An anonymous reader quotes a report from Ars Technica: The system prompt for OpenAI's Codex CLI contains a perplexing and repeated warning for the most recent GPT model to "never talk about goblins, gremlins, raccoons, trolls, ogres, pigeons, or other animals or creatures unless it is absolutely and unambiguously relevant to the user's query." The explicit operational warning was made public last week as part of the latest open source code for Codex CLI that OpenAI posted on GitHub. The prohibition is repeated twice in a 3,500-plus word set of "base instructions" for the recently released GPT-5.5, alongside more anodyne reminders not to "use emojis or em dashes unless explicitly instructed" and to "never use destructive commands like 'git reset --hard' or 'git checkout --' unless the user has clearly asked for that operation." Separate system prompt instructions for earlier models contained in the same JSON file do not contain the specific prohibition against mentioning goblins and other creatures, suggesting OpenAI is fighting a new problem that has popped up in its latest model release. Anecdotal evidence on social media shows some users complaining about GPT's penchant for focusing on goblins in completely unrelated conversations in recent days. Update: OpenAI has published a blog post explaining "where the goblins came from." In short, a training signal meant to encourage its "Nerdy" personality accidentally rewarded creature-heavy metaphors, causing words like "goblins" and "gremlins" to spread beyond that personality into broader model behavior. OpenAI says it has since retired the Nerdy personality, removed the goblin-friendly reward signal, and filtered creature-word examples from training data to keep the quirk from resurfacing in inappropriate contexts.

Read more of this story at Slashdot.

Longtime Slashdot reader schwit1 shares a report from ZeroHedge: The Justice Department on Tuesday sued Cloudera, accusing the enterprise data and artificial intelligence company of deliberately engineering a hiring process that excluded American workers from at least seven lucrative technology positions while the firm pursued permanent residency sponsorship for foreign workers on temporary visas. In a 14-page complaint filed with the Office of the Chief Administrative Hearing Officer, the department's Civil Rights Division alleges that Cloudera, from March 31, 2024, through at least January 28, 2025, instructed job candidates to submit applications to a dedicated email address, amerijobpostings@cloudera.com, that rejected all external messages with an automated bounce-back error. The company did not advertise the roles on its public careers website or accept applications through its standard portal, as it did for non-sponsorship positions. Cloudera then attested to the Department of Labor that it could not locate any qualified U.S. workers for the roles, which paid between approximately $180,000 and $294,000 annually, according to the filing. The positions included a Product Manager role in Santa Clara, California, with a listed salary range of $170,186 to $190,000. The case marks one of the most detailed enforcement actions under the Justice Department's Protecting U.S. Workers Initiative, which was relaunched last year and has already produced 10 settlements targeting employers accused of discriminating against American workers in favor of temporary visa holders. "Employers cannot use the PERM sponsorship process as a backdoor for discriminating against U.S. workers," Assistant Attorney General Harmeet K. Dhillon of the Civil Rights Division said in a statement. "The Division will not hesitate to sue companies who intentionally deter U.S. workers from applying to American jobs."

Read more of this story at Slashdot.

Tesla has produced the first Semi from its new high-volume production line at Gigafactory Nevada, a milestone for the long-delayed electric Class 8 truck program after years of pilot builds and delays. Electrek reports: The Tesla Semi has had one of the longest gestation periods in Tesla's history. First unveiled in 2017, the truck was originally promised for production in 2019. That target slipped repeatedly -- to 2020, then 2021, then 2022 -- before Tesla finally delivered a handful of units to PepsiCo in late 2022. Those early trucks were essentially hand-built on a pilot line. Tesla spent the next three years refining the design, cutting roughly 1,000 lbs from the truck, and building out a dedicated factory adjacent to Gigafactory Nevada in Sparks. The company revealed the final production specs in February, confirming two trims: a Standard Range with 325 miles at full 82,000-lb gross combination weight, and a Long Range with 500 miles of range. Tesla is quoting $290,000 for the 500-mile Long Range version and roughly $260,000 for the Standard Range -- making it the lowest-priced Class 8 battery electric tractor on the market. The shift from a pilot line to a high-volume production line is significant. Tesla's Semi factory is designed for an annual capacity of 50,000 trucks, though the company will ramp gradually. Analysts project deliveries between 5,000 and 15,000 units in 2026, but that sounds way too optimistic. [...] Both trims feature an 800-kW tri-motor drivetrain producing 1,072 hp and support 1.2-MW Megacharger speeds, restoring 60% of range in roughly 30 minutes -- conveniently timed around a driver's mandatory rest break. Tesla has opened its first Megacharger station in Ontario, California, and has mapped 66 Megacharger locations across 15 states.

Read more of this story at Slashdot.

An anonymous reader quotes a report from the San Francisco Chronicle: Elon Musk returned to the witness stand Wednesday in Oakland federal court for a second day of testimony in his case against OpenAI, detailing his shift from being an enthusiastic supporter of the nonprofit to feeling betrayed. He also clashed repeatedly with OpenAI's attorney over questions that Musk believed were unfair. He said his feelings towards OpenAI CEO Sam Altman and President Greg Brockman shifted from a "phase one" of support, "phase two" of doubts, and finally "phase three, where I'm sure they're looting the nonprofit. We're currently in phase three," Musk said with a chuckle. Musk said he was a "fool" for giving OpenAI "$38 million of essentially free funding to create what would become an $800 billion company," of which he has no equity stake. In his 2024 lawsuit, Musk alleged breach of charitable trust and unjust enrichment, arguing OpenAI abandoned its original nonprofit mission to benefit humanity to pursue financial gain. OpenAI's lawyer William Savitt argued Tuesday during his opening statement that the nonprofit entity remains in control of the for-profit public benefit corporation and is now one of the most well-funded nonprofits in the world. Musk is seeking to oust Altman from OpenAI's board and upwards of $134 billion in damages, which he said would be used to fund OpenAI's nonprofit mission. During cross-examination, Savitt clashed with Musk over questioning. Savitt asked whether Musk had contributed $38 million to OpenAI, rather than the $100 million that he later claimed to have invested on X. Musk said he also contributed his reputation to the company and came up with the idea for the name, leading Savitt to ask Musk to respond yes or no to "simple" questions. "Your questions are not simple. They're designed to trick me, essentially," Musk said, adding that he had to elaborate or it would mislead the jury. He compared Savitt's questions to asking, "have you stopped beating your wife?" Judge Yvonne Gonzalez Rogers intervened, leading Musk to answer yes to the $38 million investment amount. The world's richest man said his doubts grew and by late 2022, he thought "wait a second, these guys are betraying their promise. They're breaking the deal." "I started to lose confidence that they were telling me the truth," Musk said. A turning point was co-defendent Microsoft's investment of billions of dollars into OpenAI, Musk said. On October 23, 2022, Musk texted Altman that he was "disturbed" to see OpenAI's valuation of $20 billion in the wake of the Microsoft deal. Musk called the deal a "bait and switch," since a nonprofit doesn't have a valuation. OpenAI had "for all intents and purposes" become primarily a for-profit company, Musk argued. Altman responded to Musk by text that "I agree this feels bad," saying that OpenAI had previously offered equity in the company but Musk hadn't wanted it at the time. Altman said the company was happy to offer equity in the future. Musk said it "didn't seem to make sense to me" to hold equity in what should be a nonprofit. Musk also testified about former OpenAI board member Shivon Zilis, who lives with him, is the mother of four of his children, and served as a senior advisor at Neuralink. He denied that she shared sensitive OpenAI information with him. Court evidence showed Musk had encouraged her to stay close to OpenAI to "keep info flowing" and had approved Neuralink recruiting OpenAI employees, which he defended by saying workers are free to change jobs. "It's a free country," Musk said. Recap: Musk Testifies OpenAI Was Created As Nonprofit To Counter Google (Day Two) Elon Musk and OpenAI CEO Sam Altman Head To Court (Day One)

Read more of this story at Slashdot.