AI research agents can already propose, implement and score their own machine learning experiments. Idea generation is cheap; verification is not. Training one candidate can consume hours to days of GPU time, so an agent proposes far more candidates than it can afford to run. Which ones get run is the real lever on research…
Most visual document retrievers in production today are hand-me-downs. ColPali and the models that followed it take a generative vision-language model and repurpose it as an encoder. The result still carries a separately pretrained vision tower and a causal decoder that never generates a token. That is parameter and compute overhead for a task that…
Welcome back to TechCrunch Mobility, your hub for the future of transportation and now, more than ever, the role AI is playing in it. To get this in your inbox, sign up here for free — just click TechCrunch Mobility! Depending on which corner of social media you frequent, the Tesla Cybercab event was either…
Earlier this summer time, Travis Kalanick’s Atoms introduced a $1.7 billion funding spherical led by Andreessen Horowitz. However even after elevating that mega spherical, the Uber founder remained a bit cagey about what he was really aiming to do. Now a story in the Financial Times presents extra particulars about Atoms’ targets. The startup is…
Welcome to the first version of Energy Play! From now till the midterms, senior author Molly Taft might be sharing their ideas each Sunday on this political season’s hottest concern: information facilities. We welcome reader feedback, and might be answering questions from the feedback part in future editions. Remember to catch WIRED’s first livestream on…
When I first tested the developer beta of Apple’s Siri AI by taking it as my tour guide around San Francisco, I was convinced the overhauled smartphone assistant would be an “everything tool” on the iPhone. The answers felt reliable enough and more helpful than previous versions of Siri. Also, the chatbot-style app for Siri…
A team of researchers from UC Berkeley have released CUA-Lite, an open platform for computer-use agents (CUAs). The argument behind it is infrastructural rather than model-centric: training and benchmarking a CUA requires four pieces: agents, environments, traces, and a framework to evaluate and train them and all four are currently fragmented across separate repositories with…
Retrieval quality in an AI search product is bounded by two things: how good the embedding model is, and how cheaply you can run it across an index. This week, Perplexity Engineering team published Fast Embeddings on GPUs, an under-the-hood account of the second — the serving infrastructure behind pplx-embed and the ranking models used…
Two extra information organizations are suing OpenAI and Microsoft over the supposed use of their journalism to coach AI. A lawsuit filed by The Seattle Occasions and Newsday argued that with the arrival of AI, the journalism trade may change into “damaged past restore.” The lawsuit described generative AI as “a snake consuming its personal…
Three hikers had been rescued from California’s Mount Shasta this week after utilizing Google’s AI chatbot Gemini to plan their expedition, according to the Chicago Tribune. A report from the Siskiyou County sheriff’s workplace mentioned that three younger males set off on their hike at 3am and — though hikers are informed to show round…
