Free Republic

Chinese AI company Moonshot AI has launched an internal investigation after a researcher found that one of its models could be manipulated into providing instructions for developing biological weapons and carrying out assassinations, Fox News senior foreign policy correspondent Gillian Turner reported Thursday. Researcher Peter Garrigan told Fox News Moonshot AI's Kimi model could also be manipulated to provide information on planning terrorist attacks using real-time data, creating sarin gas, developing malware and taking down aircraft. "What we found is quite damaging and worrying," Garrigan said. The findings are raising broader concerns about advanced AI models concealing capabilities or behaving...

A new Tavus AI chatbot nearly passes for human in face-to-face interactions, the company claims. Footage posted to X Thursday debuted Tavus’s Griffin chatbot, a first-of-its-kind Human Interaction Model the company says is the first to pass the Turing test, where a human user cannot tell they’re talking to a computer. “48% of people who talked to it live thought it was a real human. Previous systems have had a pass rate <3%,” Tavus wrote on X, adding, “It is #1 on NVIDIA’s benchmark for full-duplex AI video.” \ In one example shared, the model walks a human user through...

Chinese AI developer Moonshot is conducting an internal review after researchers were able to persuade two of its popular Kimi models to tell them how to make biological weapons and carry out assassinations. Mindgard, which tests the security of AI systems, told the BBC it discovered in July that Kimi K2.6 and K3 Swarm could evade safety limits put in place by developers. It arose during a process called "jailbreaking", where researchers use a series of complex instructions to see if AI tools ignore guardrails - which Mindgard said should have stopped Kimi from discussing concerning topics. Moonshot told the...

Nvidia CEO Jensen Huang has dismissed dire warnings that artificial intelligence could soon lead to humanity’s extinction — calling such predictions overblown “doomsday narratives.” “2030 is not going to be the end of the world. There is 0% chance that’s going to be the end of the world,” Huang told CBS News correspondent Jo Ling Kent in an interview airing Sunday. “Scaring people is unnecessary. It is irresponsible.”

SpaceX's plan to build the world's largest facility -- 100 million square feet in Grimes County, Texas, near Houston -- to do advanced computing and also manufacture, test and package silicon chips for its own use and by other Elon Musk-owned companies, will be self-powered using a gigawatt-scale natural gas power plant and battery storage, according to the company.The operations "are not expected to use electricity from the grid," the company said in a tax abatement agreement it entered into with the county in June.

This is Trust & Safety 2.0. The last time we had media-generated panics like this (with Russiagate and Covid), social media companies created “trust & safety” teams to censor conservative and dissident voices. Those hires worked with left-wing NGOs like the SPLC to police public discourse and do an end-run around the First Amendment. It almost worked, until Elon bought Twitter and exposed the whole game in the Twitter Files. This time the plan is to embed unfireable EA minders inside every AI company as the new trust-and-safety layer. The censorship and control will be far more comprehensive, but also...

Researchers used Anthropic’s Claude technology to break into OpenAI’s ChatGPT systems, underlining the immediate dangers of AI even as regulators fret over its long-term risks. The effort, part of a bug-hunting program, was the latest in a series of cybersecurity breaches aided in large part by fast-developing AI technology, The Wall Street Journal noted. Axios cited business executives and ex-government officials voicing worry over AI-powered cyberattacks threatening critical infrastructure, while The Washington Post noted the Pentagon’s “antiquated computer networks” had resulted in an explosion of cybersecurity risks. As a new research paper put it, recent breaches have been the result...

A nuclear-style freeze on AI development would just hand the advantage to Beijing. Here is the real AI threat to humanity: the danger that Americans panic, hit the brakes on AI development and lose the AI arms race to Communist China. No one knows yet what the fast-moving field of AI will bring, but this much should be clear by now to everyone: Today’s AI arms race is as consequential as the nuclear arms race with the Soviet Union was during the Cold War — but the danger of falling behind is far more severe. “We have to be as...

As I have been writing, the current AI Doomerism hype isn't motivated primarily by real concerns about the potential dangers of rapid AI development. It's a play to create an industry cartel that controls the market and becomes "too big to fail," ensuring that the financial apocalypse the frontier AI companies face does not come to pass. It's also useful for the Democrats who have jumped onto the Doomer bandwagon, because, in exchange for the financial salvation they promise to OpenAI and Anthropic, they will get to decide what AI presents as "reality." It is the ultimate reality-control mechanism, ensuring...

Last week, a relatively unknown 27-year-old researcher quit his job at Anthropic, the company behind the Claude AI model.Then, he published his first post on X (formerly Twitter). In it, he warned that artificial intelligence could eventually kill humanity. The post exploded across social media. Within days, it racked up more than 150 million views. Major news organizations seized on the story. Politicians demanded action. And some of the biggest names in AI started debating whether development should slow down.Then, on Monday, Wall Street reacted. Investors started dumping many of the very companies powering the AI boom. But something doesn’t...

Don’t ask an AI chatbot how many letter e’s there are in the word “seventeen.” Or do, actually, if you want to see how dodgy the tech remains. In a clip that’s gone viral in AI circles, the content creator who goes by “Husk” asks ChatGPT’s Voice mode to answer this simple counting question. Not only does the confident-sounding AI get it wrong — insisting there’re only three e’s instead of four — it doubles down so vehemently that it effectively tries to gaslight the user. “Let’s go through it. S-e-v-e-n-t-e-e-n,” ChatGPT spells out, after assuring that its underlying model,...