Fred Oliveira 🧠
@f
Transforming inscrutable matrices into products. current: @capitalfactory @lisbonai_. prev: @techcrunch, @gumroad, @webreakstuff @oreillymedia
Most things that cause me anxiety are coordination problems.
Still fine-tuning LLMs in a lab? Or are you shipping machines that actually run in the wild? If you care about real-world AI — the kind that works without cloud GPUs or a fleet of DevOps engineers — meet Paul Hetherington (@phethers), robotics entrepreneur and former CEO of…
“We’re bringing back one of the coolest things the web ever produced, but we’re removing the humanity” Imagine “damn Daniel”, but as AI slop
We’re bringing back Vine, but in AI form
whoever gave them access to the eating tool needs to answer for it
AI agents are eating the world
asked Siri to "ring my alarm in 20 minutes and in 25 minutes". I wish I was joking, but it literally set a timer for 45 minutes later.
The new job is having impeccable taste.
Front-end engineer was a job
Nathan's survey is important, well structured, and well worth your time. Took me around 10min to answer because I tried to be thoughtful. If you have around 10 minutes too, do the thing 👇
>300 submissions now to the @stateofaireport survey you can join too at stateof(dot)ai/survey
AI safety people: "more people should hear this message" also AI safety people: "I have no time to be on stage to talk about safety"
When models start reasoning step-by-step, we suddenly get a huge safety gift: a window into their thought process. We could easily lose this if we're not careful. We're publishing a paper urging frontier labs: please don't train away this monitorability. Authored and endorsed…
imagine asking the incumbents about whether they want to be replaced
> Be European Parliament. > Write a suggestion for a pan-European legal entity for startups > Interview mostly legacy industry and their associations pickachuface when literally EVERYONE in the ecosystem says you failed at your task. It's so bad that EVERY startup association…
xAI is an interesting one to watch for an early rogue AI incident: • Does huge amounts of RL (which generates unintended reward hacking behaviour) • Moving very fast, deploys immediately • Has more compute than talented staff • Not doing any safety stuff as far as anyone can…
We ran a randomized controlled trial to see how much AI coding tools speed up experienced open-source developers. The results surprised us: Developers thought they were 20% faster with AI tools, but they were actually 19% slower when they had access to AI than when they didn't.
not enough people pay attention to the potential downsides of creating something smarter than us. People that haven't read ai-2027.com can now watch this video 👇 super happy that the @80000Hours folks created this.
The AI 2027 scenario is terrifying and important. More people should be thinking about how radical change might come over the next few years, how likely it is, and how a sane world would be reacting to it. We want to bring you into the story, and the conversation. Video here:
listen kids, I'll let you play around with the blockchain all you want. But first, this:
Ozempic, but for IQ.
if you are looking for a design job, you need to read this job listing and if you're not looking for a design job, maybe you should read this job listing too!
I’m hiring a product designer for Sudowrite. This is a dream job for a designer who loves to write. The beginning of the listing is… a mini therapy sesh for me. You’d think it’d be chill after that. But no, it goes on to talk about our company cult… sudowrite.com/jobs/designer
dear reader: maybe this is you? 👇
Probably last chance to apply to humanaligned.ai: Human-aligned AI Summer School. The school is at full capacity, but some people cancelled -> again some spots.
We’re back with AI FR (AI For Real) - a live podcast about building AI in the open. This time, we’re going physical. @pgte hosts @VilleKuosmanen, robotics engineer and frontier tinkerer, for a deep dive into AI with arms and legs. Monday, 7 July 20h BST Live @pgte
I can’t believe it’s already the 2-yr anniversary of OpenAI announcing it would solve the alignment problem in 4 years. Such a relief that Ilya, Jan, Leopold, and the rest must now be ~50% done with the breakthroughs needed to steer/control superhuman AI systems
We need new technical breakthroughs to steer and control AI systems much smarter than us. Our new Superalignment team aims to solve this problem within 4 years, and we’re dedicating 20% of the compute we've secured to date towards this problem. Join us! openai.com/blog/introduci…
I don't think I'll ever hear "chatGPT" and not think about this automatic youtube caption again.
