Google DeepMind Used a Giant Language Mannequin to Resolve an Unsolvable Math Downside
Will Douglas Heaven | MIT Expertise Assessment
“In a paper printed in Nature [on Thursday], the researchers say it’s the first time a big language mannequin has been used to find an answer to a long-standing scientific puzzle—producing verifiable and worthwhile new info that didn’t beforehand exist. ‘It’s not within the coaching knowledge—it wasn’t even identified,’ says coauthor Pushmeet Kohli, vice chairman of analysis at Google DeepMind.”
Supercomputer That Simulates Whole Human Mind Will Swap on in 2024
James Woodford | New Scientist
“Not like an strange laptop, its {hardware} chips are designed to implement spiking neural networks, which mannequin the best way synapses course of info within the mind. Such neuromorphic computer systems, as they’re identified, have been constructed earlier than, however DeepSouth would be the largest but, able to 228 trillion synaptic operations per second, which is on par with the estimated variety of synaptic operations in a human mind.”
In a World First, a Affected person’s Antibody Cells Had been Simply Genetically Engineered
Emily Mullin | Wired
“[B cells] make a lot of antibodies—hundreds of them each second. What if these antibody factories may very well be harnessed to make different issues the physique wants? That’s the thought behind a trial launched by Seattle-based biotech firm Immusoft. The corporate introduced right this moment that its scientists have genetically programmed a affected person’s B cells and put them again in his physique in an effort to deal with illness. It’s the primary time engineered B cells have been examined in an individual.”
TECH
Everyone’s Speaking About Mistral, an Upstart French Challenger to OpenAI
Benj Edwards | Ars Technica
“Mistral, primarily based in Paris and based by Arthur Mensch, Guillaume Lample, and Timothée Lacroix, has seen a speedy rise within the AI area not too long ago. It has been rapidly elevating enterprise capital to turn into a form of French anti-OpenAI, championing smaller fashions with eye-catching efficiency. Most notably, some (however not all) of Mistral’s fashions run domestically with open weights that may be downloaded and used with fewer restrictions than closed AI fashions from OpenAI, Anthropic, or Google.”
OpenAI Demos a Management Methodology for Superintelligent AI
Eliza Strickland | IEEE Spectrum
“Earlier this yr, OpenAI launched its superalignment program, an bold try to search out technical means to manage a superintelligent AI system, or ‘align’ it with human objectives. …The most important problem for this challenge: ‘This can be a future downside about future fashions that we don’t even know how you can design, and positively don’t have entry to,’ says Collin Burns, a member of OpenAI’s superalignment crew. ‘This makes it very difficult to review—however I believe we additionally haven’t any alternative.’ The primary preprint paper to come back out from the superalignment crew showcases a technique the researchers tried to get round that constraint.”
Agility Is Utilizing Giant Language Fashions to Talk With Its Humanoid Robots
Brian Heater | TechCrunch
“It’s turn into more and more clear that these kinds of applied sciences are primed to revolutionize the best way robots talk, study, look, and are programmed. Accordingly, various high universities, analysis labs, and firms are exploring the perfect strategies for leveraging these synthetic intelligence platforms. Nicely-funded Oregon-based startup Agility has been taking part in round with the tech for some time now utilizing its bipedal robotic, Digit.”
This New System Can Educate a Robotic a Easy Family Process Inside 20 Minutes
Rhiannon Williams | MIT Expertise Assessment
“A brand new system that teaches robots a home process in round 20 minutes may assist the sector of robotics overcome one among its greatest challenges: an absence of coaching knowledge. The open-source system, known as Dobb-E, was skilled utilizing knowledge collected from actual houses. It might probably assist to show a robotic how you can open an air fryer, shut a door, or straighten a cushion, amongst different duties.”
Dishonest Fears Over Chatbots Had been Overblown, New Analysis Suggests
Natasha Singer | The New York Occasions
“Final December, as highschool and school college students started making an attempt out a brand new AI chatbot known as ChatGPT to fabricate writing assignments, fears of mass dishonest unfold throughout the US. …However the alarm could have been overblown—not less than in excessive faculties. In response to new analysis from Stanford College, the popularization of AI chatbots has not boosted total dishonest charges in faculties.”
DIGITAL MEDIA
Information Publishers See Google’s AI Search Instrument as a Site visitors-Destroying Nightmare
Keach Hagey, Miles Kruppa, and Alexandra Bruell | The Wall Road Journal
“Shortly after the launch of ChatGPT, the Atlantic drew up an inventory of the best threats to the 166-year-old publication from generative synthetic intelligence. On the high: Google’s embrace of the know-how. About 40% of the journal’s internet visitors comes from Google searches, which flip up hyperlinks that customers click on on. …What was as soon as a hypothetical risk is now a really actual one. Since Might, Google has been testing an AI product dubbed ‘Search Generative Expertise’ on a bunch of roughly 10 million customers, and has been vocal about its intention to deliver it into the guts of its core search engine.”
Picture Credit score: Greg Rakozy / Unsplash