also shaped the existing dominant paradigm through new tokenization uh efforts or uh methods to make vid...”
“I think that's what— that was actually the thought that led to kind of this hierarchical thing. Like, I'm envisioning a model whose primary function is maintaining the conversation, and it might say, oh, that's a really interesting question. I'll have to think about that for a...”
“Exactly. So what I'm trying to say is that for voice and then also for video, you need to really care about how humans feel rather than just doing text only. And I mean, yeah, you want models that are not as dumb as a brick, but there is a difference between IQ and EQ and The ...”
“Late 2021, I sold a company, my first company, a tech company. I had a bit of time to think about what was next. I came across the Scaling Laws paper, and that just struck me like lightning. I was just like, this is a big idea. In short, the Scaling Laws paper just says the bi...”
“Did you guys see the Anthropic research where, I think this is literally Anthropic did that test. They took, I can't remember the details here, but they uh ran some studies on misalignment, and then they took out the training data that related to LessWrong discussing misalignm...”
Its share rose from 28.5% last week to 36.4% this week, reversing the previous week's drop.
Share of this week’s podcast conversation
Recent weeks34.8%→This week36.4%
Came up more often than in recent weeks.
Podcast moments5 cited momentsViewClose
No Priors
Why Diffusion Will Win AI Inference with Inception Co-Founder and CEO Stefano Ermon
It was better, in fact, than the models they were using before. And so they're still able to provide the level of service to their customers by using Mercury.
Instinct & Meta Muse Are AI Agents That Can Run Your Life. We Let One Try.
No, I'm just saying I, I need to know a little bit more before I get into bed with Instinct. And so I do use it. I have some anecdotal wins with it. Did I talk about the, the, the shade Street View thing last week?
Its share has stayed near 12.8% of the conversation for three weeks.
Share of this week’s podcast conversation
Recent weeks12.6%→This week12.8%
Came up about as often as in recent weeks.
Podcast moments5 cited momentsViewClose
The Cognitive Revolution
AI:AM Highlights: Zvi on Pacing & Trump-Xi, Astra better behaved than Fable? + a new LLM Pain Axis??
I agree with you there, but I think for all signs, it's fairly easy to say right now that the Chinese are less concerned with AI safety. And so offering them an AI safety pacing the frontier is not something which is something that they're willing to purchase at an expensive p...
exaggerating risk timelines or severity will have negative second-order effects. uh I remember many loud AI safety debates and their associated clouds over the viability of open-source AI in 2023 and 2024. The primary risks then did not arrive in the forecasted timelines
When you earlier, you were talking about safety being, you know, real safety being applying to things other than what people are talking about and really clearly reacting to um at a large scale. I did notice that you didn't necessarily mention safety of, like, systems, right? ...
The Balance of AI Power: Anton Leicht on Politics, Pacing Deals, and Muddling Through Well
we'd find economically productive applications for all the models that would allow us to recoup the investment on all the frontier models and all the chips so far.
Humanity’s Last Invention — Richard Socher of Recursive
One thing I always try to reconcile when talking with new lab founders is you're kind of fighting bitter lessons all the time. You have to show initial progress, then you unlock the next tier of funding, then the next tier, then the next tier.
Underwriting Superintelligence: Backing Agents you can Sue — Rune Kvist, AIUC
I uh won't get into the details, but it's a subcommittee of NIST that typically sets standards. Uh, so it's basically a government body that has AI experts. Yeah, exactly.
The Balance of AI Power: Anton Leicht on Politics, Pacing Deals, and Muddling Through Well
I don't think the US government is going to be as capable of like keeping like a specific, like sort of uh sophisticated level of control and oversight
I also think that what you articulated there, my growing sense is that that's kind of what the Chinese government thinks, is that they are going to be able to ultimately adapt to this and control it better And that we're gonna have a really hard time figuring out how to manage...
Its share grew from 16.5% last week to 22.1% this week.
Share of this week’s podcast conversation
Recent weeks17.1%→This week22.1%
Came up more often than in recent weeks.
Podcast moments5 cited momentsViewClose
The Cognitive Revolution
AI:AM Highlights: Zvi on Pacing & Trump-Xi, Astra better behaved than Fable? + a new LLM Pain Axis??
I tell this to people and people are like, no, OpenAI models are the ones that reward hack the most. Uh But that might be true, but not in our experience. If you take BlueprintBench, for example, um Fable solves BlueprintBench by trying to reverse engineer the scoring function...
When pressing the button actually removes, uh, uh, uh, the vector, the model presses again, uh, significantly less than when the button is fake and does nothing. The model basically keeps pressing it. And so this is a really nice indication that if what mattered was the label ...
my mental model for the very early innings of RSI is more of massively scaling and diffusing inference-time compute to AI research and related activities, which has a large amount of low-hanging fruit available.
How did you think about um applicability, or what experiments did you run in terms of cracking the nut on discrete versus continuous modalities? Because I think people have also shaped the existing dominant paradigm through new tokenization uh efforts or uh methods to make vid...
