Three days after Amazon announced its AI chatbot Q, some employees are sounding alarms about accuracy and privacy issues. From a report: Q is “experiencing severe hallucinations and leaking confidential data,” including the location of AWS data centers, internal discount programs, and unreleased features, according to leaked documents obtained by Platformer. An employee marked the incident as “sev 2,” meaning an incident bad enough to warrant paging engineers at night and make them work through the weekend to fix it.

[…] In a statement, Amazon played down the significance of the employee discussions. “Some employees are sharing feedback through internal channels and ticketing systems, which is standard practice at Amazon,” a spokesperson said. “No security issue was identified as a result of that feedback. We appreciate all of the feedback we’ve already received and will continue to tune Q as it transitions from being a product in preview to being generally available.”

Read more of this story at Slashdot.

Source: Slashdot – Amazon’s Q Has ‘Severe Hallucinations’ and Leaks Confidential Data in Public Preview, Employees Warn

An anonymous reader quotes a report from Ars Technica: In a preprint research paper titled “Does GPT-4 Pass the Turing Test?”, two researchers from UC San Diego pitted OpenAI’s GPT-4 AI language model against human participants, GPT-3.5, and ELIZA to see which could trick participants into thinking it was human with the greatest success. But along the way, the study, which has not been peer-reviewed, found that human participants correctly identified other humans in only 63 percent of the interactions — and that a 1960s computer program surpassed the AI model that powers the free version of ChatGPT. Even with limitations and caveats, which we’ll cover below, the paper presents a thought-provoking comparison between AI model approaches and raises further questions about using the Turing test to evaluate AI model performance.

In the recent study, listed on arXiv at the end of October, UC San Diego researchers Cameron Jones (a PhD student in Cognitive Science) and Benjamin Bergen (a professor in the university’s Department of Cognitive Science) set up a website called turingtest.live, where they hosted a two-player implementation of the Turing test over the Internet with the goal of seeing how well GPT-4, when prompted different ways, could convince people it was human. Through the site, human interrogators interacted with various “AI witnesses” representing either other humans or AI models that included the aforementioned GPT-4, GPT-3.5, and ELIZA, a rules-based conversational program from the 1960s. “The two participants in human matches were randomly assigned to the interrogator and witness roles,” write the researchers. “Witnesses were instructed to convince the interrogator that they were human. Players matched with AI models were always interrogators.”

The experiment involved 652 participants who completed a total of 1,810 sessions, of which 1,405 games were analyzed after excluding certain scenarios like repeated AI games (leading to the expectation of AI model interactions when other humans weren’t online) or personal acquaintance between participants and witnesses, who were sometimes sitting in the same room. Surprisingly, ELIZA, developed in the mid-1960s by computer scientist Joseph Weizenbaum at MIT, scored relatively well during the study, achieving a success rate of 27 percent. GPT-3.5, depending on the prompt, scored a 14 percent success rate, below ELIZA. GPT-4 achieved a success rate of 41 percent, second only to actual humans. “Ultimately, the study’s authors concluded that GPT-4 does not meet the success criteria of the Turing test, reaching neither a 50 percent success rate (greater than a 50/50 chance) nor surpassing the success rate of human participants,” reports Ars. “The researchers speculate that with the right prompt design, GPT-4 or similar models might eventually pass the Turing test. However, the challenge lies in crafting a prompt that mimics the subtlety of human conversation styles. And like GPT-3.5, GPT-4 has also been conditioned not to present itself as human.”

“It seems very likely that much more effective prompts exist, and therefore that our results underestimate GPT-4’s potential performance at the Turing Test,” the authors write.

Read more of this story at Slashdot.

Source: Slashdot – 1960s Chatbot ELIZA Beat OpenAI’s GPT-3.5 In a Recent Turing Test Study

The world’s biggest operational experimental nuclear fusion center has been inaugurated in Naka, Japan. The Guardian reports: The goal of the JT-60SA reactor is to investigate the feasibility of fusion as a safe, large-scale and carbon-free source of net energy — with more energy generated than is put into producing it. The six-storey-high machine, in a hangar in Naka, north of Tokyo, comprises a doughnut-shaped “tokamak” vessel set to contain swirling plasma heated up to 200mC (360mF). It is a joint project between the European Union and Japan, and is the forerunner for its big brother in France, the under-construction International Thermonuclear Experimental Reactor (ITER).

Sam Davis, the deputy project leader for the JT-60SA, said the device would “bring us closer to fusion energy.” “It’s the result of a collaboration between more than 500 scientists and engineers and more than 70 companies throughout Europe and Japan,” Davis said at Friday’s inauguration. The EU energy commissioner, Kadri Simson, said the JT-60SA was “the most advanced tokamak in the world,” and called the start of operations “a milestone for fusion history.” “Fusion has the potential to become a key component for energy mix in the second half of this century,” Simson added.

Read more of this story at Slashdot.

Source: Slashdot – World’s Biggest Experimental Nuclear Fusion Reactor Launched In Japan

An anonymous reader quotes a report from Motherboard: On Wednesday, a group of military and military-adjacent scientists gathered at a conference to discuss the possibility of creating a super soldier. They discussed breeding programs, Marvel movies, The Matrix, and the various technologies the Pentagon is researching with the goal of creating a real life super soldier complete with cybernetic implants and thorny ethical issues surrounding bodily autonomy. The talk happened at the The Interservice/Industry Training, Simulation and Education Conference, or I/ITSEC, an annual conference where military leaders come to talk shop and simulation corporations gather to demo new products. It’s the kind of place where execs and generals don virtual reality helmets and talk about the virtues of VR sims. You could even catch members of congress talking about the importance of simulations and war. “Winning the war of cognition by pushing readiness and lethality boundaries,” reads the official poster for the 2019 I/ITSEC.

It was here, in Orlando, Florida, where five illustrious members of the military-industrial complex gathered to discuss super soldiers at the “Black Swan — Dawn of the Super Soldier” panel. Lauren Reinerman-Jones, an analyst from Defense Acquisition University, moderated a panel that included U.S. Army Developmental Command representatives George Matook and Irwin Hudson, research scientist J.J. Walcutt, and Richard McKinley, who works on “non-invasive brain stimulation” for the Air Force. I/TSEC advertised the panel in its program with a picture of the experts next to a posing Master Chief, the genetically enhanced super soldier from the Halo video game franchise. Throughout the conversation, which covered the nuts and bolts of what’s possible now and what’s about to be possible along with various ethical concerns, references to science fiction and fantasy stories were common. Some of the ideas discussed include synthetic blood, pain-numbing stimulants, limb regeneration, and non-invasive brain stimulation. The discussion references the John Scalzi book about a near future where Earth wages war by offering the elderly new youthful bodies in exchange for military service.

They also discuss the ethical and legal concerns surrounding the creation of super soldiers, as well as the societal norms and potential risks. “What risks are we willing to take? There’s all these wonderful things we can do,” Matook said. “We don’t want a fair fight. We really don’t, this is not an honorable thing. We want our guys to be over-matching any possible enemies, right? So why aren’t we giving them pharmaceutical enhancements? Why are we making them run all week when we could just be giving them steroids? There’s all these other things you could do if you change societal norms and ethics. And laws, in some cases.”

The discussion concludes with considerations about the long-term effects, reversibility of enhancements, and the potential ownership of enhanced individuals by the government. “So if you do these kinds of changes to an individual, what do you do when their service is up? What happens? Or are they just literally owned by the government for life,” asks Reinerman-Jones. Hudson replied with a grim joke: “Termination.”

Read more of this story at Slashdot.

Source: Slashdot – Pentagon Scientists Discuss Cybernetic ‘Super Soldiers’

On Monday, Merriam-Webster announced its word of the year is “authentic — the term for something we’re thinking about, writing about, aspiring to, and judging more than ever.” The Associated Press reports: Authentic cuisine. Authentic voice. Authentic self. Authenticity as artifice. Lookups for the word are routinely heavy on the dictionary company’s site but were boosted to new heights throughout the year, editor at large Peter Sokolowski told The Associated Press in an exclusive interview. “We see in 2023 a kind of crisis of authenticity,” he said ahead of Monday’s announcement of this year’s word. “What we realize is that when we question authenticity, we value it even more.”

Sokolowski and his team don’t delve into the reasons people head for dictionaries and websites in search of specific words. Rather, they chase the data on lookup spikes and world events that correlate. This time around, there was no particularly huge boost at any given time but a constancy to the increased interest in “authentic.” […] “Can we trust whether a student wrote this paper? Can we trust whether a politician made this statement? We don’t always trust what we see anymore,” Sokolowski said. “We sometimes don’t believe our own eyes or our own ears. We are now recognizing that authenticity is a performance itself.”

There’s “not false or imitation: real, actual,” as in an authentic cockney accent. There’s “true to one’s own personality, spirit or character.” There’s “worthy of acceptance or belief as conforming to or based on fact.” There’s “made or done the same way as an original.” And, perhaps the most telling, there’s “conforming to an original so as to reproduce essential features.”

Read more of this story at Slashdot.

Source: Slashdot – Merriam-Webster’s Word For 2023 Is ‘Authentic’

The first wireless charging public roadway in the United States has been installed in Detroit’s Corktown neighborhood. Electrek reports: Wireless charging provider Electreon provided inductive-charging copper coils that were installed below the road’s surface. The coils will charge EVs equipped with Electreon receivers as they drive over the road. The road’s charging segments transfer electricity wirelessly through a magnetic field, which is then transferred as energy to the vehicle’s battery, charging it. Detroit’s wireless charging roadway is a pilot that will test and aim to perfect the wireless charging technology in a real-world environment. Researchers are using a Ford E-Transit equipped with an Electreon receiver.

The plan is to open it up to the public in the next few years. MDOT and Electreon have entered a five-year commitment to develop and pilot the electric road system on Michigan roads. The pilot is on a quarter-mile stretch on 14th Street between Marantette and Dalzelle Streets in Detroit’s historic Corktown. It runs alongside the Newlab at Michigan Central Building, home to more than 60 tech and mobility startups, where the wireless charging tech will be further tested and developed beginning in early 2024. In 2024, MDOT will begin seeking bids to rebuild part of Michigan Avenue (US-12) and will install additional inductive charging.

Read more of this story at Slashdot.

Source: Slashdot – Michigan Installs First Wireless EV Charging Road In US