x

75% of people globally want unbiased news (Pew Research) but are concerned they're not getting that.

Get balanced coverage with Symby. See how stories are covered across diverse outlets and spot when one side covers a story the other skips.

PODCAST

Redwood Research Blog

Redwood Research

Narrations of Redwood Research blog posts. Redwood Research is a research nonprofit based in Berkeley. We investigate risks posed by the development of powerful artificial intelligence and techniques for mitigating those risks.

technology philosophy

All Episodes

21:47

“Will reward-seekers respond to distant incentives?”...

Redwood Research Blog ·

2026/02/16

Subtitle: Reward-seekers are supposed to be safer because they respond to incentives under developer control. But what if they also respond to incentives that aren't?. Reward-seekers are usually modeled as responding only to local incentives administered by developers. Here I ask: Will AIs or humans be able to influence their incentives at a distance—e.g., by retroactively reinforcing actions substantially in the future or by committing to run many copies of them in simulated deployments with

en-gb

127:01

“How do we (more) safely defer to AIs?” by Ryan...

Redwood Research Blog ·

2026/02/12

Subtitle: How can we make AIs aligned and well-elicited on extremely hard to check open ended tasks?. As AI systems get more capable, it becomes increasingly uncompetitive and infeasible to avoid deferring to AIs on increasingly many decisions. Further, once systems are sufficiently capable, control becomes infeasible.1 Thus, one of the main strategies for handling AI risk is fully (or almost fully) deferring to AIs on managing these risks. Broadly speaking, when I say “deferring to AIs”2 I mean

en-gb

5:24

“Distinguish between inference scaling and “larger...

Redwood Research Blog ·

2026/02/11

Subtitle: Most recent progress probably isn't from unsustainable inference scaling. As many have observed, since reasoning models first came out, the amount of compute LLMs use to complete tasks has increased greatly. This trend is often called inference scaling and there is an open question of how much of recent AI progress is driven by inference scaling versus by other capability improvements. Whether inference compute is driving most recent AI progress matters because you can only scale up

en-gb

45:21

“Fitness-Seekers: Generalizing the Reward-Seeking...

Redwood Research Blog ·

2026/01/29

Subtitle: If you think reward-seekers are plausible, you should also think “fitness-seekers” are plausible. But their risks aren’t the same.. The AI safety community often emphasizes reward-seeking as a central case of a misaligned AI alongside scheming (e.g., Cotra's sycophant vs schemer, Carlsmith's terminal vs instrumental training-gamer). We are also starting to see signs of reward-seeking-like motivations. But I think insufficient care has gone into delineating this category. If you were to

en-gb

232:24

The inaugural Redwood Research podcast

Redwood Research Blog ·

2026/01/06

After five months of me (Buck) being slow at finishing up the editing on this, we’re finally putting out our inaugural Redwood Research podcast. I think it came out pretty well—we discussed a bunch of interesting and underdiscussed topics and I’m glad to have a public record of a bunch of stuff about our history. Tell your friends! Whether we do another one depends on how useful people find this one. You can watch on Youtube here, or as a Substack podcast.---Published:January 4th,

en-gb

232:24

The inaugural Redwood Research podcast

Redwood Research Blog ·

2026/01/05

After five months of me (Buck) being slow at finishing up the editing on this, we’re finally putting out our inaugural Redwood Research podcast. I think it came out pretty well—we discussed a bunch of interesting and underdiscussed topics and I’m glad to have a public record of a bunch of stuff about our history. Tell your friends! Whether we do another one depends on how useful people find this one. You can watch on Youtube here, or as a Substack podcast.---Published:January 4th,

en-gb

31:37

“Recent LLMs can do 2-hop and 3-hop latent (no CoT)...

Redwood Research Blog ·

2026/01/01

Subtitle: Recent AIs are much better at chaining together knowledge in a single forward pass. Prior work has examined 2-hop latent (as in, no Chain-of-Thought) reasoning (e.g., “What element has atomic number (the age at which Tesla died)?”) and found that LLM performance was limited aside from spurious successes (from memorization and shortcuts). I find that recent LLMs can now do 2-hop and 3-hop latent reasoning with moderate accuracy. I construct a new dataset for evaluating n-hop latent

en-gb

12:17

“Measuring no CoT math time horizon (single forward...

Redwood Research Blog ·

2025/12/26

Subtitle: Opus 4.5 has around a 3.5 minute 50%-reliablity time horizon. A key risk factor for scheming (and misalignment more generally) is opaque reasoning ability. One proxy for this is how good AIs are at solving math problems immediately without any chain-of-thought (CoT) (as in, in a single forward pass). I’ve measured this on a dataset of easy math problems and used this to estimate 50% reliability no-CoT time horizon using the same methodology introduced in Measuring AI Ability to Complete

en-gb

25:34

“Recent LLMs can use filler tokens or problem repeats...

Redwood Research Blog ·

2025/12/22

Subtitle: AI can sometimes distribute cognition over many extra tokens. Prior results have shown that LLMs released before 2024 can’t leverage ‘filler tokens’—unrelated tokens prior to the model's final answer—to perform additional computation and improve performance.1 I did an investigation on more recent models (e.g. Opus 4.5) and found that many recent LLMs improve substantially on math problems when given filler tokens. That is, I force the LLM to answer some math question immediately without

en-gb

32:08

“BashArena and Control Setting Design” by Adam...

Redwood Research Blog ·

2025/12/18

We’ve just released BashArena, a new high-stakes control setting we think is a major improvement over the settings we’ve used in the past. In this post we’ll discuss the strengths and weaknesses of BashArena, and what we’ve learned about how to make settings for high stakes control research. BashArena Overview BashArena is a control setting: a dataset designed to support the kind of adversarial game we studied in the original AI control paper and Ctrl-Z. In general, control settings require two

en-gb

36:12

“The behavioral selection model for predicting AI...

Redwood Research Blog ·

2025/12/04

Subtitle: The basic arguments about AI motivations in one causal graph. Highly capable AI systems might end up deciding the future. Understanding what will drive those decisions is therefore one of the most important questions we can ask. Many people have proposed different answers. Some predict that powerful AIs will learn to intrinsically pursue reward. Others respond by saying reward is not the optimization target, and instead reward “chisels” a combination of context-dependent cognitive

en-gb

31:37

“Will AI systems drift into misalignment?” by Josh...

Redwood Research Blog ·

2025/11/15

Subtitle: A reason alignment could be hard. Written with Alek Westover and Anshul Khandelwal This post explores what I’ll call the Alignment Drift Hypothesis: An AI system that is initially aligned will generally drift into misalignment after a sufficient number of successive modifications, even if these modifications select for alignment with fixed and unreliable metrics.The intuition behind the Alignment Drift Hypothesis. The ball bouncing around is an AI system being modified (e.g. during

en-gb

39:23

“What’s up with Anthropic predicting AGI by early...

Redwood Research Blog ·

2025/11/03

Subtitle: I operationalize Anthropic's prediction of "powerful AI" and explain why I'm skeptical. As far as I’m aware, Anthropic is the only AI company with official AGI timelines1: they expect AGI by early 2027. In their recommendations (from March 2025) to the OSTP for the AI action plan they say: As our CEO Dario Amodei writes in ‘Machines of Loving Grace’, we expect powerful AI systems will emerge in late 2026 or early 2027. Powerful AI systems will have the following properties:

en-gb

35:41

“Sonnet 4.5’s eval gaming seriously undermines...

Redwood Research Blog ·

2025/10/30

Subtitle: and this seems caused by training on alignment evals. Sonnet 4.5's eval gaming seriously undermines alignment evals and seems caused by training on alignment evals According to the Sonnet 4.5 system card, Sonnet 4.5 is much more likely than Sonnet 4 to mention in its chain-of-thought that it thinks it is being evaluated; this seems to meaningfully cause it to appear to behave better in alignment evaluations. So, Sonnet 4.5's behavioral improvements in these evaluations may partly be

en-gb

18:30

“Should AI Developers Remove Discussion of AI...

Redwood Research Blog ·

2025/10/23

There is some concern that training AI systems on content predicting AI misalignment will hyperstition AI systems into misalignment. This has been discussed previously by a lot of people: Anna Salamon, Alex Turner, the AI Futures Project, Miles Kodama, Gwern, Cleo Nardo, Richard Ngo, Rational Animations, Mark Keavney and others. In this post, I analyze whether AI developers should filter out discussion of AI misalignment from training data. I discuss several details that I don’t think have been

en-gb

13:05

“Is 90% of code at Anthropic being written by AIs?”...

Redwood Research Blog ·

2025/10/22

Subtitle: I'm skeptical that Dario's prediction of AIs writing 90% of code in 3-6 months has come true. In March 2025, Dario Amodei (CEO of Anthropic) said that he expects AI to be writing 90% of the code in 3 to 6 months and that AI might be writing essentially all of the code in 12 months.1 Did this prediction end up being true? We’ll charitably interpret the prediction as being about Anthropic itself. Recently, various articles have claimed that Dario said that AIs are writing 90% of the code

en-gb

20:15

“Reducing risk from scheming by studying trained-in...

Redwood Research Blog ·

2025/10/16

Subtitle: Can we study scheming by studying AIs trained to act like schemers?. In a previous post, I discussed mitigating risks from scheming by studying examples of actual scheming AIs.1 In this post, I’ll discuss an alternative approach: directly training (or instructing) an AI to behave how we think a naturally scheming AI might behave (at least in some ways). Then, we can study the resulting models. For instance, we could verify that a detection method discovers that the model is scheming or

en-gb

14:37

“Iterated Development and Study of Schemers (IDSS)”...

Redwood Research Blog ·

2025/10/10

Subtitle: A strategy for handling scheming. In a previous post, we discussed prospects for studying scheming using natural examples. In this post, we’ll describe a more detailed proposal for iteratively constructing scheming models, techniques for detecting scheming, and techniques for preventing scheming. We’ll call this strategy Iterated Development and Study of Schemers (IDSS). We’ll be using concepts from that prior post, like the idea of trying to make schemers which are easier to catch. Two

en-gb

12:01

“The Thinking Machines Tinker API is good news for AI...

Redwood Research Blog ·

2025/10/09

Subtitle: It's a promising design for reducing model access inside AI companies.. Last week, Thinking Machines announced Tinker. It's an API for running fine-tuning and inference on open-source LLMs that works in a unique way. I think it has some immediate practical implications for AI safety research: I suspect that it will make RL experiments substantially easier, and increase the number of safety papers that involve RL on big models. But it's also interesting to me for another reason: the

en-gb

12:17

“Plans A, B, C, and D for misalignment risk” by Ryan...

Redwood Research Blog ·

2025/10/08

Subtitle: Different plans for different levels of political will. I sometimes think about plans for how to handle misalignment risk. Different levels of political will for handling misalignment risk result in different plans being the best option. I often divide this into Plans A, B, C, and D (from most to least political will required). See also Buck's quick take about different risk level regimes. In this post, I’ll explain the Plan A/B/C/D abstraction as well as discuss the probabilities and

en-gb

102 results

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