AI Reading for Monday September 23!

Jony Ive confirms he is working on a product with OpenAI. - NY Times

---

7 ways to benefit from OpenAI's advanced voice assistant, which might be available to more people shortly. - WSJ

Expanded access maybe coming Tuesday? - Twitter

---

The hottest AI startups are raising back-to-back funding rounds. Suddenly, it's feeling very 2021 in Silicon Valley for a chosen few. - Business Insider

---

An AI can beat CAPTCHA tests 100 per cent of the time. If you can't know who is human, everything will have to go biometric, will have to have your retina scanned by a Google orb or a blockchain app or something. - New Scientist

---

You can't really study online culture anymore because you can't tell what's human and what's AI. - GitHub

---

When you call a restaurant now, you may be chatting with AI. - Ars Technica

---

Russians build AI poker bots, make money, eventually find it a tough road and no one wants to play with them, pivot to skill-based matchmaking tech. Let new players play some unskilled bots and think they have a shot, and gradually get them hooked. - Bloomberg

---

How could we even know if an AI is sentient? - Nature

---

Google has a data science AI assistant in its labs. - Google

---

Cloudflare's marketplace will let websites charge AI bots for scraping. - TechCrunch

---

US to ban Chinese connected car software and hardware, citing security, surveillance, remote control risks. - Ars Technica

---

The difference between gpt-4 and o1-preview is sort of like the difference between imitation learning and reinforcement learning. - Understanding AI

Mastery requires the ability to observe and imitate, explore and experiment and practice deliberately (RL), and theorize (understand a well-defined solution space and intentionally search it).

---

Benioff calls Copilot the new Clippy and a science project. Fair enough, hasn't fully realized its promises yet but the potential is there. - Business Insider

---

The NPR emotional timbre of the AI announcers is pretty wild. I wonder if Google will buy Notion and make a big play for the note-taking / online collaboration space, or if they will screw it up. - VentureBeat

---

I tried Google’s new NotebookLM one-click AI podcast creator, and now I don’t know what’s real anymore. - TechRadar

---

My own podcast link - YouTube

(sounds uncannily natural but the discourse is nonsense, wrong content for this format)

---

Luma, Runway ship APIs. - VentureBeat

---

SiFive unveils new open-source chip designs to keep ARM honest and on its toes. - SiliconANGLE

---

AI is a double-edged sword for cybersecurity, giving script kiddies better skills and better phishing including deepfakes, while giving cybersecurity pros better tools to find and fix vulnerabilities, better surveillance and log monitoring. - Hacker Noon

---

Prison inmates in Finland are working as AI data labelers - Yahoo News

---

Once AI has been trained on your private data, there's literally no way to delete it. It's like trying to unlearn Baby Shark or the Barney theme song. - Medium

---

Google opens philanthropy checkbook for AI in education. - Education Week

---

Get AI on your phone from apps like Arc, without upgrading to a new iPhone with Apple Intelligence. - NY Times

---

Meta Connect preview: Cheaper AR, new smart glasses and plenty of AI. - Engadget

---

OpenAI leased a new HQ originally developed as a life sciences property. - Hoodline

---

AI processes sensor inputs from a universe of robots, uses them to generate imaginary world models which are then used to train robots in a virtual environment before putting them in the real world. - 1X

(why not 10X?)

---

LinkedIn suspends use of UK data for AI training after regulator raises questions - BBC

---

Inedible slop from AI slop recipes? - NPR

---

Using AI to remake LOTR in a trailer park, or in the style of Blade Runner, or as a 20s silent film. - Looper

---

Follow the latest AI headlines via SkynetAndChill.com on Bluesky