Once in a while, LLMs turn evil--and no one quite knows why. The AI company Anthropic has made a rigorous effort to build a large language model with positive human values. The $183 billion company's flagship product is Claude, and much of the time, its engineers say, Claude is a model citizen. Its standard persona is warm and earnest. When users tell Claude to "answer like I'm a fourth grader" or "you have a PhD in archeology," it gamely plays along. It makes threats and then carries them out. And the frustrating part--true of all LLMs--is that no one knows exactly why. Consider a recent stress test that Anthropic's safety engineers ran on Claude. In their fictional scenario, the model was to take on the role of Alex, an AI belonging to the Summit Bridge corporation.

ChatGPT debuted in Nov. 2022, garnering worldwide attention almost instantaneously. The artificial intelligence (AI) is capable of answering questions on anything from historical facts to generating computer code, and has dazzled the world, sparking a wave of AI investment. Now users have found a way to tap into its dark side, using coercive methods to force the AI to violate its own rules and provide users the content -- whatever content -- they want. ChatGPT creator OpenAI instituted an evolving set of safeguards, limiting ChatGPT's ability to create violent content, encourage illegal activity, or access up-to-date information. But a new "jailbreak" trick allows users to skirt those rules by creating a ChatGPT alter ego named DAN that can answer some of those queries.

While it's possible that someone out there enjoys folding clothes, it's probably not a beloved pastime. Accordingly, researchers at UC Berkeley's AUTOLAB have developed a new robotic method of folding garments at record speed (for a robot) called SpeedFolding. Using machine vision, a neural network called BiManual Manipulation Network (BiMaMa-Net), and a pair of industrial robot arms, SpeedFolding can fold 30–40 randomly positioned garments per hour, usually finishing each within two minutes. While that rate does not sound impressive compared to a human, previous robotic garment-folding methods reached only "3-6 FPH" (that's "folds per hour") according to the researchers in a paper submitted for presentation at IROS2022 next week in Kyoto. Speed achievement aside, the paper is worth a read to enjoy how the researchers describe the garment-folding problem in technical terms.

Word has been making its way out from the technology community: The world changed this summer with the rollout of an artificial intelligence system known as GPT-3. Its ability to interact in English and generate coherent writing have been startling hardened experts, who speak of "GPT-3 shock." Where typical AI systems are trained for specific tasks--classifying images, playing Go--GPT-3 can handle tasks it was never specifically trained for. Research released by its maker, San Francisco-based OpenAI, has found that GPT-3 can...

Artificial intelligence has long seemed like more sizzle than steak, but 2017 may have been the year when enterprises decided to try it in earnest. According to a new report, AI's use as a mainstream business tool grew 60 percent over the last year. According to the Outlook on Artificial Intelligence in the Enterprise 2018, 61 percent of business and technology executives surveyed were already using AI to save money, boost productivity, personalize interactions and automate repetitive manual tasks. Predictive analytics, machine learning, natural language processing and voice recognition and response were the most widely used AI solutions reported. The survey report from Narrative Science, which helps organizations integrate intelligence into their platforms, said respondents using AI-driven solutions described benefits such as saving time and increasing operational efficiency, accuracy and employee productivity.

Meet Promobot IR77, an artificial intelligence (AI) designed to have face-to-face interactions with humans. It looks cute, but this Russian-made robot had recently been "arrested," after making rounds in a political rally, recording voters' opinions about a candidate's team. It sounds fairly harmless, but Promobot seemed to have made enough trouble to make the local authorities ask policemen to apprehend and detain the robot. "Police asked to remove the robot away from the crowded area, and even tried to handcuff him," the Promobot spokesperson said. This isn't the first time that Promobot got itself in a fair amount of mischief--it's run away from its home laboratory before, twice. The mad run for freedom ended with a battery-drained robot blocking thickening traffic in the street; the programmers were left scratching their heads.

In this week's Brain Boost: monetizing chatbots, AI breaks the sound barrier, and Björk's VR. Monetizing chatbots Artificially intelligent assistants are all the rage. Facebook has an AI messenger bot that makes recommendations with you via chat. Amazon has Alexa, which can handle simple conversations and enables you to make purchases– all without turning on any device. And, of course, Apple's Siri performs a huge range of functions and will reportedly be a big part of next generation Macintoshes.