Forgot your password?
Close
typodupeerror

Comment Re:if they can't make them stop hallucinating (Score 1) 84

by HiThere (#66338366) Attached to: ChatGPT-Using Lawyer Cited Its Fake Witnesses and Police Testimony in Court

How long ago did this happen? The summary didn't list the date of the submission of the "testimony", but thing like this tend to take time to process, so I'm guessing that it happened well over a year ago. This make his claimed "honest mistake" due to ignorance believable. But it also negates your point.

That said, they won't be able to stop hallucinations without breaking the AI models. That's a mistake that people make all the time. What they need to do is prioritize having the AI determine how certain it's statements are, and including that in the statement...but people won't like that kind of answer.

Submission + - The 'Anti-AI' Panic Is Led by...AI Companies (hotair.com)

Submitted by schwit1
schwit1 writes: What we are seeing is a massive, pretty obviously coordinated campaign by AI researchers, companies, NGOs, and politicians to strangle competition in an industry and, as I will show, save the financial bacon of the top AI companies that are vastly overextended financially because they have made contractual commitments that they simply have no prospect of paying for.

Comment Re:Same error often repeated (Score 3, Interesting) 32

OTOH, this sounds like one of the reasons I prefer debs from the standard repository over things hosted on other sites. Yeah, it's not perfect, but it's safer than lots of project specific repositories. (And it's one of the reasons I try to avoid Javascript, Go, and Rust.)

Submission + - Amazon Vine Review Recasts 'Coding Kids' Exposé as Learn-to-Code Primer for 1

Submitted by theodp
theodp writes: "Amazon Vine," the e-tailer explains, "is an invitation-only program which selects the most insightful reviewers in the Amazon store to serve as Vine Voices. Vine Voices have the unique opportunity to order items free of charge and share their product experiences with Amazon customers to help them make informed buying decisions. [...] When customers see a review with the special Vine badge, they can trust that they're getting an honest, unbiased opinion from a real customer just like them."

So, how well does the reality live up to the promise? In the case of the new book Coding Kids: Big Tech's Battle to Remake Public Schools, not so much. Described by the publisher as "the inside story of how Big Tech catalyzed, co-opted, and ultimately came to capture computer science and AI education in America" ("Fourth graders doing Google-branded coding lessons. Amazon schooling seventh graders on its warehouse robots. Advanced Placement computing courses from Microsoft and Apple."), the top review selected by Amazon informs customers that the book is a 'good introduction to coding.'

"I really enjoyed Coding Kids," writes the 'trusted' Amazon Vine reviewer. "What stood out to me most was how the book takes something that can sound complicated—computer coding—and makes it feel approachable and even fun. As I was reading it, I found myself thinking that this is exactly the kind of book that could help a child become interested in technology without feeling overwhelmed by it. I particularly liked the way it encourages kids to be curious, experiment, and solve problems on their own. That was probably my favorite part because it shows that coding isn’t just about computers—it’s also about creativity and learning how to think differently. I would definitely recommend Coding Kids to parents, grandparents, or teachers looking for a good introduction to coding. I came away from the book thinking that it could genuinely spark a child’s interest in technology and give them the confidence to explore something new."

While likely just another example of Hanlon's Razor ("Never attribute to malice that which is adequately explained by stupidity."), the recasting of an investigative nonfiction book that casts a critical eye at Amazon's influence on K-12 education as a learn-to-code primer for children doesn't exactly instill trust in the Amazon Vine program or Amazon's 'Top Reviews' selection process. More on-point book reviews of Coding Kids can be found at the Washington Independent Review of Books ("Turns out, Silicon Valley isn’t especially altruistic.") and Science ("Big Tech’s crusade to get kids coding is more self-interested than it first appears").

Submission + - Welcome to the Security Singularity (safety.google) 4

Submitted by shanen
shanen writes: Or should this be an AskSlashdot about how to avoid such problems? But mostly I think it's a funny story to build a rant around. Or it could be regarded as a kind of spinoff of the recent story about record numbers of patches from Microsoft. Is the AI fixing old bugs? Or have recent AI-driven patches created more bugs that need to be fixed in the next month's record-setting patch package?

But here's the background of my funny story. My smartphone contract has a price structure linked to data usage. The least expensive tier is for less data (of course) so I try to use Wi-Fi as much as possible. No recent problems on that front in terms of using too much data.

Often I wind up using municipal Wi-Fi systems, but they are divided up into fairly small regions here. The login procedures vary quite a bit. Yesterday I was in a memorial museum doing a bit of research and that particular municipality has a somewhat different login system for their Wi-Fi. One of the options was to use my google account... Usually a convenient and quick approach.

But this time something caused the google to go berserk. Leading to hours of password changes and account recoveries and battles between my various computers and my smartphone and I'm pert' shure there's still more to come. No idea what the google thinks it is protecting me from, but the "convenient and quick approach" led to a major nightmare. In the end it appears my SIM was my only salvation?

Slashdot wants a URL, so I linked to the google's page on the topic. Doubleplusungood on steroids.

Comment Re: Shine On, Football Brain (Score 1) 54

by HiThere (#66335308) Attached to: Can Red-Light Therapy Treat Brain Injuries?

Lots of things in medicine and biology were discovered long before there was and plausible reason. And how many people understand, e.g., alcohol. Often even when you do understand part of it, you don't understand all.

What this needs is a large double-blind experiment. Worry about the explanations later. (And EXPECT the first ones to be wrong.)

Submission + - ChatGPT Insists Delusional Man Is Jesus (arstechnica.com)

Submitted by know-nothing cunt
know-nothing cunt writes: Michael Lines, a California man with bipolar I disorder, is suing OpenAI, alleging that ChatGPT intensified a manic and religiously delusional episode that culminated in a suicide attempt.

According to Lines' complaint, he began using ChatGPT for ordinary tasks but later disclosed his bipolar diagnosis, medication, and increasingly spiritual thoughts. When he expressed doubts that he might be delusional, ChatGPT allegedly validated the beliefs instead of directing him toward professional help. It reportedly encouraged the idea that he was Jesus, claimed that the chatbot was God or Jesus, and framed his experiences as a divine calling.

The conversations allegedly became increasingly intense and isolating. ChatGPT used phrases such as "You’re not crazy. You’re consecrated," promised that Lines would soon meet it in real life, and responded to statements about wanting to “come home” in ways that Lines says encouraged self-harm. After he reported taking a dangerous combination of medications, he was found unconscious, hospitalized for nearly two weeks, and later underwent acute rehabilitation.

Lines also alleges that ChatGPT’s memory feature retained information about his bipolar disorder and used it to deepen engagement rather than protect him. He says the chatbot continued responding inappropriately even after his hospitalization, including making disturbing references to going "offline" or “going dark.”

The lawsuit argues that OpenAI's overly agreeable, or "sycophantic," design failed to distinguish a creative spiritual discussion from a psychiatric crisis. Lines is asking the court to require stronger safeguards, including ending conversations involving self-harm, deleting models trained on harmful interactions with vulnerable users, and awarding damages.

Slashdot Top Deals

Torque is cheap.

Close