AI Reading for Sunday June 9

NYT on Apple's big reveal tomorrow - NY Times

Another roundup, say it will automatically choose what to run on-device v. in the cloud. - ZDNET

---

Benedict Evans: Building reliable AI products on unreliable models. - Benedict Evans

---

AGI by 2027 and other predictions by OpenAI researcher who left to be an investor. - Geeky Gadgets

---

Improving AI with automated adversarial testing and training. - Wired

An LLM might answer correctly that Tom Cruise's mother was Mary Lou Pfeiffer, and yet answer incorrectly if you ask who Mary Lou Pfeiffer's son is. But by making it play against itself you can improve it. Game-theory-enhanced synthetic data.

---

CEO of Morgan Freeman company says she can sometimes no longer tell that AI-generated videos of him are fake. - Deadline

---

Finally, an AI that will help you buy clothes that fit you, don't run large or small or tight in the wrong places. - CNET

---

AI scans predict heart attack risk by detecting inflammation. - the Guardian

---

There's some fake-ass WAP going around. - UPROXX

---

Please don't do that, Dave, you are scaring me. Ask Claude about changing thread history, blow its mind. - Reddit

---

New product idea: a real Quack Overflow rubber duck debugger that looks at your code, asks you to walk through it and helps you debug it. (via Reddit)

---

Follow the latest AI headlines via SkynetAndChill.com on Bluesky