AI Reading for Wednesday January 15
ChatGPT can now handle scheduled tasks, reminders and to-dos - The Verge
OpenAI's AI reasoning model 'thinks' in Chinese sometimes and no one really knows why - TechCrunch
Maybe Chinese is better for reasoning about some tasks, like French is more poetic and has more emotional payload.
Microsoft, OpenAI ask federal judge to toss news orgs’ copyright claims - Courthouse News Service
An interesting model, OpenAI will fund local newsrooms if it can use the content. - Axios
Why can’t Google do this.
Give police facial recognition, they rely on it and stop doing police work.
You really have to drill into people that AI can process a lot of info and advise, but you can’t arrest someone and send them to jail based on a blurry photo because AI thinks they are the best match. The info surfaced by AI has to stand on its own and be credible to an independent observer. People have to be responsible for choices, without using some appeal to AI authority.
AI Agents Are Here. How Much Should We Let Them Do? - WIRED
Half of UK companies to invest in AI instead of staff. - The Register
AI for governance and representative democracy - The Conversation
There’s Arrow’s impossibility theorem, multiple equilibria, temporal consistency, how people choose depends on how you frame the choices. AI could help read the preferences of voters and rank and explain choices. You really have to drill into people that you AI can be a (possibly biased and hallucinatory) advisor but people have to be responsible for their choices.
Follow the latest AI headlines via SkynetAndChill.com on Bluesky