“Last Beatles song”, restored by AI, is nominated for 2 Grammys

AI has brought the Beatles back together for one last time. In 1970 John Lennon recorded a song called “Now and Then” on grainy tape at his home in New York. The tape sat on a shelf until Lennon’s wife Yoko Ono gave it to the remaining members of the band after John’s death in 1980, for possible use in a posthumous album. However, the sound quality of the tape was so bad that there was no way to use it at the time. Then, last year, Paul McCartney joined an effort to revive the song from the battered tape using modern AI audio restoration tools. The result was so good that McCartney and Ringo Starr, the last surviving members of the band, added new instrumentals and vocals, and employed AI once again to add in guitar tracks by George Harrison from 1995. “Now and Then” was released in November of last year, and is considered the last Beatles song. It has now received 2 Grammy nominations, Record of the Year and Best Rock Performance, 54 years after it was first recorded on grainy tape in John Lennon’s home studio.

AI-restored Beatles song nominated for 2 Grammies

Judge rules in favor of OpenAI in landmark copyright case

OpenAI is the target of multiple lawsuits from media companies alleging copyright violation because the company trained its ChatGPT AI chatbot on data from a wide swath of the internet, including media sites. OpenAI has maintained that such training on publicly available material constitutes legally protected “fair use.” Last week a federal judge in New York essentially agreed with OpenAI, dismissing copyright violation complaints from two plaintiffs. This could be a precedent-setting decision that could make the courts much more friendly to AI companies using copyrighted material in their training data. The judge did give the plaintiffs one more chance to show damages from OpenAI’s actions, but expressed doubt that any such damages could be found.

Judge rules for OpenAI in landmark copyright case

Wendy’s bets on defense-tech giant Palantir’s AI to track your $1 Frosty

Burger chain Wendy’s has made a surprising foray into AI-enabled supply chain logistics by tapping Palantir, a defense tech giant who supplies the Pentagon with battlefield management systems. Palantir’s AI will help Wendy’s avoid shortages of its popular items, such as their $1 Frosty “frozen dairy treat.” Is battling for market share really just like an actual battle? Wendy’s intends to find out.

Wendy’s bets on AI from defense giant Palantir to track your $1 Frosty

The other election night winner was AI searchbot Perplexity

All of the major AI chatbots - ChatGPT, Claude, Google’s Gemini, and Elon Musk’s Grok - were flooded with questions about the presidential race on election night, and most survived the night without a major misstep. (The exception was Grok, which started giving out pro-Trump erroneous results and was shut down.) ChatGPT exercised perhaps too much restraint by refusing to answer many innocuous questions about the current results on the grounds of not wishing to interfere in the election. But the clear winner of the AI election news-vending race was upstart AI search bot, Perplexity. Perplexity prepared explanatory charts, graphs, and news snippets in real time and on the fly, and by most accounts, it performed brilliantly. Perplexity’s election news hub received 4 million visitors that night, as compared to CBS News’ 14 million online visitors. Not bad for an upstart.

The other election night winner was AI searchbot Perplexity

Vatican unveils AI “digital twin” of St. Peter’s Basilica

The Vatican has entered into a partnership with Microsoft to develop an AI-enabled “digital twin” of St. Peter’s Basilica. A 3D model of the ancient church was developed based on thousands of scans of the structure taken by drones carrying cameras and lasers. The resulting model will be available for virtual tours during the Vatican’s Jubilee 2025, which begins on Christmas Eve 2024.

Vatican unveils AI “digital twin”of St. Peter’s Basilica

AI improved scientists’ productivity by 39% - and they hated it

A study of the introduction of AI to the R&D lab of a large US company found that innovation went up significantly, with 44% more novel materials found and 39% more patents filed. However, less-talented researchers saw almost no gain, while the productivity of the top scientists doubled. This is thought to be due to the fact that the AI took the lead in generating ideas for novel compounds, while the scientists decided which of the AI’s proposed candidates to pursue with verification experiments. Better scientists had better intuition, while less talented researchers would pursue blind alleys. Strikingly 82% of participants disliked working with the AI, because it performed the “fun part” of the research, generating novel ideas, while humans were relegated to picking the compounds to pursue and doing the grunt work of verification.

AI improved scientists’ productivity 39% - and they hated it

AI robot artist’s painting sells for $1.1 million

A humanoid robot named Ai-Da, built by a UK robotics company, uses AI to design and paint portraits. Recently, one of the robot’s creations, an impressionistic portrait of computer pioneer Alan Turing titled “AI God”, sold at auction at Sotheby’s for $1.1 million, ten times its estimated value before the auction.

AI robot artist painting sells for $1.1 million

Swarovski Optik’s AI binoculars help birders identify birds in the wild

Swarovski Optik has released new AI-enabled binoculars that use AI image recognition together with GPS to identify birds sighted in their natural habitat.

Swarovski Optik’s AI binoculars help birders identify birds in the wild

Robot learns surgical techniques from watching videos

Researchers at Johns Hopkins have developed an AI robotic surgery system that can learn how to perform surgical techniques by watching videos of experienced surgeons. Current robotic surgery systems such as industry-leading Da Vinci are primarily teleoperated by human surgeons using input devices such as joysticks. The Johns Hopkins AI-enabled system brings man-machine collaboration to a new and more integrated level.

Robot learns surgical techniques from watching videos

Detecting medication errors with AI monitoring of wearable cameras

University of Washington researchers have devised an AI system that monitors the feed from wearable cameras in the OR, in order to detect and prevent potential medication errors. ORs are busy places, with multiple personnel circulating, and the chances of error can be high. The always-on AI monitors the camera feed from all personnel in the OR that have access to medications, and will alert if it sees a potential error in the making, such as wrong medicine or wrong dose.

Detecting medication errors with AI monitoring wearable cameras

AI detects high blood pressure and diabetes with 5 seconds of video

Researchers from the University of Tokyo have developed an AI system that can view as little as 5 seconds of video of a patient’s face and hands, and detect high blood pressure nearly as well as an automated blood pressure cuff, and detect diabetes nearly as well as the standard Hemoglobin A1c blood test. Work is ongoing to turn this system into a commercial product.

AI detects high blood pressure and diabetes with 5 seconds of video

AI team of agents designs novel nanobodies to block COVID

Stanford scientists have developed an interdisciplinary research team of AI agents, each given an individual role from Principal Investigator to Immunologist to Computational Biologist. A human supervisor set the research agenda, which was to find novel and effective treatments for COVID. The team of AI agents communicated among one another both in all-hands lab meetings, and in individual meetings between pairs of agents. With minimal guidance from the human, the team of agents devised a workable research strategy, which was to focus on antibody fragments called nanobodies, which are simpler to synthesize in a lab than full antibodies. The AI agents then designed experiments, and proposed tests of effectiveness of the candidate molecules. When the proposed protein fragments were tested, several were found to have good efficacy against the currently-circulating variants of COVID, while retaining efficacy against older variants. The age of AI-directed research is now upon us.

AI team of agents designs novel nanobodies to block COVID