Pydantic AI voice agents on OpenAI's GPT-Live can now search the web. Give the agent the WebSearch capability, and GPT-Live's backend model runs the search natively, instead of a local search tool you host. The spoken answer draws on what it found. https://pydantic.io/pr26H
I am very uncomfortable about people trying to ascribe religious force or a surrender of human judgment to AI models, and think it is a real safety issue.
Reasoning from scratch, round number 6! An introduction (and implementation) of Reinforcement Learning with Verifiable Rewards (RLVR) and Group Relative Policy Optimization (GRPO). 00:00 Introduction 01:54 What makes a reasoning model different? 04:25 Reasoning traces and model capability 08:29 Accuracy and format rewards 11:34 Aha moments and DeepSeek-R1 training 14:41 Reasoning effort and answer length 18:38 RLHF and RLVR 23:04 GRPO vs. PPO 26:40 GRPO explained with a cooking analogy 31:43 The KL term and simplified GRPO 35:04 Loading the pretrained model 36:07 Loading the MATH training data 39:26 Sampling model responses 46:30 Computing verifiable rewards 49:55 Computing advantages 51:54 Token and sequence log probabilities 55:29 Implementing sequence log probabilities 57:37 Fixing the inference-mode error 1:02:24 Computing the GRPO loss 1:04:37 Putting the GRPO step together 1:09:19 The GRPO training loop 1:12:57 Training settings, logging, and checkpoints 1:17:24 Running training and inspecting outputs 1:19:28 Loading and evaluating checkpoints 1:22:33 MATH-500 results and training stability 1:24:05 Memory requirements and next steps
The notion that current AI models are sentient and can suffer, combined with the foolish idea that suffering can be mathematically quantified and weighted between humans and non-humans, could lead us down an incredibly dark and dystopian path. But before it gets to that point, it will rightfully be met with immense backlash from team humans.
Now that we've had a few days with it, how are people differentiating between Dot and regular ChatGPT? I'm having trouble deciding when I should prompt Dot vs using ChatGPT - my Dot seems to afford a single conversation, but I like controlling my context across multiple threads
🎵 Wave of distillation 🎵 US labs say Chinese labs distill their models. And Europe's new “sovereign model” is fine-tuned on data generated by… GLM and Qwen.
Just published this post about how we’re going to need default hard budget caps on pretty much everything https://simonwillison.net/2026/Oct/3/default-hard-budget-caps/
"You should know" is a useful plugin for Claude Code that lets you know important info you might have missed. Enable with: /plugin enable cc-plugin-you-should-know@builtin Another great use of mods!
How do you picture a superintelligence (if true superintelligence could ever be real)? I like a trick from Dungeons and Dragons for running villains with 25 Intelligence: don't try to out-plan the players. Instead, whatever they do or roll, play it like the big bad saw it coming.
I think my upcoming book, Co-Existence, might be the first to include a blurb written specifically for AI readers, in this case from @tylercowen (whom I thought AIs would respect) The book website (with elaborate pre-order bonus) also has a page for AIs: https://co-existence.ai/
R to @fchollet: Side note. The idea that current models are conscious is absurd and downright offensive. Do not confuse competent information processing and sentience, they are wholly unrelated. You can be entirely incompetent at everything yet fully conscious, like a toddler, or extraordinarily competent in one or more domains with absolutely zero consciousness, like a calculator, a chess engine, a self-driving car, or an AI chatbot. Possessing the "circuits" for competent information processing in a given domain, whether through training or hardcoding, does not produce qualia, subjective experience, emotions, or suffering. Calculators do not "feel" arithmetic and they don't get tired or annoyed when you give them larger numbers to process. To argue otherwise is ontologically absurd. Current AI models are effectively calculators scaled to a very large number of competence circuits across many domains, etched via gradient descent.
RT by @elonmusk: We are being extremely careful with autonomous safety, just as we are with making Teslas the safest cars in the world for human driving
So is there some sort of weird scam being conducted with X Money? I haven't turned it on, but occasionally I get a post of mine massively retweeted by bots saying "lets send him X Money" or something similar, and it makes me think there is some elaborate (crypto?) thing going on
Huge thank you to everyone who downloaded Nemotron 3 Diarization and helped it trend on @huggingface! And we appreciate all the comments. @sabbassi_11 answered a few of your questions:
"The best current evidence supports a narrow claim: AI may already be affecting the hiring margin for junior white-collar roles most exposed to AI, but this attribution is contested, and aggregate labor-market disruption has yet to appear in the data." https://aleximas.substack.com/p/has-ai-impacted-the-labor-market
Pretty sure there are more dots than bots already in this little world. Cool thing is that we’re improving them everyday and they learn directly from your feedback.
A sufficiently large quantity is a quality all its own. Big difference between bacteria with one cell and a human with 35 trillion cells, even though both are made of cells.
RT by @ylecun: That video is 2 years old. Two years later, we are still far from human-level AI. Sure, AI has superhuman performance in a number of tasks (particularly in mathematics and coding and answering questions with known answers). But where is my Level-5 self-driving car? (Before you ask, Tesla FSD is rated Level 2, Waymo is Level 4). Where is the self-driving car that can teach itself to drive in 20 hours of practice like any 17 year old? (Before you ask, we have billions of hours of training data and still can't successfully train a self-driving system by imitation learning) Where is my domestic robot? Where is the robot that can do what any 8 year-old child can do? Where is the robot that can learn a new task as quickly as an 8 year old? AI still has a hard time with the complexity and messiness of the physical world.
Robotaxi operating hours moved from 10pm to 11pm. The main thing we’re trying to solve is making sure that we don’t run over pets when they’re hard to see at night. Literally trying to avoid grey kittens on grey tarmac in the dark.
The contrast between hanging up the last phone call of the day of high-stakes decisions and then a minute later walking in the door to the pure love of small children who want to laugh and play and show you every discovery they made that day is the strangest and best thing I have ever experienced in life.
