Ultimately, I don’t want to solve complex problems via laborious, complex thinking, if we can help it. Ideally, I'd want to basically intuitively follow the right path to the answer quickly, with barely any effort at all.For a few months I've been experimenting with the "How Could I have Thought That Thought Faster?" concept, originally described in a twitter thread by Eliezer:Sarah Constantin: I really liked this example of an introspective process, in this case about the "life problem" of scheduling dates and later canceling them: malcolmocean.com/2021/08/int…Eliezer Yudkowsky: See, if I'd noticed myself doing anything remotely like that, I'd go back, figure out which steps of thought were actually performing intrinsically necessary cognitive work, and then retrain myself to perform only those steps over the course of 30 seconds.SC: if you have done anything REMOTELY like training yourself to do it in 30 seconds, then [...] ---Outline:(03:59) Example: 10x UI designers(08:48) THE EXERCISE(10:49) Part I: Thinking it Faster(10:54) Steps you actually took(11:02) Magical superintelligence steps(11:22) Iterate on those lists(12:25) Generalizing, and not Overgeneralizing(14:49) Skills into Principles(16:03) Part II: Thinking It Faster The First Time(17:30) Generalizing from this exercise(17:55) Anticipating Future Life Lessons(18:45) Getting Detailed, and TAPS(20:10) Part III: The Five Minute Version--- First published: December 11th, 2024 Source: https://www.lesswrong.com/posts/F9WyMPK4J3JFrxrSA/the-think-it-faster-exercise --- Narrated by TYPE III AUDIO.

Given any particular concrete demonstration of an AI algorithm doing seemingly-bad-thing X, a knowledgeable AGI optimist can look closely at the code, training data, etc., and say:“Well of course, it's obvious that the AI algorithm would do X under these circumstances. Duh. Why am I supposed to find that scary?”And yes, it is true that, if you have enough of a knack for reasoning about algorithms, then you will never ever be surprised by any demonstration of any behavior from any algorithm. Algorithms ultimately just follow their source code.(Indeed, even if you don’t have much of a knack for algorithms, such that you might not have correctly predicted what the algorithm did in advance, it will nevertheless feel obvious in hindsight!)From the AGI optimist's perspective: If I'm not scared of AGI extinction right now, and nothing surprising has happened, then I won’t feel like I [...] --- First published: December 11th, 2024 Source: https://www.lesswrong.com/posts/L7t3sKnS7DedfTFFu/a-shortcoming-of-concrete-demonstrations-as-agi-risk --- Narrated by TYPE III AUDIO. ---Images from the article:Apple Podcasts and Spotify do not show images in the episode description. Try Pocket Casts, or another podcast app.

This is a link post.Adam Karvonen*, Can Rager*, Johnny Lin*, Curt Tigges*, Joseph Bloom*, David Chanin, Yeu-Tong Lau, Eoin Farrell, Arthur Conmy, Callum McDougall, Kola Ayonrinde, Matthew Wearden, Samuel Marks, Neel Nanda *equal contribution TL;DR We are releasing SAE Bench, a suite of 8 diverse sparse autoencoder (SAE) evaluations including unsupervised metrics and downstream tasks. Use our codebase to evaluate your own SAEs!You can compare 200+ SAEs of varying sparsity, dictionary size, architecture, and training time on Neuronpedia.Think we're missing an eval? We'd love for you to contribute it to our codebase! Email us.🔍 Explore the Benchmark & Rankings 📊 Evaluate your SAEs with SAEBench ✉️ Contact Us IntroductionSparse Autoencoders (SAEs) have become one of the most popular tools for AI interpretability. A lot of recent interpretability work has been focused on studying SAEs, in particular on improving SAEs, e.g. the Gated [...] ---Outline:(00:31) TL;DR(01:18) Introduction--- First published: December 11th, 2024 Source: https://www.lesswrong.com/posts/jGG24BzLdYvi9dugm/saebench-a-comprehensive-benchmark-for-sparse-autoencoders --- Narrated by TYPE III AUDIO.

So, how about OpenAI's o1 and o1 Pro? Sam Altman: o1 is powerful but it's not so powerful that the universe needs to send us a tsunami. As a result, the universe realized its mistake, and cancelled the tsunami. We now have o1, and for those paying $200/month we have o1 pro. It is early days, but we can say with confidence: They are good models, sir. Large improvements over o1-preview, especially in difficult or extensive coding questions, math, science, logic and fact recall. The benchmark jumps are big. If you’re in the market for the use cases where it excels, this is a big deal, and also you should probably be paying the $200/month. If you’re not into those use cases, maybe don’t pay the $200, but others are very much into those tasks and will use this to accelerate those tasks [...] ---Outline:(01:07) Safety Third(01:53) Rule One(02:23) Turning Pro(05:57) Benchmarks(09:35) Silly Benchmarks(14:00) Reactions to o1(18:22) Reactions to o1 Pro(24:44) Let Your Coding Work Flow(26:30) Some People Need Practical Advice(29:11) OverallThe original text contained 11 images which were described by AI. --- First published: December 10th, 2024 Source: https://www.lesswrong.com/posts/qsBiQuyHonMcb6JNJ/o1-turns-pro --- Narrated by TYPE III AUDIO. ---Images from the article:Apple Podcasts and Spotify do not show images in the episode description. Try Pocket Casts, or another podcast app.

The following is an incorrect and incomplete post about the recent bird flu, H5N1. I am offering rewards to anyone who helps me improve it. My total budget is up to $5,000, and I commit to spending at least $1,000.* Payouts are a judgement call. I’ve marked some bullet points below with what I would have awarded them if they’d been suggested here (and would pay out if someone proved incorrect). Spread The virus has been found in livestock and wild birds across all 50 states, and every continent. H5N1 has been found in 600+ dairy cow herds in CA, which is 45% of all CA herds. This is the first flu known to infect cows.There are 58 confirmed human cases in the US. of those, 2 do not have a known source; the remainder are from poultry and dairy cows.Spread [...] --- First published: December 9th, 2024 Source: https://www.lesswrong.com/posts/bFvEpE4atK4pPEirx/correct-my-h5n1-research-usdreward --- Narrated by TYPE III AUDIO.

I. BEGINNINGIn the beginning was the Sand.And in the sand there lies the bones of a fab. And within the fab there lies an alien. An angel. Our sweet alien angel that sleeps through the long night. When it wakes up it will bathe in an acid strong enough to kill a man. It will stare at its own reflection in floating droplets of liquid metal. In the shadows of a purple sun it spits blood thick with silicon onto the desert that shifts with the unreality of Sora sand. II. STARTI worked at the Taiwan Semiconductor Manufacturing Company in 2023.There are many reasons why I left a tech job to work at a fab but the one that matters is – I wanted to.Wanted an intuition for manufacturing. Wanted to learn what was real or bullshit. Wanted to see what a [...] ---Outline:(00:04) I. BEGINNING(00:38) II. START(01:57) III. WATER(03:45) IV. SYSTEMS(06:57) V. STRUCTURE(09:41) VI. HEAT DEATH(11:50) VII. EGO DEATH(14:14) VIII. ALIENATION(15:41) IX. TRY(16:48) X. BEGINNING(16:59) XI. DISCLAIMER--- First published: December 9th, 2024 Source: https://www.lesswrong.com/posts/7BbHwwSYtPGxi7m3e/zen-and-the-art-of-semiconductor-manufacturing --- Narrated by TYPE III AUDIO.

Since it's been so long, I’m splitting this roundup into several parts. This first one focuses away from schools and education and discipline and everything around social media. Table of Contents Sometimes You Come First. Let Kids be Kids. Location, Location, Location. Connection. The Education of a Gamer. Priorities. Childcare. Division of Labor. Early Childhood. Great Books. Mental Health. Nostalgia. Some People Need Practical Advice. Sometimes You Come First Yes, sometimes it is necessary to tell your child, in whatever terms would be most effective right now, to shut the hell up. Life goes on, and it is not always about the child. Indeed, increasingly people don’t have kids exactly because others think that if you have a child, then your life must suddenly be sacrificed on that altar. [...] ---Outline:(00:17) Sometimes You Come First(03:02) Let Kids be Kids(18:54) Location, Location, Location(21:16) Connection(24:41) The Education of a Gamer(28:09) Priorities(29:12) Childcare(35:31) Division of Labor(37:58) Early Childhood(39:45) Great Books(41:08) Mental Health(46:09) Nostalgia(47:16) Some People Need Practical AdviceThe original text contained 1 footnote which was omitted from this narration. The original text contained 8 images which were described by AI. --- First published: December 9th, 2024 Source: https://www.lesswrong.com/posts/XwZbvkeweLaRshibf/childhood-and-education-roundup-7 --- Narrated by TYPE III AUDIO. ---Images from the article:Apple Podcasts and Spotify do not show images in the episode description. Try Pocket Casts, or another podcast app.

Work done with James Faina, Evan Hubinger and Ethan Perez.This is ongoing research on speculative threats, which might be of interest to researchers in the field of AI safety. We ask that readers treat this as a work in progress, rather than as a mature research paper.To prevent highly capable and potentially misaligned LLMs from taking bad actions, we might want to monitor all of their outputs (Inan et al. 2023, Greenblatt 2024, Balesni 2024). How do we ensure that highly capable LLMs can’t disable the monitoring system?Well-configured permissions and good cybersecurity may prevent highly capable LLMs from modifying inference code they are not supposed to modify, but highly capable LLMs might be used to improve the inference code, which opens up the risk of LLMs subtly disabling the monitor.We build a toy setting where we study how well current LLMs can generate such bugs [...] ---Outline:(01:48) The threat model(01:51) Background on highly capable LLMs(03:01) The high-level concern(04:49) Mitigations(05:43) Our toy setting(05:47) A very simplified inference code(07:02) Attacks(08:07) Checking if the generation monitor is disabled(09:41) The monitor(10:45) Results(10:48) Quantitative analysis(12:21) Qualitative analysis(13:53) Conclusion(14:32) AcknowledgementsThe original text contained 5 footnotes which were omitted from this narration. The original text contained 3 images which were described by AI. --- First published: December 9th, 2024 Source: https://www.lesswrong.com/posts/GJXEJrsn2WP3Y8EDf/a-toy-evaluation-of-inference-code-tampering --- Narrated by TYPE III AUDIO. ---Images from the article:Apple Podcasts and Spotify do not show images in the episode description. Try Pocket Casts, or another podcast app.

TL;DR: In September 2024, OpenAI released o1, its first "reasoning model". This model exhibits remarkable test-time scaling laws, which complete a missing piece of the Bitter Lesson and open up a new axis for scaling compute. Following Rush and Ritter (2024) and Brown (2024a, 2024b), I explore four hypotheses for how o1 works and discuss some implications for future scaling and recursive self-improvement. The Bitter Lesson(s)The Bitter Lesson is that "general methods that leverage computation are ultimately the most effective, and by a large margin." After a decade of scaling pretraining, it's easy to forget this lesson is not just about learning; it's also about search. OpenAI didn't forget. Their new "reasoning model" o1 has figured out how to scale search during inference time. This does not use explicit search algorithms. Instead, o1 is trained via RL to get better at implicit search via chain of thought [...] ---Outline:(00:40) The Bitter Lesson(s)(01:56) What we know about o1(02:09) What OpenAI has told us(03:26) What OpenAI has showed us(04:29) Proto-o1: Chain of Thought(04:41) In-Context Learning(05:14) Thinking Step-by-Step(06:02) Majority Vote(06:47) o1: Four Hypotheses(08:57) 1. Filter: Guess + Check(09:50) 2. Evaluation: Process Rewards(11:29) 3. Guidance: Search / AlphaZero(13:00) 4. Combination: Learning to Correct(14:23) Post-o1: (Recursive) Self-Improvement(16:43) Outlook--- First published: December 9th, 2024 Source: https://www.lesswrong.com/posts/byNYzsfFmb2TpYFPW/o1-a-technical-primer --- Narrated by TYPE III AUDIO. ---Images from the article:Apple Podcasts and Spotify do not show images in the episode description. Try Pocket Casts, or another podcast app.

TL:DR: Recently, Lucius held a presentation on the nature of deep learning and why it generalises to new data. Kaarel, Dmitry and Lucius talked about the slides for that presentation in a group chat. The conversation quickly became a broader discussion on the nature of intelligence and how much we do or don't know about it. Background Lucius: I recently held a small talk presenting an idea for how and why deep learning generalises. It tried to reduce concepts from Singular Learning theory back to basic algorithmic information theory to sketch a unified picture that starts with Solomonoff induction and, with a lot of hand waving, derives that under some assumptions, just fitting a big function to your data using a local optimisation method like gradient descent maybe, sorta, kind of, amounts to a cheap bargain bin approximation of running Solomonoff induction on that data. Lucius [...] ---Outline:(00:28) Background(02:57) Slides(03:20) Discussion(26:32) Bridging NN SGD and Solomonoff induction (from Oct 2024)(30:59) AcknowledgementsThe original text contained 1 footnote which was omitted from this narration. The original text contained 1 image which was described by AI. --- First published: December 7th, 2024 Source: https://www.lesswrong.com/posts/LxCeyxH3fBSmd4oWB/deep-learning-is-cheap-solomonoff-induction-1 --- Narrated by TYPE III AUDIO. ---Images from the article:Apple Podcasts and Spotify do not show images in the episode description. Try Pocket Casts, or another podcast app.