“The paper adds to a small but growing body of evidence that today’s most advanced AI models are becoming capable of strategic deception.”
From Time Magazine, 3 days before publication of the article. Is this interpretation inflammatory or just exaggerated?
“Under Fukushima’s new rules, if the rubbish remains unsorted for a week, city workers can go through it and try to identify the offenders via items such as mail. The violators will be issued a verbal warning, followed by a written advisory, before the last resort: having their names published on the government website.”
From BBC.
Got too much red algae? Feed it to the cows. It reduces methane by 40%.
US media is not following this development in AI. A whistleblower from OpenAI who recently left the company is found dead. Police say it is suicide, but it is not revealed for more than a week. And it takes the BBC to report on it before it makes news in the US.
NYTimes interviewed him a few weeks ago, but has not printed anything about his death yet. Mysterious.
I polled my students and the ONLY AI they used was ChatGPT. It felt like the iPhone phenomenon, but even worse. They were unaware of any others. I am set to remedy that in the next couple of weeks with a 3-week intro to “Using AI for Language Learning”, the final module of three of my classes.
It seems that AI researchers use different AIs for different purposes. The new favorite in Silicon Valley is Anthropic’s Claude AI. Here is Kevin Roose from the NYTimes on this. (Gift article)
I’m curious about what you use for yourself and for students. I’ve been playing around with TalkPal to improve my Japanese. It seems to work. I do get a flavor of how nervous my students are using AI for learning. I hope to teach them to feel more in control. I’m beta testing IDoRecall where you can add spaced repetition learning to almost anything you encounter, and link back to it if you don’t remember the flash cards. I’ve signed up for a similar site at YouLearn. Still looking into both.