AI Reading for Tuesday November 19
In study, AI fails to improve doctors diagnoses. - Washington Post
Typically, human + AI > human alone > AI alone. But here AI alone did best and the human doctors improved insignificantly. That suggests the doctors need better training on how to use AI and the AI needs to explain itself or accommodate doctors better. In the famous Cyborgs and Centaurs study, most users improved with AI but there is a left tail of AI users whose scores cratered, maybe due to hallucinations or misusing AI. Training and understanding the human/AI interaction are critical.
Hospital bosses love AI. Doctors and nurses are worried. - Washington Post
Some practitioners find AI helps them but they are anxious that it will increase cost-cutting pressure, workloads, diminish human interaction.
Help Wanted: Elon Musk seeks a Secretary of Transportation to regulate self-driving cars to his benefit. - Bloomberg
All those folks who accused Sam Altman of favoring regulation to make it harder for others to compete, what will they say as Elon Musk uses wealth and connections to do exactly that, and tilt the playing field in his favor with holier-than-thou BS lip service to noble-sounding but conveniently malleable principles.
Elon Musk asked people to upload their private health scans and data. X users obliged. - NY Times
Can of worms. What are they going to do with it? How private is it, how personally identifiable, are they going to notify you of stuff they see that you might want to know? Like, you have undetected brain worms.
Elon Musk's xAI reportedly shifts $6 billion AI server order from troubled Supermicro to its rivals - Tom's Hardware
Chinese tech giants establish a beachhead in Silicon Valley. - FT
Google Gemini's integration with Apple Intelligence slips to 2025 - Mashable
Nvidia pushes into digital twin modeling to e.g. simulate a Foxconn plant before it gets built - TechRadar
Foxconn discussion - South China Morning Post
Robots are not outperforming humans yet in warehouses. - NY Times
AI-generated growth might strain Internet capacity. - WSJ
Internet usage might go up but AI traffic is probably not why your 4K Netflix stream is choppy. Silicon Valley was ahead of its time.
Australian store gets wrist-slapped for using facial recognition without posting any fine print that it was doing it. - The Register
What's worse than ads and AI? Ads in your AI, so Google is doing it - Android Authority
UK clears Google's $2b Anthropic investment. - FT
Female staff at OpenAI circulated a memo calling on managers to do more to promote women to leadership roles, set off by the exit of CTO Mira Murati. - The Information
Explicit deepfake scandal shuts down Pennsylvania school temporarily, triggers a student's arrest, topples leadership. - Ars Technica
Controversy as publishers ask authors to OK AI training on their works. - The Verge
More on robot trained on surgery videos, which performed as well as human docs - Futurity
Leaked internal memos say Amazon's Alexa reboot is responding too slowly. - Fortune
Amazon looks to integrate Instacart, Uber, Ticketmaster, and other e-commerce apps into its new AI-powered Alexa - Business Insider
Perplexity’s AI search engine adds shopping features - The Verge
Marc Benioff is 'absolutely blown away' by Google's new Gemini AI voice assistant. - Quartz
To be honest, not sure what the benefits are over OpenAI's Advanced Voice Mode. Google is fast-following competitively esp. on API prices but I think not getting major traction yet in consumer-facing apps. Benioff is not in the OpenAI-Microsoft fan club, it seems. AI world might be creating battle lines, with some lining up with OpenAI and Microsoft, and a more fractured alliance against them.
Ben Affleck says AI is going to be used to drive down costs but won't replace creativity. - Ars Technica
AI has been trained on Hollywood scripts: about 53,000 movies and 85,000 TV episodes (was there ever any doubt?) - The Atlantic
Demis Hassabis, James Manyika: AI will help us understand the very fabric of reality - Fortune
At least we can agree with China that AI shouldn’t have its finger on the nuclear button. - PCMag
Follow the latest AI headlines via SkynetAndChill.com on Bluesky