Users often interact with large language models through black-box inference APIs, both for closed- and open-weight models (e.g., Llama models are popularly accessed via Amazon Bedrock and Azure AI Studio). In order to cut costs or add functionality, API providers may quantize, watermark, or finetune the underlying model, changing the output distribution -- often without notifying users. We formalize detecting such distortions as Model Equality Testing, a two-sample testing problem, where the user collects samples from the API and a reference distribution and conducts a statistical test to see if the two distributions are the same. We find that tests based on the Maximum Mean Discrepancy between distributions are powerful for this task: a test built on a simple string kernel achieves a median of 77.4% power against a range of distortions, using an average of just 10 samples per prompt. We then apply this test to commercial inference APIs for four Llama models, finding that 11 out of 31 endpoints serve different distributions than reference weights released by Meta.

Who is behind the most notorious "deepfake" app on the internet? Trying to answer that question these past few months, for a new Guardian podcast series, Black Box, has been like wandering through a hall of mirrors. The app, ClothOff, has hundreds of thousands of followers and has already been used in a least two cases to generate dozens of images of underage girls – pictures that have left the girls traumatised, their parents outraged and the police baffled at how to stop it. Producers Josh Kelly, Alex Atack and I have followed ClothOff's trail to nondescript addresses in central London that appear to be unoccupied. We have encountered sham businesses, distorted voices and photographs of fake employees.

In May 2020, OpenAI released a report on algorithmic efficiency improvements in deep learning. Main headline: Compared to 2012, it now takes 44 times less compute to train a neural network to the level of AlexNet (by contrast, Moore’s Law would yield an 11x cost improvement over this period). Our results suggest that for AI tasks with high levels of recent investment, algorithmic progress has yielded more gains than classical hardware efficiency. A lot people were surprised by this; there’s a common narrative in which AI progress has come mostly from throwing more and more compute at relatively-dumb algorithms. (This is a common interpretation of The Bitter Lesson, though I would argue it is largely a misinterpretation.) I’ve had various experiences over the years which made the result not-that-surprising. Algorithms beating compute is the sort of thing I expect by default, on a gut level. The point of this post is to tell a few of the stories which underlie that intuition, aimed especially toward people who don’t have much first-hand experience with software engineering, ML, or simulation. (There will still be some jargon, though.) Disclaimer: this does not mean that you should put tons of confidence on this view. The goal is just to provide a possible lens through which “algorithmic progress has yielded more gains than classical hardware efficiency” makes sense; I want to raise that hypothesis from entropy. I’m not going to provide the sort of evidence which would justify very high confidence, I’m just going to point it out as a hypothesis to keep in the back of your mind, and update on when results like OpenAI’s come along. REWRITE IN C Back in college, I spent a summer simulating an unusual type of biochemical oscillator, officially under the aegis of the Minnesota Supercomputing Institute. The algorithm was conceptually simple: every time a reaction occurs between two molecules, update the counts of each molecule, then randomly sample to figure out when th

Let's dive into the realms of Artificial Intelligence. We're thrilled to announce our latest podcast with Lara Heller, British Filmmaker, Actress and Producer of an upcoming movie SYNTH. SYNTH – Winner of Venice Film Awards for Best Sci-Fi is inspired by the true story of an AI application called Sophia, a Humanoid Robot made by Hanson Robotics in Hong Kong. Sophia is the first AI in the world to receive citizenship in Saudi Arabia. She further shares her views on how disruptive technologies are transforming everyday lives and impacting the film industry.

I love the sci-fi movie genre. Futuristic scenarios, jaw-dropping visuals, a tight storyline knitting it all together – that's a recipe for a box office hit. Anyone who grew up in the 80s and 90s will be intimately familiar with the Terminator franchise. And once I moved into the machine learning space, my appreciation and interest in these movies grew multifold! What we once thought of as unrealistic scenarios are now playing out in the real world.

The art of utilizing machine learning (ML) is therefore in perfecting how it augments human intuition and curiosity, and in automating this unity to the maximum extent. The following is a true story from a pilot Cortex XDR Managed Threat Hunting customer, and it showcases the security outcomes that can be achieved today when you pair powerful AI with elite threat hunting expertise.

I'm a physicist who works at a YC startup. Our job is to help new grads get hired into their first machine learning jobs. Some time ago, I wrote about the things you should do to get hired into your first machine learning job. I said in that post that one thing you should do is build a portfolio of your personal machine learning projects. But I left out the part about how to actually to do that, so in this post, I'll tell you how.

I'm a physicist who works at a YC startup. Our job is to help new grads get hired into their first machine learning jobs. Some time ago, I wrote about the things you should do to get hired into your first machine learning job. I said in that post that one thing you should do is build a portfolio of your personal machine learning projects. But I left out the part about how to actually to do that, so in this post, I'll tell you how.

David Byrne performs at the New Orleans Jazz and Heritage Festival in April.Amy Harris/Invision/AP Since the late-1970s, when David Byrne formed the iconic (and alas, now-defunct) Talking Heads, his career has been an endless stream of fascinating side projects, starting with his super-weird, super-cool Brian Eno collab, My Life in the Bush of Ghosts, and his scoring of choreographer Twyla Tharp's The Catherine Wheel. He founded his own World Music label, Luaka Bop, and wrote half a dozen books, including the best-selling quasi-memoir How Music Works. His obsession with the National Color Guard Championships led to a documentary called Contemporary Color. Most recently, his American Utopia tour featured dancers and musicians untethered from the standard concert setup by means of wireless and wearable instruments--nary an amp nor drumset in sight. In November, as the tour wrapped up, came the re-release of Byrne's 1986 film, True Stories, which explores the inner lives and outer quirks of residents of a fictional Texas town and is based on stories from tabloid newspapers.

It's an interesting time to be making a case for philosophy in science. On the one hand, some scientists working on ideas such as string theory or the multiverse--ideas that reach far beyond our current means to test them--are forced to make a philosophical defense of research that can't rely on traditional hypothesis testing. On the other hand, some physicists, such as Richard Feynman and Stephen Hawking, were notoriously dismissive of the value of the philosophy of science. Original story reprinted with permission from Quanta Magazine, an editorially independent publication of the Simons Foundation whose mission is to enhance public understanding of science by covering research developments and trends in mathematics and the physical and life sciences. That value is asserted with gentle but firm assurance by Michela Massimi, the recent recipient of the Wilkins-Bernal-Medawar Medal, an award given annually by the UK's Royal Society. Massimi's prize speech, delivered earlier this week, defended both science and the philosophy of science from accusations of irrelevance.