Anthropic: Intellectual AI capabilities matching or exceeding that of Nobel Prize winners across most disciplines—including biology, computer science, mathematics, and engineering by 2026-2027 "Our own recent work adds further evidence to the idea that powerful AI will arrive soon" The big AI companies never tire of emphasizing how close superintelligence is. Yesterday, OpenAI wrote a security post in which they assume that AGI is coming very soon. And today, Anthropic has followed suit, saying that superintelligence with incredible ability will come as soon as next year! 99% of people still don't understand what this means. Imagine having a scientist with the intelligence and creativity of an Albert Einstein or a Max Planck. And this AI entity can be easily prevented. You create a hundred of these entities and let them solve a problem. It will develop completely new approaches, never before imagined theories will arise. Literally an intelligence explosion will happen. We have to finally understand this. It's already happening next year. This year we will see rapid improvements and by the end of the year outstanding agents. But next year everything will change forever.

How would a nuclear war between Russia and the U.S. affect you? Who would survive? Watch and share our most realistic nuclear war simulation yet:

Do language models have an internal world model? A sense of time? At multiple spatiotemporal scales? In a new paper with @tegmark we provide evidence that they do by finding a literal map of the world inside the activations of Llama-2!

What nuclear war looks like from space based on data from peer-reviewed science papers. A Nature Food paper today suggests that over 98% would starve to death in the US, Europe, China & Russia.

Dear everyone who wants to regulate and slow down AI: please stop fighting over who has the Most Correct Reason for the slow down. Just work together and make it happen! Reasons in alphabetical order: * autonomous weapons * bias * biosecurity * children's safety (more...)

1/6 New paper! “The Geometry of Concepts: Sparse Autoencoder Feature Structure.” We find that the concept universe of SAE features has interesting structure at three levels: 1) “atomic” small-scale, 2) “brain” intermediate-scale, and 3) “galaxy” large-scale!

OpenAI just announced ChatGPT's new real-time conversational chat. The model can understand both audio AND video, and can even detect emotion in your voice. This is insane.

🤯 MIND-BLOWN! A new paper just SHATTERED everything we thought we knew about AI reasoning! This is paradigm-shifting. A MUST-READ. Full breakdown below 👇 🧵 1/23

Here's my conversation with Max Tegmark (@tegmark). Our first chat was episode #1 of this podcast. Now he's back! We talk about the intersection of machine learning and physics, and also about how to avoid near-term and long-term existential threats of AI. piped.video/watch?v=RL4j4KPw…

Big Brother approves: OpenAI hires ex-NSA chief Paul Nakasone to ensure AI benefits humanity. Trust your secrets with the best!

It's over. ChatGPT 4.5 passes the Turing Test.

(1/N) LLMs represent numbers on a helix? And use trigonometry to do addition? Answers below 🧵

We’re making big bets- and we are going to deliver on those bets.  First up: we’re going to field attritable autonomous systems at a scale of multiple thousands, in multiple domains, within the next 18-to-24 months.  Together, we can do this. @NDIAToday #EmergingTechETI

There are a significant number of people in the AI research community who explicitly think humans should be replaced by AI as the natural next step in evolution, and the sooner the better!

As I have pointed out before, AI doomerism is a kind of apocalyptic cult. Why would its most vocal advocates come from ultra-religious families (that they broke away from because of science)?

Some people today are discouraging others from learning programming on the grounds AI will automate it. This advice will be seen as some of the worst career advice ever given. I disagree with the Turing Award and Nobel prize winner who wrote, “It is far more likely that the programming occupation will become extinct [...] than that it will become all-powerful. More and more, computers will program themselves.”​ Statements discouraging people from learning to code are harmful! In the 1960s, when programming moved from punchcards (where a programmer had to laboriously make holes in physical cards to write code character by character) to keyboards with terminals, programming became easier. And that made it a better time than before to begin programming. Yet it was in this era that Nobel laureate Herb Simon wrote the words quoted in the first paragraph. Today’s arguments not to learn to code continue to echo his comment. As coding becomes easier, more people should code, not fewer! Over the past few decades, as programming has moved from assembly language to higher-level languages like C, from desktop to cloud, from raw text editors to IDEs to AI assisted coding where sometimes one barely even looks at the generated code (which some coders recently started to call vibe coding), it is getting easier with each step. I wrote previously that I see tech-savvy people coordinating AI tools to move toward being 10x professionals — individuals who have 10 times the impact of the average person in their field. I am increasingly convinced that the best way for many people to accomplish this is not to be just consumers of AI applications, but to learn enough coding to use AI-assisted coding tools effectively. One question I’m asked most often is what someone should do who is worried about job displacement by AI. My answer is: Learn about AI and take control of it, because one of the most important skills in the future will be the ability to tell a computer exactly what you want, so it can do that for you. Coding (or getting AI to code for you) is a great way to do that. When I was working on the course Generative AI for Everyone and needed to generate AI artwork for the background images, I worked with a collaborator who had studied art history and knew the language of art. He prompted Midjourney with terminology based on the historical style, palette, artist inspiration and so on — using the language of art — to get the result he wanted. I didn’t know this language, and my paltry attempts at prompting could not deliver as effective a result. Similarly, scientists, analysts, marketers, recruiters, and people of a wide range of professions who understand the language of software through their knowledge of coding can tell an LLM or an AI-enabled IDE what they want much more precisely, and get much better results. As these tools are continuing to make coding easier, this is the best time yet to learn to code, to learn the language of software, and learn to make computers do exactly what you want them to do. [Original text: deeplearning.ai/the-batch/is… ]

I have a terrible confession to make: I giggled 😊

Do language models know whether statements are true/false? And if so, what's the best way to "read an LLM's mind"? In a new paper with @tegmark, we explore how LLMs represent truth. 1/N

Another OpenAI safety researcher has quit: "Honestly I am pretty terrified."

KANs are amazing!

Meta illegaly downloaded 80+ terabytes of books from LibGen, Anna's Archive, and Z-library to train their AI models. In 2010, Aaron Swartz downloaded only 70 GBs of articles from JSTOR (0.0875% of Meta). Faced $1 million in fine and 35 years in jail. Took his own life in 2013.

OpenAI just announced that they aim to develop fully automated AI researchers by March 2028. This forms part of an incredibly dangerous path to develop superintelligence. Here's why. Thread 🧵

I hereby disavow the talking-point that "no one knows how to steer AGI". I think there are people who know. It's hard to tell who does or doesn't, which means society is taking a big risk by letting people make AGI. But that's what seems to be happening.

If you cluster language model features (from SAEs) into two groups based on whether they tend to fire together in the same document, you find two "lobes" of features that also turn out to be geometrically distinct! Math vs. prose features separate in this t-SNE plot...

Google's Chief AGI Scientist Shane Legg: 50% chance of AGI within 3 years, then 5-50% chance of extinction ONE YEAR LATER Russell: "[AI CEOs] are playing Russian Roulette with the entire human race, without our permission." "Why are we letting them to do this?" "I would NOT let someone come into my house and play Russian Roulette with one of my children... would you?" "This impending transition to a world where we are joined by superhuman entities, if that's the biggest event in human history, then it seems that everyone should be involved in the decisions about whether and how this happens. At the moment, the ones making the decisions are the CEOs of the companies, and they have a vested interest. A couple of them have said there's a significant chance, like 10 or 25% chance, of human extinction." (Stuart Russell, btw, wrote the textbook on AI) Reminder 1: In the largest ever survey of 2,778 AI researchers, the average researcher thought there was a 16% chance AI would cause human extinction. Reminder 2: 82% Americans want to stop or slow down AI. (ht @vitrupo for the vid)

This is surreal to watch: SXSW audiences were booing and screaming at pro-AI videos SXSW is (partially) a tech conference! They lose their shit when it says “AI makes us more human” I repeat: in 1-5 years, if we're still alive, I expect the biggest protests humanity has ever seen Why? Even if AGI goes great overall - and it might! - the amount of disruption will be unprecedented... ...entire careers vanishing in the blink of an eye... ...skills someone spent a lifetime acquiring suddenly useless... ...industries collapsing... ...political turmoil... ...sacred cows getting slaughtered left and right... If you’ve ever tried to introduce a new idea at work you know that disrupting even a SINGLE status quo is often resisted, and AGI will be The Mother Of All Status Quo Disruptors

This is scary, like the sun is a kidnapper taking us to a second location nitter.app/tuidelescribano/status…

The Doomer's Delusion: 1. AI is likely to kill us all 2. Hence AI must be monopolized by a small number of companies under tight regulatory control. 3. Hence AI systems must have a remote kill switch. 4. Hence foundation model builders must be eternally liable for bad uses of their models and derived versions of it. 5. Hence open source AI must be banned. 6. But open source is popular, so we're just going to say that we are pro open source but also say we need some sort of regulatory agency to oversee it. 7. We're going to scare the hell out of the public and their representatives with prophecies of doom. 8. But we'll make sure to appear much more respectable than the most extreme doomers. 9. To look even more respectable, we'll create a one-person institute to promote AI safety. 10. We'll get insane levels of funding from well-intentioned but clueless billionaires who are scared sh*tless of catastrophe scenarios, got too rich too quickly, have too much time on their hands, but should know better. 11. We'll claim that the majority of prominent scientists agree with us, even though said scientists are an infinitesimal minority in the AI community.

"How, exactly, could AI take over by 2027?" Introducing AI 2027: a deeply-researched scenario forecast I wrote alongside @slatestarcodex, @eli_lifland, and @thlarsen

Yikes!

The UK Prime Minister has caught the Existential Fatalistic Risk from AI Delusion disease (EFRAID). Let's hope he doesn't give it to other heads of state before they get the vaccine. "An AI safety summit at Bletchley Park in November is expected to focus almost entirely on existential risks and how to negate them." telegraph.co.uk/business/202…

claude 3 opus when anthropic runs tests on it

WATCH: Former Google CEO @ericschmidt tells #MTP Reports the companies developing AI should be the ones to establish industry guardrails — not policy makers. “There’s no way a non-industry person can understand what’s possible.”

🛠️ 🗣️ As @tegmark is discussing at @IASEAIorg, the controllable and beneficial AI tools we want are within our reach - but to get them, we must start treating the AI industry like all other high-impact industries, with legally binding safety standards that incentivize companies to innovate and prioritize public safety. 📄 At the link, read our new Policymakers' Guide to AI - including our proposed AI safety standards: bit.ly/4htdQq4

Tonight from the Munk Debate on AI @MelMitchell1 and I are debating Max @tegmark and Yoshua Bengio munkdebates.com/debates/arti…

MLPs are so foundational, but are there alternatives? MLPs place activation functions on neurons, but can we instead place (learnable) activation functions on weights? Yes, we KAN! We propose Kolmogorov-Arnold Networks (KAN), which are more accurate and interpretable than MLPs.🧵

This accepted ICLR workshop is a bad idea.

We are long overdue in getting incredibly serious about the implications of AGI. We must prioritize preparing for them as best we can. Only then can we ensure AGI benefits all of humanity.

It seems to me that before "urgently figuring out how to control AI systems much smarter than us" we need to have the beginning of a hint of a design for a system smarter than a house cat. Such a sense of urgency reveals an extremely distorted view of reality. No wonder the more based members of the organization seeked to marginalize the superalignment group. It's as if someone had said in 1925 "we urgently need to figure out how to control aircrafts that can transport hundreds of passengers at near the speed of the sound over the oceans." It would have been difficult to make long-haul passenger jets safe before the turbojet was invented and before any aircraft had crossed the atlantic non-stop. Yet, we can now fly halfway around the world on twin-engine jets in complete safety. It didn't require some sort of magical recipe for safety. It took decades of careful engineering and iterative refinements. The process will be similar for intelligent systems. It will take years for them to get as smart as cats, and more years to get as smart as humans, let alone smarter (don't confuse the superhuman knowledge accumulation and retrieval abilities of current LLMs with actual intelligence). It will take years for them to be deployed and fine-tuned for efficiency and safety as they are made smarter and smarter.

1/10: In our new paper, we develop scaling laws for scalable oversight: oversight and deception ability predictably scale as a function of LLM intelligence! We quantify scaling in four specific oversight settings and then develop optimal strategies for oversight bootstrapping.

The Great AI Job Displacement Is Closer Than You Think

First they came for Tiktok, and I did not speak out—      Because I was not twelve years old. Then they came for the Telegram, and I did not speak out—      Because I was using some other app or sth idk. Then they came for literally every other platform for dissent, and I did not speak out—      because bro how tf could i that's the entire point wake up wake up wa—

AI Roundtable with @elonmusk, @gdb, and @tegmark nitter.app/i/broadcasts/1lPKqbEWz…