Controlling computers with your mind - and without brain surgery

Much has been made of brain implant devices that allow the user to control computers just by their thoughts, a great boon for quadriplegic individuals who are paralyzed from the neck down. Startups like Elon Musk’s Neuralink are pursuing this technology actively. The biggest problem is that brain implants require drilling a hole in the patient’s head to insert the chip, and then push the associated wires into the brain to the desired target areas. The process is currently icky, painful, and dangerous.

Now researchers have developed not one, but two different noninvasive alternatives to implants for a brain-computer interface (BCI).

First, UCLA researchers have developed a wearable cap with electrodes that touch the scalp and detect brain waves like a medical EEG (electro-encephalogram). AI is able to decipher a user’s brain wave patterns to control a computer or a robotic arm.

Even more impressively, an MIT spin-out named AlterEgo uses a small wearable sensor to detect tiny involuntary movements of the patient’s facial muscles when thinking about words, and AI then translates those patterns of movement to determine the word being thought of. A recent demo of this technology has gone viral, with participants speaking silently to one another, translating words from one language to another, and even writing computer code, all by thinking.

If either of these two technologies is perfected, then Elon Musk’s Neuralink brain implant company becomes irrelevant.

If the AlterEgo minimalist wearable sensor becomes commercially viable, then large numbers of on-the-go users will want one. It would be very helpful in environments where typing is difficult (such as when walking), or where speech is not possible either due to ambient noise or a need for privacy. Watch the demo of the AlterEgo technology here.

UCLA researchers develop a wearable cap to control computers with brain waves

MIT spinout company senses the word you are thinking

Tesla’s Board wants to make Elon the world’s first trillionaire

Elon Musk is already the world’s richest man. Now the Board of Directors of Tesla is developing a compensation plan that could make him 3 times richer - the world’s first trillionaire. To achieve trillionaire status under this plan, Elon will have to produce some highly aggressive results, resulting in Tesla’s market value increasing 8-fold, to over $8 trillion dollars. For comparison, this is approximately twice the value of Nvidia, currently the world’s most valuable public company at $4.25 trillion. And Tesla faces headwinds, including slumping auto sales due to brand damage from Elon’s ill-fated tenure as the head of DOGE, and stiff competition from Chinese EV maker BYD. It’s clear that the Board is desperate to keep Elon’s attention focused on Tesla, rather than on his multiple other ventures, including the AI company, the rocket company, and the brain implant company (which our article just above suggests may soon be toast.)

Tesla’s Board wants to make Elon the world’s first trillionaire

Microsoft to use Claude as well as ChatGPT in its CoPilot AI

Once the closest of partners, Microsoft and OpenAI are drifting ever further into becoming frenemies, if not worse. You’d think Microsoft’s $13 billion investment in OpenAI when it was just a struggling startup would buy a little loyalty, right? Microsoft Satya Nadella certainly did. But in Silicon Valley, egos are as high as an elephant’s eye, and OpenAI CEO Sam Altman is on a mission to make his company the winner in AI, not just be a sidekick to Microsoft.

Both sides have been playing footsie with other players, but in the last 2 weeks Microsoft has made two moves that signal that the Microsoft-OpenAI situationship has moved into “it’s complicated” territory. Last week, Microsoft announced 2 new, powerful, homegrown AI models, loosening its dependence on OpenAI’s models. This week they revealed that they have a struck a deal with OpenAI’s arch-rival Anthropic. Microsoft will be able to power its CoPilot AI features in its software with ChatGPT, or Anthropic’s Claude, or both.

Nadella wanted this deal with Anthropic so much that he agreed to pay Amazon, Anthropic’s exclusive cloud services provider, for the cloud costs of using Anthropic. This is a stunning concession, given that Amazon’s AWS cloud services platform is the major competitor to Microsoft’s own Azure cloud services.

OpenAI backs major animated film to be made with AI

OpenAI is backing a major new animated film, Critterz, that will be created primarily with its AI tools. They hope to prove to Hollywood that these tools are ready for prime time and will usher in a new era of film making. To date, Hollywood has been distinctly skittish about using AI. The 118-day SAG-AFTRA strike in 2023, the longest strike in the union’s 90-year history, was primarily over issues of protecting actors and screenwriters from predatory uses of AI which threatened to decimate employment in the industry.

Now OpenAI is providing their AI technology for an animated film (no human actors are displaced), which will use human voice actors, as a showcase for the ways that AI can reduce film costs and speed production. Major animated films can cost up to $200 million to produce, and generally require 2 to 3 years to make. This film, Critterz, is projected to cost around $30 million and to require only 9 months to create.

If the cost and time savings turn out to be real, and the film is a critical and/or commercial success, no doubt more AI-enhanced films will follow.

OpenAI backs major animated film to be made with AI

Godfather of AI now says we CAN coexist with super-intelligent AI - if we program it with maternal instinct

Nobel Prize-winning AI scientist Geoffrey Hinton, known as the “Godfather of AI”, has been decidedly bleak about humanity’s chance of survival if we develop superintelligent AI models. He has often said that in the history of evolution on Earth, the more intelligent species always won out over lesser intelligences. (The facts that AI models are not species, are not sentient, and have no evolutionary drives for survival or domination are almost never mentioned in these types of pronouncements.)

Recently he has changed his tune, in an almost comically clueless Boomer way. He now points to the fact that there IS an example of a creature of lesser intelligence controlling one of higher intelligence - a crying baby controlling the behavior of its mother. So the answer is that we have to program maternal instinct into our super-intelligent AI models. (Facepalm.) Now there is so much wrong with this formulation, not least the casual misogyny and the outdated cultural context, that it would take more space than this brief newsletter can devote to demolishing it. Nonetheless, however bass-ackwardly Hinton arrived at his conclusion, I believe that he is essentially correct. AI models are our creations, and we have all the tools we need to program them as we see fit. The challenge for us humans is to have the wisdom to craft their motivations so that they act to help us, not hurt us.

The Godfather of AI may be a cranky old Boomer like your Grandpa, but genius Nobel Prize-winning Grandpas can sometimes say insightful things (in a way that makes you cringe.)

Godfather of AI says we need to give AI a “maternal instinct”

China mandates AI instruction in all grades from age 6 up

As part of its nationwide push to embed AI in all aspects of its economy, China has mandated that schools include instruction in all schools from age 6 through university. Implementation of this mandate is left to the provinces, so curricula will vary from place to place, at least to start. In the Chinese system, the central government is responsible for setting goals, and the provinces are allowed to innovate and adapt to local conditions. Progress is measured, and provinces whose programs do well are held up as models to the others (and the provincial governor just may get a promotion to a role in the central government.)

China mandates instruction in AI in all grades from age 6 up

Orchard AI is building the farm of the future

Agtech startup Orchard AI is developing a comprehensive automation platform for fruit farming. Autonomous robots (or your own tractor for the budget-conscious) carrying an array of sensors patrol the orchard, logging soil conditions, microclimate conditions, and the condition of each tree and even the condition of each fruit on each tree. This blizzard of near-real time data is analyzed by AI, which then gives suggestions for optimal management of the orchard, the individual tree, down to the individual fruit. Although being applied now to fruit crops, this sort of AI-assisted detailed farm management is a harbinger of the future for nearly all commercial crops.

Orchard AI is building the farm of the future

OpenAI announces AI Jobs Platform and AI Academy

OpenAI has made no secret of its ambition to be a massive consumer products company. Last week they announced 2 linked initiatives: an AI jobs bank, and an AI Academy which will grant certificates of mastery of AI tools and techniques. The AI Jobs Platform will have profiles of candidates who are AI-savvy, or who have particular AI skills. This will help companies hire more AI-proficient employees. The AI Academy will offer certificate courses in a wide variety of AI tools and techniques, helping companies upskill their existing workforce to prepare for the coming AI transformation. OpenAI is planning to certify 10 million Americans in AI by 2030. Walmart, the world’s largest private employer, has already committed to offering the AI Academy to its 1+ million associates. Jobs disruption and company transformations are already happening - OpenAI is offering ways for workers and companies to cope.

OpenAI announces AI Jobs Platform and AI Academy

Snakelike robot with tentacles may replace divers in risky environments

Researchers in the UK and Brazil have collaborated to develop a soft, flexible robot with tentacles to perform risky undersea operations too dangerous for human divers. This would be particularly useful for inspections and maintenance of offshore oil platforms and wind farms, as well as undersea pipelines and cable routes. Early tests of the robot find it both capable and rugged in harsh conditions.

Snakelike robot may replace divers in risky environments

Chinese robotics startup plans IPO at $7 billion valuation

Leading Chinese robotics startup Unitree Robotics is planning to file for an initial public offering (IPO) in the near future, targeting a share price that will make the company worth $7 billion. Founded with one employee in 2016, Unitree quickly rose to prominence with high-performing, inexpensive robot quadrupeds (commonly called “robot dogs”) which were useful for inspections and security patrols at factories and other businesses. More recently, Unitree has released popular and inexpensive humanoid robots, which dominated the recent Robot Olympics in Beijing. Unitree now has over 1,000 employees, annual revenue of $140 million, and is profitable. Even so, in any other sector except AI/Robotics, the offering price of 50 times revenue would be laughably high. Unitree’s founder, 35-year-old Wang Xingxing, plans on laughing all the way to the bank.

Chinese robotics startup plans IPO at $7 billion valuation

FDA wants AI to speed up drug approvals

The US Food and Drug Administration has always valued safety over speed. This caution famously helped protect the US from the epidemic of babies born with deformed, seal-like limbs in the 1960s, a tragic side effect of the drug thalidomide that was marketed in Europe to pregnant women as a cure for morning sickness. Now the FDA is looking to get safety and speed with various AI tools. The agency has rolled out Elsa, an internal AI tool that speeds up administrative tasks, including scientific reviews. The agency is also piloting use of “organs on a chip” technology, formally known as Microphysiological Systems, as a substitute for slow and expensive animal trials. Human cells from a target organ such as the liver are grown on a chip which keeps the cells alive with microfluidics - tiny channels that direct fluids to bring in nutrients and wash away wastes. Impact of drugs on the functions of these human cells can be rapidly assessed without breeding, feeding, dosing, and sacrificing any animals.

FDA wants AI to speed up drug approvals

AI voice agents help seniors report home blood pressure readings

A study by Emory Healthcare in Atlanta showed that commercially available AI voice agents helped older patients with high blood pressure to take and record their blood pressure at home. 2000 patients, most over age 65, were contacted at home by phone in their preferred language, and then coached through the process of taking a valid blood pressure with a home blood pressure monitor. This reading was recoded by the voice agent and entered into the patient’s medical chart, where it would be reviewed by a clinician. Evaluation of this outreach program showed that the proportion of patients with controlled blood pressures increased by 17%, enough to earn a 4-star rating on Medicare’s HEDIS quality measurement. Patient satisfaction with the calls exceeded 90%. This study points the way to using more home monitoring of patients to improve outcomes while enhancing patient satisfaction.

AI voice agents help seniors report home blood pressure readings