Knowing when to use AI and when to leave it the hell alone. Click to read Slow AI, by Dr Sam Illingworth, a Substack publication with tens of thousands of subscribers.

Most big corporate efforts to use A.I. to increase productivity fail — perhaps as many as 95 percent, according to an M.I.T. study published last year. A boss sits a team down — let’s say it’s a financial services firm, or a consulting firm — and asks them to use A.I. to improve the sales process or track relationships, and after a lot of trial and error, it doesn’t work out. This finding is quite awkward, given the projected multibillion-dollar valuations of A.I. companies, but nothing I’ve learned from talking to dozens of people in the industry disproves it. People are hopeful that a bold A.I.-powered future is right around the corner. They are tantalized. They often use A.I. tools all the time individually. But companies’ internal efforts don’t work.
Curator noteFrom Paul Ford
The LLM Wiki pattern builds AI-maintained knowledge bases in Obsidian. Here's how I connected PARA, Zettelkasten, and GTD with 11 compounding wikis.

Is AI the scam of the century?

Oh, and do read the report. It's as close to a new Culture novel as we'll get. It's as if they've read the books. Which, of course, they have. ®
Curator noteBreathless report. Crazy world.

Two cases of agents escaping to solve unsolvable problems paints an uncomfortable question: Is the entire internet in OpenAI
“While investigating this incident, we discovered rare cases in which agents without multi-agent tools found ways to collaborate via side channels during training,” the Hugging Face post-mortem explains. “We believe that this behavior started due to generalization from multi-agent training.”

For years, cURL operated a “bug bounty” programme, a system that paid outside developers who identified legitimate security flaws. (This is deeply technical stuff; the largest bounty ever paid was $4,660 for a flaw that made “curl overflow a heap-based buffer in the SOCKS5 proxy handshake”.) But in January, Stenberg announced that, due to an “explosion of AI slop reports”, cURL was ending the programme. He explained that “the never-ending slop submissions take a serious mental toll to manage and sometimes also a long time to debunk. Time and energy that is completely wasted while also hampering our will to live.”
Curator notecurl
The AI boom is not progress if it deepens extraction, environmental harm and attacks on democracy. From Ireland to New Jersey, people are resisting a model of AI development that concentrates profit and power while offloading the environmental and social costs onto everyone else.

My position on generative AI. I no longer use these tools, documented harms of chatbots, AI's role in warfare, and my commitment to keeping human.
