OpenAI announces it has begun training its new flagship AI model
Cade Metz | The New York Times
“The San Francisco startup, one of the world’s leading AI companies, said in a blog post that it aims to build ‘artificial general intelligence,’ or AGI, a machine that can do anything the human brain can do, and hopes the new model will bring ‘next-level capabilities.’ The new model will power AI products such as chatbots, digital assistants similar to Apple’s Siri, search engines and image generators.”
Will scaling solve robotics problems?
Nishant J. Kumar | IEEE Spectrum
“Develop general-purpose robots that can perform a wide variety of tasks efficiently and robustly.” Any Getting robots to navigate home and office environments in the same way as humans do may have been the Holy Grail of robotics since the field’s inception. And given recent advances in fundamental models, it seems that extending existing network architectures by training on very large datasets may indeed be the key to that Holy Grail.”
Genetically edited salad greens to hit U.S. store shelves this fall
Emily Marin | Wired
“Last year, startup Pairwise began selling the first food product in the U.S. made with CRISPR technology: a new type of flavor-tuned mustard greens. … The company has introduced the greens to select restaurants, cafeterias, hotels, nursing homes, catering and other foodservice industries in several cities. One grocery store in New York City also sold the greens. Now, biotech giant Bayer has licensed the greens from Pairwise and will distribute them to grocery stores nationwide.”
New ChatGPT offers a lesson in AI hype
Brian X. Chen | The New York Times
“When OpenAI launched the latest version of its hugely popular ChatGPT chatbot this month, it introduced a new voice with human-like inflections and emotions. An online demo even showed the bot teaching a child how to solve a geometry problem. Unfortunately, the demo was essentially a bait-and-switch. The new ChatGPT was released without most of its new features, including an improved voice (which the company told me was delayed for revisions). The ability to use your phone’s video camera to analyze things like math problems in real time is also still unavailable.”
The world’s thinnest lens is just three atoms thick
Michael Irving | New Atlas
“[A Fresnel lens uses] Lenses are concentric rings of material that diffract light to a single point, allowing them to be much thinner at the expense of some image sharpness. Now, scientists have pushed it almost to the limit, creating a lens just 0.6 nanometers (nm), or just three atoms thick. That’s the thinnest lens ever made, beating the previous record of 6.3 nm set in 2016 by a factor of 10.”
Why Google’s AI-generated summaries are wrong
Rhiannon Williams | MIT Technology Review
“On Thursday, Google Search head Liz Reid announced that the company was making technical improvements to its systems to make them harder to generate false answers, including improving mechanisms for detecting nonsensical queries. It’s also limiting satirical, humorous and user-generated content from being included in answers, because such content can lead to misleading advice. But why is AI Overview returning unreliable and potentially dangerous information, and is there anything that can be done to fix it?”
1-bit LLM could solve AI’s energy needs
Matthew Hutson | IEEE Spectrum
“To make LLM cheap, fast and green, it needs to be small enough to run directly on devices like phones. … Researchers have long compressed their networks by reducing their precision. [their] A process called parameter quantization ensures that instead of each parameter taking up 16 bits, it takes up 8 or 4 bits. Now researchers are pushing the limit all the way down to 1 bit.”
New computer uses AR glasses to create a 100-inch virtual workspace
Kyle Barr | Gizmodo
“The problem with the term ‘spatial computer’ is that most devices using this arcane marketing term don’t actually look like computers. Sure, the Apple Vision Pro and Meta Quest 3 meet the definition of a ‘computer,’ but most people still think of a ‘PC’ as a desktop or laptop. So here comes the Spacetop G1 AR laptop, ready to kick the arse out of both the desktop and VR markets because they’re both too set in their ways.”
World’s first tooth regenerative drug to be administered to humans in September
Bronwyn Thompson | New Atlas
“The world’s first human trials of a drug capable of regrowing teeth will begin within the next few months, less than a year after news of successful animal testing, which would make the drug commercially available as soon as 2030. … The intravenous treatment will test its effectiveness on the human dentition after successfully growing new teeth without significant side effects in ferret and mouse models.”
Humanity needs an ethical upgrade to keep up with new technology
Marcelo Glaser | BigThink
“Technological advances such as nuclear weapons, genetic engineering with CRISPR, and artificial intelligence raise significant ethical challenges and responsibilities. It is reasonable to be concerned about technological threats and dilemmas, but many people respond by blaming scientists and science itself, failing to recognize the difference between those who control the application of scientific discoveries.”
Image credit: NASA, ESA, Joel Kasner (RIT)
