Nobel Prize awarded to two AI pioneers

Much to the surprise of almost everyone, including the recipients, (“I was flabbergasted,” said Hinton), this year’s Nobel Prize in Physics has been awarded to two pioneers in the theory of artificial intelligence: John Hopfield of Princeton and Geoffrey Hinton who did his seminal work at UC San Diego. Both researchers have been taken aback by how powerful their brainchild has become. Hopfield declared himself “very unnerved” by how little we understand how AI produces its results, and Hinton, known as “the Godfather of AI” is also sometimes called the “figurehead of doomerism” for his very public warnings about the dangers of an out-of-control AI - misgivings which caused him to quit a cushy and lucrative job at Google to go back to teaching at the University of Toronto. Love it or fear it, it’s clear that AI has world-changing potential, and the Swedish Royal Academy of Sciences reached way out of its usual bailiwick to award a Nobel - in Physics, no less, generally regarded as the most prestigious of the Nobels - in order to recognize that transformational power.

Hinton and Hopfield awarded Nobel Prize for research in AI

OpenAI raises $6.6 billion, giving it a $157 billion valuation

Last week OpenAI CEO Sam Altman made history by pulling together the largest venture capital raise in history, $6.6 billion from some of the deepest pockets in tech investing, such as Softbank, Microsoft, and Nvidia. The resulting $157 billion post-money valuation makes OpenAI not only one of the most valuable startups on the planet, it is now one of the most valuable private companies in the world, with a valuation comparable to publicly traded companies such as AT&T, Shell Oil Company, and Wells Fargo. By ordinary metrics this seems an insane valuation for a company with revenue of “only” $3.7 billion and a loss of $5 billion for the last 12 months. If OpenAI delivers on Altman’s rosy projections, no price is too high. And if not, the savvy investors have a very attractive Plan B - just make sure that Sam can keep the proverbial plates spinning in the air long enough for OpenAI to have an IPO, and the smart money can cash out to the dumber money of retail investors eager to grab a stake in one of the most-hyped technologies in history.

OpenAI raises $6.6 billion, giving it a $157 billion valuation

Meta releases text-to-video app to compete with OpenAI’s Sora

Text-to-video apps have been The Next Big Thing now for over 6 months, a geological eon in AI time. OpenAI released demo clips of Sora, its groundbreaking entry in the field, last February (remember the clips of the mastodons, and the puppies in the snow?). Since then, it has been quite coy about a release date while it is “testing for safety.” In the interim, a slew of competitors have emerged. Now Meta/Facebook has thrown another hat in the ring, with Movie Gen, which purports to include all the features users want, including editing the resulting videos by text, generating videos based on images or other videos, and more. The capacity for Movie Gen to turbocharge production of deepfakes must be making Meta’s lawyers blanch. The link below gives you a peek at what they have created. You can look, but not touch - Meta is also “testing for safety.”

Meta’s Movie Gen adds to a crowded field of text-to-video wannabes

Waymo partners with Hyundai to assemble driverless taxicabs

Waymo is on a roll (pun intended.) The driverless taxi company owned by Google parent Alphabet has racked up 21 million miles of autonomous taxi rides, recently released data proving that its robo-taxis have one-fourth the accidents per mile of humans, is expanding to Austin and Atlanta in partnership with Uber, and now has announced a multi-year partnership with carmaker Hyundai. Hyundai will install Waymo’s autonomous driving tech into units of its IONIQ 5 SUV electric vehicles destined to be added to Waymo’s fleet of driverless taxis.

Waymo partners with Hyundai to assemble driverless taxicabs

Harvard students convert Meta AR RayBans into stalker specs

Just as a demonstration of what off-the-shelf AI technology is capable of, two Harvard students used Meta’s best-selling RayBan augmented reality glasses to perform facial recognition on fellow students and passersby, then used publicly available databases to find out the individual’s name, home address, job, and much more. They were even able to convince strangers waiting for a subway car that they had some personal connection, based on the personal data they could work into the conversation. Thankfully, the students did not release this software into the world, but are using their demo as a cautionary tale that highlights how vulnerable we all are to having our personal data misused. Regulators and AI vendors, take note.

Harvard students convert Meta AR RayBans into stalker specs

Environmental fund launches AI-powered satellite to track methane

Environmental Defense Fund, an eco-advocacy NGO, has launched a satellite that will use AI to process images in order to detect and track sources of methane emissions into the atmosphere. Methane is 81 times more potent as a greenhouse gas than carbon dioxide, but to date it has been difficult to measure emissions accurately. EDF’s MethaneSat will use sophisticated AI algorithms to aggregate the impact of multiple small, dispersed sources of methane in order to get a clearer picture of the size and scope of the problem, and to help identify where efforts at mitigation may be most effective.

EDF launches AI-powered satellite to track methane emissions

US Army testing robot dogs carrying anti-drone AI rifles

The US military is going all in on AI-powered weapons. Now they are testing robot dogs, already used for autonomous transport of ammunition and supplies to troops in the field, as anti-drone marksmen with AI-powered rifles in the Middle East. Currently, the US Army often uses expensive missiles to counter attacks from cheap drones. Since bullets are much cheaper than missiles, or even drones, it is practical to deploy a durable multi-use robot dog carrying a rifle with an AI targeting system that can detect, track, and destroy drones.

US Army testing robot dogs carrying anti-drone rifles

Home-made AI software finds people when Search and Rescue can’t

British Mountain Rescue is an all-volunteer force of individuals who search for missing hikers in rugged mountain terrain in the UK. Now a volunteer in the organization has developed software that automatically pilots drones with cameras in a search pattern within a designated area, and uses AI to detect anomalies in the images that might indicate a downed hiker. A human examines the tagged images, and the coordinates of those that seem promising are communicated to the search and rescue teams to focus their efforts. The system has already proven its worth in finding hikers not found by conventional means. More evidence that AI is empowering individuals to solve problems not addressed by the market.

Home-made AI software finds people when Search and Rescue teams can’t

Four National Cancer Centers launch AI Alliance

Four national Cancer Institute-designated cancer centers - Dana Farber, Fred Hutch, Memorial Sloan Kettering, and the Sidney Kimmel Center - have joined together to create the Cancer AI Alliance (CAIA). The aim is to deploy AI for research on the centers’ data, in order to unlock insights that can lead to innovative treatments and improved outcomes. The initial $40 million in funding for the CAIA comes from Amazon’s AWS cloud service, Deloitte, Microsoft, Nvidia, and Slalom Consulting.

Four National Cancer Centers launch Cancer AI Alliance

AI at Penn finds diseases in scans taken for other reasons

Millions of imaging scans are performed on patients each year, and until now, much of the information captured has been wasted. Radiologists understandably have concentrated on reading scans to answer the question that the ordering physician is asking, while giving less attention to subtle indications of possible disease in other organs. Now researchers at the University of Pennsylvania have developed an AI image analysis system called AInSights, which scans multiple organs for evidence of disease, regardless of the original purpose of the scan. AInSights is currently being used on 2,000 abdominal and chest CTs per month. The process is integrated into the radiologists’ workflow, and requires only 2.8 minutes per scan, thus not impeding the busy radiologist from completing their work. The cost of the system for this volume is estimated at $700 monthly.

AI at Penn finds diseases in scans taken for other reasons

VA integrating AI in clinical care and in administration

The Veterans Administration health system is adopting AI both to improve clinical practice and to achieve administrative efficiencies. AI is being deployed to identify veterans at risk for suicide or homelessness, to help detect pulmonary nodules on chest imaging, and to reduce waiting times for appointments.

VA integrating AI in clinical care and in administration

AI identifies candidate molecules for new pain drugs

G-protein coupled receptors (GPCRs) are cell-surface receptors that are involved in a wide variety of cell signaling pathways. Now researchers at the Cleveland Clinic and at the IBM Watson Research Center have developed a deep learning AI system that can predict interactions between a given molecule and GPCRs. The researchers validated the system by searching for FDA-approved drugs and gut bacteria metabolites that can be repurposed to treat pain.

AI identifies candidate molecules for new pain drugs