Today's Hallucination HQSalesforce Drops $3.6 Billion on a Chatbot. Modest, Really.
Salesforce has acquired Fin, an AI customer service platform, for $3.6 billion — roughly the GDP of a small island nation, spent on something that will tell your customers their order is "being processed." The goal is to bolt Fin's technology into Agentforce, Salesforce's platform for building custom AI agents that automate business tasks. Whether this makes Agentforce meaningfully better, or just meaningfully more expensive, remains the kind of question $3.6 billion rarely answers upfront.
Source: TechCrunch
The White House Pulled the Plug. The Rest of the World Took Notes.
Over the weekend, the White House instructed Anthropic to take its newest AI models — Fable and Mythos — offline, blocking access for all foreign nationals, including Anthropic's own staff abroad. Anthropic complied, noting it had "little choice," which is one way to describe having no choice. For governments outside the US already nervous about depending on American AI infrastructure, the incident was less a warning shot and more a full demonstration of exactly what they'd been worried about.
Source: The Verge
The Man Who Taught a Brain Implant to Actually Be Useful
Casey Harrell has ALS, is paralysed, and has spent nearly three years with electrodes embedded in his brain — which, for the record, makes him considerably more committed to this technology than most people who claim to be "early adopters." His brain-computer interface, or BCI, first helped him produce spoken sentences in 2023 with a research team guiding him. He has since clocked thousands of hours using it independently, making him, by any reasonable measure, the most dedicated human-computer interface tester alive.
Source: MIT Technology Review
Dozens of Cybersecurity Experts Agree: This Ban Is a Bad Idea
A coalition of seasoned cybersecurity professionals has written to the White House urging it to reverse the export controls on Anthropic's Fable and Mythos models, which they describe as "dangerous." Their argument is straightforward: the defenders who protect critical software need access to the most capable AI tools available, and handing that advantage exclusively to domestic actors while foreign threats carry on regardless is, to use a technical term, backwards. The government has not yet responded, presumably because it is very busy.
Source: TechCrunch
Annual AI Report Asks Whether Humans Have Any Idea What They're Doing
Stanford's AI Index has published its ninth annual report, and the headline finding is not about what AI can do — it's about whether the systems surrounding it can keep pace. Governance frameworks, evaluation methods, and data infrastructure are all, politely speaking, struggling to catch up with the technology they're meant to manage. Nine editions in, the report has the weary authority of a doctor who keeps giving the same advice and watching the patient order a large fries anyway.
Source: ArXiv AI
As always, the technology is fine. It's everything else that needs work.
|