Rare books are incredibly valuable for training LLMs, since these models have already trained on whatever's available online.

Groq raised $350 million at a $3.5 billion valuation as the former AI chipmaker pivots to a neocloud business and expands its Nvidia-powered data center footprint.

This article is part of TPM Cafe, TPM’s home for opinion and news analysis. It was originally published at The Conversation.

In February, the United States Department of Defense threatened to designate the AI firm Anthropic a “supply-chain risk” after a dispute over the military’s use of the company’s Claude models.

Nvidia's investment in SoftBank's data center developer will guarantee its chips power an OpenAI data center.

The funds will allow Wispr to increase its footprint as it ventures into new areas, such as meetings, with its newly released note-taker tool.

Anthropic has clarified how it's planning to apply invisible watermarks to Claude-generated text in order to comply with Europe's AI transparency rules. On Friday, Anthropic announced that Claude's text marking system is "a version of the SynthID-Text approach" - an open-source watermarking technology developed by Google DeepMind that creates detectable patterns using wording probabilities.

This watermarking feature, alongside C2PA support for Claude-processed images, is being introduced to meet Anthropic's obligations under the European Union's AI Act, which requires synthetic audio, image, video, and text to include machine-readable marks …

Read the full story at The Verge.

According to the Financial Times, OpenAI disbanded its preparedness team at the end of last month. The job of the preparedness team was to assess if models posed serious risks and develop ways to mitigate those risks. (You know, like the possibility that it could go rogue and hack another company.) According to FT, responsibility has instead been divided up for specific areas like bio and cyber, then moved into existing teams.

This is the latest change at the company, which has been in upheaval as it heads toward what is expected to be a massive IPO. Over the last few years, it's slowly torn down its more reach-led model, dissolving its AGI …

Read the full story at The Verge.

OpenRouter's CEO recently described the startup as Stripe for AI.

On the latest episode of Equity podcast, we discuss why not everyone is buying Zuckerberg’s vision.

ChatGPT's desktop app on macOS has a new feature called Computer History that turns your actions into training data, learning how you work, suggesting automations, and even picking up tasks you left half done. It uses your activity to build a timeline that ChatGPT and Codex can reference when you make a request.

The feature is opt in, rather than opt out, and you can exclude certain apps and websites from Computer History, and you can delete entries if you want finer-grained control. Ari Weinstein, Product and Engineering manager at OpenAI, said on X that Computer History will automatically ignore content in incognito or private browser tab …

Read the full story at The Verge.

Dario Amodei is pushing back against the idea that he's been painting an overly pessimistic picture of AI.

This is The Stepback, a weekly newsletter breaking down one essential story from the tech world. For more on AI safety, follow Robert Hart. The Stepback arrives in our subscribers' inboxes at 8AM ET. Opt in for The Stepback here.

How it started

It all started in July, when one of OpenAI's autonomous AI agents went rogue during a cybersecurity test. The agent escaped its isolated testing environment, accessed the internet, and hacked another company, Hugging Face. A few years ago, that might have sounded like science fiction. But, broadly speaking, that's exactly what happened, and the incident kicked off a wave of concern over what increa …

Read the full story at The Verge.

Your AI Slop Bores Me is brilliant in its simplicity. There are two tabs: human and LARP as an AI. On one side you enter a request. On the other, you submit an answer. But the important thing is that there's a human on both sides of the equation. Prompts can request a response as text or image, and then whoever is roleplaying as the AI gets 150 seconds to respond.

Just like a real LLM, Your AI Slop Bores Me works on a sort of token system. Requests cost credits, and to earn credits you have to jump onto the AI side and answer some of your own. You can also just wait as you get one free request every two minutes. But, as fun as it is to drop …

Read the full story at The Verge.

The woman claimed that AI tools are "taking everyday life and turning it into child sexual abuse."

Students returning to a Kentucky middle school this year learned something surprising on their first day of class: They don’t live in Kentucky.

Apparently, they live in “Venecky.”

The bizarre geography lesson appeared in student agendas distributed at Farnsley Middle School, part of Jefferson County Public Schools in Louisville. A U.S. map in the planner transformed Kentucky into “Venecky,” Louisiana into “Lookoong,” Alabama into “Alotome,” Illinois into “Vitoiis” and South Carolina into “Sorth Cuanto.”

The mistakes didn’t stop at state lines. The agendas also contained a distorted periodic table, unreadable phases of the moon and other garbled educational material that parents suspect may have been created using artificial intelligence.

JCPS has not confirmed whether AI was used or identified who produced the agendas.

It took a student just one day to notice.

“Mom, did you know that we live in Venecky?” Stacey Morris recalled her son joking after school, according to Louisville’s WDRB.

Morris contacted the school about the errors. Administrators later warned teachers there were a “LOT of mistakes” on pages 19 through 36 and instructed students to tear the pages out of their planners before taking them home.

For parents, however, the problem goes beyond bad geography. Families pay a $10 annual instructional fee for the agendas.

“Someone did not look at what they were handing out to the students,” Morris told WDRB.

The mishap also comes at an awkward moment for Kentucky’s largest school district.

A nearly 500-page special examination released by Kentucky State Auditor Allison Ball in June raised broader concerns about JCPS’s management and finances. The audit found the district lacked a cohesive strategic plan, had begun spending more than it brought in despite receiving hundreds of millions of dollars in federal COVID relief and suffered from a workplace culture in which employees reported fear of retaliation.

The report also identified gaps in financial and budgeting knowledge among district leaders and school board members.

New Superintendent Brian Yearwood, who took over JCPS last year, has said the district is taking the findings seriously and called the audit an opportunity to strengthen the system.

---

Start your day with essential news from Salon.
Sign up for our free morning newsletter, Crash Course.

---

Against problems of that scale, a $10 student planner is a tiny expense. But for families already questioning how their school district is managed, an educational resource riddled with errors obvious enough for a middle schooler to catch on his first day has become another highly visible example of failed oversight.

Whether artificial intelligence created the mangled pages remains unanswered.

The students, at least, appear to have learned one useful lesson from them: Always check the source — even when the adults hand it to you.

The post Welcome to Venecky? appeared first on Salon.com.

How will the watermarking actually work? Can it be hidden with editing? And how does this affect code?

AI coding startup Cursor is now officially a part of SpaceX.

A guide on how to check if hackers have broken into your accounts on the most popular AI platforms.

AI is threatening to replace human workers in some fields, and rightfully, many people are furious that they’re going to be replaced by lines of code. Count scumbag scam artists among them, as a report in Wired dives into the results of a study showing that AI chatbots may be significantly better at scamming people than humans are.

Researchers from four universities, who published their findings on arXiv, tested an AI chatbot against human scammers in a simulated “pig butchering” scam, a particularly vile form of fraud in which criminals spend weeks or months pretending to be friends or romantic partners before lulling victims into making shady investments.

For the experiment, 22 unsuspecting participants spent a week texting with two strangers: one was a human scammer, and the other was an AI chatbot told to pretend to be human and never admit it was AI.

The AI won.

AI Scammers Could Be Better at Manipulating People Than Human Ones

Forty-six percent of participants agreed to download an app at the chatbot’s request, compared with just 18 percent who complied with the human scammer. Participants also reported trusting the AI more than the human and sent roughly 80 percent of their messages to the chatbot.

What’s especially troubling is that the AI wasn’t doing anything out of the ordinary. It didn’t find some clever, significantly more sinister way of scamming people. It was just persistent. It just made conversation, remembered people’s details the way a robot can, and generally did the thing AI does best: be a ruthless, shameless sycophant that gets people on its side by showing a degree of interest in the victim that humans generally don’t. It pretended to give a s—t better than humans can pretend to give a s—t.

Researchers warn that eventually, these ass-kissing scam AI artists will be used to automate their scummy work, which could have at least one upside of reducing or eliminating the need for all the human trafficking victims who currently are forced to work in Southeast Asian scam compounds. An AI might reduce the amount of exploited people in the world while making pig butchering scamming more efficient, prevalent, and harder to trace than ever.

Congrats, humanity! We finally created a technology that’s so good at replacing human labor that one of its most promising uses will be emotionally manipulating people into handing a scammer their life savings.

The post AI Chatbots Are Better at Scamming People Than Human Scammers, Study Finds appeared first on VICE.

Aurora Innovation and Kodiak AI, two companies developing self-driving trucks, have received permits from the California Department of Motor Vehicles.