After weeks of testimony, in which the dirty laundry of several of the most prominent figures in AI was thoroughly aired, the jury took less than 2 hours to decide that Elon was a loser. Characteristically, he was a sore loser, and vowed to appeal. Figures.

The stakes could not have been higher. Elon wanted OpenAI, the company that set off the current AI frenzy by releasing ChatGPT in November of 2022, a company that has over 1 billion weekly users, a company that is now generating over $25 billion in annual revenue, a company planning to go public later this year at a $1 trillion valuation - Elon wanted that company to be forced back into a not-for-profit structure that would have crippled its ability to raise more capital in one of the most capital-intensive industries on the planet. Oh, and Sam Altman should be fired as CEO.

The trial itself was a rare glimpse at the inner workings of a titanically successful startup that initiated a whole new industry - and the look was not pretty. Almost everyone involved came off looking petty and self-interested while spouting idealistic protestations about wanting to serve the good of humanity.

Lowlights of the trial (there were no highlights):

Elon’s lawsuit claimed that Altman had improperly converted OpenAI from a not-for-profit structure and unjustly enriched himself. Altman pointed out that the transition had been blessed by the Attorneys General of both California and Delaware, as well as the board of OpenAI itself. It was a highly public effort, with millions of dollars in lawyer’s fees being paid to top firms to make sure that the structure was as bulletproof as possible.

The jury said that none of that mattered. Under California law, there is only a three year window of time for anyone to object to a breach of an agreement, and Elon had known of OpenAI’s intent to have a for-profit arm since 2018. Elon was out of time, and his case was thrown out.

The way is cleared for OpenAI to have its trillion-dollar IPO, slated for later this year or early next. Elon will have to console himself with his own upcoming IPO of SpaceX, tentatively priced at $1.75 trillion.

Such are the lifestyles of the rich and (in)famous.

Elon becomes AI’s biggest loser, his case thrown out on a technicality

Anduril, a defense-tech AI startup that makes smart drones for the military, has teamed up with Meta/Facebook to make smart glasses for soldiers on the battlefield. These smart glasses will allow the soldier to handle battlefield tasks like displaying drone feeds, highlighting targeting data, and allowing soldiers to cue artillery or drone strikes via voice and eye-tracking.

Palmer Luckey, the 33-year-old founder of Anduril, is a familiar figure online, wearing his signature Hawaiian shirts and his bedraggled goatee. Luckey invented the modern virtual reality glasses in his parents’ garage when he was 19. He sold his invention to Meta/Facebook two years later for $2 billion. He was then fired from Meta in 2024 over his controversial financial backing of Donald Trump.

Apparently harboring no hard feelings, Luckey invited Meta into Anduril’s contract with the Pentagon to develop smart glasses for warfighters. Why does the man who invented VR glasses need Meta? Because Meta has invested billions of dollars into extending Luckey’s original invention into a popular consumer technology. This means that Meta has the hardware expertise to produce lightweight, high performance augmented reality glasses at scale, and the global supply chain to back it up.

This allows Anduril, which is actually the lead company on the Pentagon contract, to focus on developing device integration into its proprietary Lattice command and control software platform, which is currently being used for management of drones.

If the two companies can deliver on their promises, the Pentagon is prepared to sign a $22 billion multi-year contract for the new warfighter smart glasses.

That’s 22 billion reasons for Meta and Luckey to kiss and make up.

Meta and Palmer Luckey team up to pursue $22 billion smart glasses contract

China’s first offshore wind-powered underwater data center achieved full commercial operation last week, after 2 months of testing.

Data centers consume enormous amounts of electricity, essentially all of which ends up as waste heat. Cooling the data centers with today’s technology can consume a third or more of all the electricity used. China’s design puts the data center 100 feet under the sea, where the surrounding ocean serves as a heat sink. This slashes cooling needs by two-thirds or more.

This first underwater data center is placed by an offshore wind farm, allowing it to be powered by renewable generation sources.

A major hurdle for the undersea data center is maintenance and repair of the servers inside. This project is using sealed, modular, high-reliability computing units that can be easily swapped out. If this first project succeeds, it could point the way to one more way that data centers could be proliferated, with less impact on the environment and local populations than land-based installations.

/

China puts data centers under the sea

The internet went nuts in 2024 when Google upgraded NotebookLM, its AI-powered personalized research assistant, to give it the ability to create realistic audio podcasts that explained any document that was uploaded. The capacity to hear a podcast-like back-and-forth chitchat between two realistic AI-generated voices about any uploaded topic seemed magical.

Now Amazon has added podcasting to the abilities of its AI-enabled Alexa+ smart speaker. Users don’t have to upload anything. Amazon has partnered with a wide range of major news and information publishers, including over 200 local newspapers from across the US. So the user can just ask Alexa+ about a topic of interest, and the smart speaker will generate a personalized podcast episode covering that information. Users have the ability to direct Alexa+ regarding the depth and length of the podcast, as well as other customization features.

Amazon is asking the question: in the age of AI, why are you listening to news and information programs that other people have designed, when you can get personalized programming just by asking?

Amazon’s Alexa+ creates personalized podcasts on demand

ChatGPT is trying to colonize almost every sector of your life. It recently added an optional Health tab to its home screen, which allows you to securely upload information from your fitness watch, or even from your electronic medical record. ChatGPT can then explain the information to you in understandable terms, and help you analyze trends and draw conclusions.

OpenAI is now rolling out a similar playbook for your financial information. They are rolling out a Finance tab for paying users, which will use the Plaid app to securely connect to all of your financial institutions. Plaid will not reveal your account numbers, and ChatGPT will have no ability to initiate transactions. ChatGPT will synthesize the data from the separate institutions and build a central dashboard that shows spending habits, recurring subscriptions, upcoming bills, portfolio performance, etc.

With all of your information aggregated, you can ask ChatGPT for suggestions for achieving financial goals, such as paying off credit card debt, saving for a house, or planning for retirement. OpenAI has announced that there will be a future integration with Intuit (parent company of TurboTax and QuickBooks) , which will allow things like estimating tax liabilities on sales of stocks or real estate.

ChatGPT adds a Finance tab, partnering with Plaid

Japan has a bear problem. Attacks by bears on humans have skyrocketed in recent years, reaching 238 recorded attacks in the past year, with 13 humans killed. Nonfatal attacks have included a bear that terrorized vacationers at a swimming pool, and another bear that marauded through a supermarket. The response by some Japanese is characteristically high-tech: robots. Specifically, an animatronic robot Monster Wolf that repels bears by scaring the bejeesus out of them. The robot sports a simulated wolf’s head, which swivels menacingly back and forth while emitting high-decibel wolf snarls and howls. LED lights blink on and off.

Apparently it works. The small, niche firm that manufactures the robot wolf is swamped with orders.

The roots of the current bear crisis in Japan go back decades. In the 1970’s, Asian black bears in Japan were an endangered species. Conservation measures in the 1980’s helped protect them in their natural habitat, and their numbers rebounded sharply. However, over the past decade, warm winters due to climate change have disrupted the seasonal cycles of nut-bearing trees in Japan, a major food source for bears in the autumn. Simultaneously, the warm winters have caused many of the bears to shorten their hibernation, or even skip it altogether. The result is a flood of hungry bears that come out of their usual habitat into urban areas, searching for food.

Unfortunately, scaring hungry bears away from one location only causes them to move to another. The Japanese government will need to devise a solution that addresses the root causes of the bear invasion.

Japan’s bear problem causes a run on robot wolves

The AI division of Japanese electronics giant Sony (who knew they had one?) has recently published a paper in the prestigious scientific journal Nature, about their world-class ping-pong playing robot. I am not making this up.

We have gotten used to AI being able to write better than we can, play videogames better than we can, even generate computer code better than we can. But the physical world has been a challenge for AI. In fact, conventional wisdom has swung to advising young people to become plumbers and electricians, because so much of their work is physical, and therefore supposedly safe from AI replacement.

Conventional wisdom may have to update its assumptions. Sony AI has developed a robot, called ACE, that can outplay elite amateur human table tennis players, and compete with professional table tennis players. Table tennis is a challenge for robotics because it requires fast, precise actions near obstacles and at the edge of human reaction time. Sony AI hooked high-speed tracking cameras to a robotic arm with a novel control system based on Reinforcement Learning (learning through playing against itself.)

Paired against five different elite amateur players, ACE won 3 of the 5 matches. Pitted against two professional table tennis players, the robot won only one of the seven games played, but was competitive in all seven games. Professional players won points on shots with extremely high spin, beyond the capacity of the current robot to defend against.

The performance of ACE calls to mind earlier efforts to match human skill at a particular game. The time between the first game won by a computer against a chess Grand Master (Bent Larsen, 1988) and the time when a computer chess program defeated a reigning chess World Champion (Garry Kasparov, 1997) was nine years. It seems unlikely that it will take that long for a robot to become the best table tennis player in the world.

And perhaps only a little longer for it to become a plumber or electrician.

Robot ping-pong player bests elite humans

Much of the ballyhoo about AI in medical care has been focused on the rapid adoption of AI Scribes, systems that convert the spoken interactions between a physician and a patient into a transcript, which is then summarized by AI, for editing and review by the physician, before being entered into the patient’s medical record. Such systems save physicians an hour or more per day, relieve administrative burdens, and reduce physician burnout.

Now a new crop of AI transcriptionists are popping up, aimed at giving the patient a transcript of their visit with the doctor, which AI can explain and summarize, and which can be shared with family. This helps avoid patients misunderstanding or forgetting important information and advice from their doctor. It also serves as a patient empowerment tool, since patients can question the AI about the information that they were given, and the AI can connect them to other sources of valid medical information.

Los Angeles-based Kin Health was founded last year by two physicians (and brothers) Arpan and Amit Parikh, along with lead engineer Kyle Alwyn. Alwyn previously co-founded HeyDoctor, an app that was acquired by GoodRx, the digital prescription drug discount company. Two co-founders of GoodRx are also on Kin’s Board.

The startup is positioning its app as a service built by doctors for patients. They received $9 million in funding this month in a round led by Maveron, a venture capital firm that focuses on early stage, consumer-facing companies.

Kin Health raises $9 million to develop an AI note taker for patients.