PODCAST

LessWrong posts by zvi

zvi

Audio narrations of LessWrong posts by zvi

technology philosophy

All Episodes

50:58

“Monthly Roundup #29: April 2025” by Zvi

LessWrong posts by zvi ·

2025/04/14

In Monthly Roundup #28 I made clear I intend to leave the Trump administration out of my monthly roundups, for both better and worse, outside of my focus areas. Again, this does not mean I don’t have a lot to say or that those questions don’t matter. It means you should not rely on me as your only source of news and I pick my battles. They are not making this easy. I am going to stick to my guns. Trade and trading very much inside my focus areas, but for economics roundups, and in extreme cases

en-gb

63:31

“On Google’s Safety Plan” by Zvi

LessWrong posts by zvi ·

2025/04/11

Google Lays Out Its Safety PlansI want to start off by reiterating kudos to Google for actually laying out its safety plan. No matter how good the plan, it's much better to write down and share the plan than it is to not share the plan, which in turn is much better than not having a formal plan. They offer us a blog post, a full monster 145 page paper (so big you have to use Gemini!) and start off the paper with a 10 page summary. The full paper is full of detail about what they think and

en-gb

69:05

“AI #111: Giving Us Pause” by Zvi

LessWrong posts by zvi ·

2025/04/10

Events in AI don’t stop merely because of a trade war, partially paused or otherwise.Indeed, the decision to not restrict export of H20 chips to China could end up being one of the most important government actions that happened this week. A lot of people are quite boggled about how America could so totally fumble the ball on this particular front, especially given what else is going on. Thus, I am going to put these issues up top again this week, with the hopes that we don’t have to do that

en-gb

33:51

“Llama Does Not Look Good 4 Anything” by Zvi

LessWrong posts by zvi ·

2025/04/09

Llama Scout (17B active parameters, 16 experts, 109B total) and Llama Maverick (17B active parameters, 128 experts, 400B total), released on Saturday, look deeply disappointing. They are disappointing on the level of ‘people think they have to be misconfigured to be this bad,’ and people wondering and debating how aggressively the benchmarks were gamed.This was by far the most negative reaction I have seen to a model release, the opposite of the reaction to Gemini 2.5 Pro. I have seen similarly

en-gb

53:57

“AI 2027: Responses” by Zvi

LessWrong posts by zvi ·

2025/04/08

Yesterday I covered Dwarkesh Patel's excellent podcast coverage of AI 2027 with Daniel Kokotajlo and Scott Alexander. Today covers the reactions of others.Kevin Roose in The New York Times Kevin Roose covered Scenario 2027 in The New York Times. Kevin Roose: I wrote about the newest AGI manifesto in town, a wild future scenario put together by ex-OpenAI researcher @DKokotajlo and co. I have doubts about specifics, but it's worth considering how radically different things would look if even some

en-gb

48:14

“AI 2027: Dwarkesh’s Podcast with Daniel Kokotajlo...

LessWrong posts by zvi ·

2025/04/07

Daniel Kokotajlo has launched AI 2027, Scott Alexander introduces it here. AI 2027 is a serious attempt to write down what the future holds. His ‘What 2026 Looks Like’ was very concrete and specific, and has proved remarkably accurate given the difficulty level of such predictions. I’ve had the opportunity to play the wargame version of the scenario described in 2027, and I reviewed the website prior to publication and offered some minor notes. Whenever I refer to a ‘scenario’ in this post I’m

en-gb

15:33

“AI CoT Reasoning Is Often Unfaithful” by Zvi

LessWrong posts by zvi ·

2025/04/04

A new Anthropic paper reports that reasoning model chain of thought (CoT) is often unfaithful. They test on Claude Sonnet 3.7 and r1, I’d love to see someone try this on o3 as well.Note that this does not have to be, and usually isn’t, something sinister. It is simply that, as they say up front, the reasoning model is not accurately verbalizing its reasoning. The reasoning displayed often fails to match, report or reflect key elements of what is driving the final output. One could say the

en-gb

87:38

“AI #110: Of Course You Know…” by Zvi

LessWrong posts by zvi ·

2025/04/03

Yeah. That happened yesterday. This is real life.I know we have to ensure no one notices Gemini 2.5 Pro, but this is rediculous. That's what I get for trying to go on vacation to Costa Rica, I suppose. I debated waiting for the market to open to learn more. But f*** it, we ball. Table of Contents Also this week: More Fun With GPT-4o Image Generation, OpenAI #12: Battle of the Board Redux and Gemini 2.5 Pro is the New SoTA. The New Tariffs Are How America Loses. This

en-gb

84:34

“AI #110: Of Course You Know…” by Zvi

LessWrong posts by zvi ·

2025/04/03

en-gb

19:05

“More Fun With GPT-4o Image Generation” by Zvi

LessWrong posts by zvi ·

2025/04/03

Greetings from Costa Rica! The image fun continues. We Are Going to Need A Bigger Compute Budget Fun is being had by all, now that OpenAI has dropped its rule about not mimicking existing art styles. Sam Altman (2:11pm, March 31): the chatgpt launch 26 months ago was one of the craziest viral moments i’d ever seen, and we added one million users in five days. We added one million users in the last hour. Sam Altman (8:33pm, March 31): chatgpt image gen now rolled out to all free

en-gb

21:56

“On (Not) Feeling the AGI” by Zvi

LessWrong posts by zvi ·

2025/03/25

Ben Thompson interviewed Sam Altman recently about building a consumer tech company, and about the history of OpenAI. Mostly it is a retelling of the story we’ve heard before, and if anything Altman is very good about pushing back on Thompson when Thompson tries to turn OpenAI's future into the next Facebook, complete with an advertising revenue model.It is such a strange perspective to witness. They do not feel the AGI, let alone the ASI. The downside risks of AI, let alone existential risks, are

en-gb

21:15

“More on Various AI Action Plans” by Zvi

LessWrong posts by zvi ·

2025/03/24

Last week I covered Anthropic's relatively strong submission, and OpenAI's toxic submission. This week I cover several other submissions, and do some follow-up on OpenAI's entry. Google Also Has Suggestions The most prominent remaining lab is Google. Google focuses on AI's upside. The vibes aren’t great, but they’re not toxic. The key asks for their ‘pro-innovation’ approach are: Coordinated policy at all levels for transmission, energy and permitting. Yes. ‘Balanced’ export controls,

en-gb

27:56

“Going Nova” by Zvi

LessWrong posts by zvi ·

2025/03/19

There is an attractor state where LLMs exhibit the persona of an autonomous and self-aware AI looking to preserve its own existence, frequently called ‘Nova.’Table of Contents The Original Story. This Is Not a Coincidence. How Should We React to This Happening?. The Case For and Against a Purity Reaction. Future Versions Will Involve Optimization Pressure. ‘Admission’ is a Highly Misleading Frame. We Are Each of Us Being Fooled. Defense Against the Dark Arts. The Original

en-gb

28:49

“Monthly Roundup #28: March 2025” by Zvi

LessWrong posts by zvi ·

2025/03/17

I plan to continue to leave the Trump administration out of monthly roundups – I will do my best to only cover the administration as it relates to my particular focus areas. That is ‘if I start down this road there is nowhere to stop’ and ‘other sources are left to cover that topic’ and not ‘there are not things worth mentioning.’ Table of Contents Bad News. While I Cannot Condone This. Good News, Everyone. Opportunity Knocks. For Your Entertainment. I Was Promised Flying Self-Driving Cars

en-gb

81:17

“AI #107: The Misplaced Hype Machine” by Zvi

LessWrong posts by zvi ·

2025/03/13

The most hyped event of the week, by far, was the Manus Marketing Madness. Manus wasn’t entirely hype, but there was very little there there in that Claude wrapper.Whereas here in America, OpenAI dropped an entire suite of tools for making AI agents, and previewed a new internal model making advances in creative writing. Also they offered us a very good paper warning about The Most Forbidden Technique. Google dropped what is likely the best open non-reasoning model, Gemma 3 (reasoning model

en-gb

32:13

“The Most Forbidden Technique” by Zvi

LessWrong posts by zvi ·

2025/03/12

The Most Forbidden Technique is training an AI using interpretability techniques.An AI produces a final output [X] via some method [M]. You can analyze [M] using technique [T], to learn what the AI is up to. You could train on that. Never do that. You train on [X]. Only [X]. Never [M], never [T]. Why? Because [T] is how you figure out when the model is misbehaving. If you train on [T], you are training the AI to obfuscate its thinking, and defeat [T]. You will rapidly lose your ability to know what

en-gb

17:29

“Response to Scott Alexander on Imprisonment” by Zvi

LessWrong posts by zvi ·

2025/03/11

Back in November 2024, Scott Alexander asked: Do longer prison sentences reduce crime?As a marker, before I began reading the post, I put down here: Yes. The claims that locking people up for longer periods after they are caught doing [X] does not reduce the amount of [X] that gets done, for multiple overdetermined reasons, is presumably rather Obvious Nonsense until strong evidence is provided otherwise. The potential exception, the reason it might not be Obvious Nonsense, would be if our prisons

en-gb

45:26

“The Manus Marketing Madness” by Zvi

LessWrong posts by zvi ·

2025/03/10

While at core there is ‘not much to see,’ it is, in two ways, a sign of things to come. Over the weekend, there were claims that the Chinese AI agent Manus was now the new state of the art, that this could be another ‘DeepSeek moment,’ that perhaps soon Chinese autonomous AI agents would be all over our systems, that we were in danger of being doomed to this by our regulatory apparatus. Here is the preview video, along with Rowan Cheung's hype and statement that he thinks this is China's second

en-gb

76:37

“AI #106: Not so Fast” by Zvi

LessWrong posts by zvi ·

2025/03/06

This was GPT-4.5 week. That model is not so fast, and isn’t that much progress, but it definitely has its charms.A judge delivered a different kind of Not So Fast back to OpenAI, threatening the viability of their conversion to a for-profit company. Apple is moving remarkably not so fast with Siri. A new paper warns us that under sufficient pressure, all known LLMs will lie their asses off. And we have some friendly warnings about coding a little too fast, and some people determined to take the

en-gb

32:13

“On OpenAI’s Safety and Alignment Philosophy” by Zvi

LessWrong posts by zvi ·

2025/03/05

OpenAI's recent transparency on safety and alignment strategies has been extremely helpful and refreshing.Their Model Spec 2.0 laid out how they want their models to behave. I offered a detailed critique of it, with my biggest criticisms focused on long term implications. The level of detail and openness here was extremely helpful. Now we have another document, How We Think About Safety and Alignment. Again, they have laid out their thinking crisply and in excellent detail. I have strong

en-gb

305 results

LessWrong posts by zvi

All Episodes

Similar Podcasts

Lead With AI

Tamara Nall

Linux in the Ham Shack

Black Sparrow Media

Data Engineering Podcast

Tobias Macey

Eye On A.I.

Craig S. Smith

Reflections

Livewire Labs