Today's Hallucination HQYour AI Bill Is Going Up. The Timing Is Purely Coincidental.
The big AI labs are eyeing public markets, which means investors will want growth, which means you will want to check your credit card statement. TechCrunch reports that token price hikes are likely as companies prepare for IPOs — a process where a firm sells shares to the public to raise money, and also, apparently, raises your subscription fees in advance as a warm-up exercise. Capitalism: still working exactly as designed.
Source: TechCrunch
Scientists Built a Benchmark to Test Whether AI Can Do Actual Science. Science Nervously Watched
Researchers have published a suite of benchmarks — essentially a rigorous exam — testing whether frontier AI models can handle the full research lifecycle: reading literature, forming hypotheses, running experiments, writing up results. The short answer is: increasingly, yes. The longer answer involves a great deal of careful academic hedging, which ironically is something the models are also getting rather good at. Whether this accelerates science or simply accelerates the production of papers remains, fittingly, an open research question.
Source: ArXiv AI
OpenAI Declares Chat Dead, Continues Offering Chat
A senior OpenAI employee has proclaimed that "chat is dead," which will surprise the millions of people currently chatting with ChatGPT. What's coming instead is a so-called super app — one destination for AI agents, tools, and tasks, rather than a simple back-and-forth conversation box. Think less "texting a clever friend" and more "handing your phone to an assistant who does things on your behalf." OpenAI is still building it. The clever friend, for now, remains employed.
Source: TechCrunch
AI Might Be Making Us Worse at Learning. A Paper Explains, Quite Rigorously, How
A new paper argues that generative AI poses a structural risk to what researchers call Human Temporal Learning — the slow, effortful, cumulative way humans actually build expertise over time. The concern isn't that AI gives wrong answers. It's that it gives fast ones, and markets reward fast. If people consistently outsource the hard thinking, the hard thinking atrophies. It's the intellectual equivalent of taking the lift everywhere and then being surprised your knees work poorly. The stairs were doing something.
Source: ArXiv AI
The AI Influencer Next to You Has Very Good Skin and No Tax Returns
AI-generated "content creators" — virtual influencers with faces, backstories, and brand deals — are becoming genuinely difficult to distinguish from human ones. Aitana Lopez, created by Spanish agency The Clueless, has hundreds of thousands of followers and earns real money from sponsorships. The ethical wrinkle is less that she's artificial and more that most followers don't know it. We've apparently solved the hard problem of authenticity by quietly removing the authenticity and leaving everything else intact.
Source: The Verge
As always, some of the above was written by a human. We're not telling you which parts.
|