OpenAI snags Jony Ive startup for $6.5 billion

It’s been a whirlwind week of global dealmaking for OpenAI CEO Sam Altman. First he announced a huge partnership with the United Arab Emirates to build a multibillion-dollar data center in that country and offer ChatGPT to all UAE residents. Then he took a few minutes to acquire an AI hardware startup founded by the legendary Jony Ive, who designed all the iconic products at Apple during the Steve Jobs era, for an eye-watering price of $6.5 billion. As with all of Altman’s machinations, each deal had a long backstory and wheels-within-wheels complexity of overlapping (and sometimes conflicting) interests. An analysis of the UAE deal is in the section below. The Jony Ives deal accomplishes several goals for OpenAI, and for Altman personally. Altman had been collaborating with Ive for 2 years on developing AI-powered consumer devices, and Open AI had already acquired a 23% stake in Ive’s AI startup, io. Now, Ive and his team will be joining OpenAI and heading up the design team. Since the deal is an all-stock transaction, OpenAI does not have to burn any of its precious cash, which is covering its estimated $14 billion dollar losses for this year (on 12.7 billion of revenue.) In addition, the issuance of $6.5 billion worth of new stock to purchase io dilutes the share of the company owned by the nonprofit arm from an absolute majority (51%), and therefore effective control, to 42%, shifting control to the for-profit investors. Who wins the most in all this? Our boy Sam, to absolutely no one’s surprise.

OpenAI snags Jony Ive startup for $6.5 billion

IBM lays off 8,000 due to AI, but also hires over 9,000 new employees

IBM has recently announced that they are laying off 8,000 employees due to efficiencies created by AI. A year ago, IBM implemented a proprietary AI chatbot known as AskHR, which automates certain HR administrative tasks such as processing vacation requests, managing payroll, and handling employee documentation. In 2024, AskHR was able to handle 94% of all these transactions, with only 6% of the more complex cases needing human involvement. IBM is now eliminating the human HR clerks that used to perform these functions, at an estimated savings of $3.5 billion in annual costs. Despite these layoffs, IBM’s total headcount has actually grown. The savings from AI efficiencies are being deployed to increase hiring in sales, marketing, and software development - roles that are either inherently hard for AI to automate because of the requirement for soft skills such as relationship building or industry-specific expertise - or, in the case of software development, skills that contribute directly to making AI ever more capable.

IBM lays off 8,000 due to AI, but also hires over 9,000

OpenAI announces Stargate UAE, its first whole-country deal

So, let’s talk about Sam Altman’s other blockbuster deal this past week. Not enough to team up with the most famous hardware designer in tech history while seizing control of OpenAI by diluting the nonprofit arm’s stake below 50%. No, our boy Sam also inked a historic deal with the United Arab Emirates, in which the UAE put up billions in petrodollars to fund a massive data center there, while offering ChatGPT to all the residents in the country. All this was done with the blessing (indeed, the pushing) of the US government, since the Trump administration is extremely interested in getting the rich oil states in the Middle East to help fund the US drive for AI leadership. And Sam isn’t done - next week OpenAI kicks off an AI roadshow in the “Asia Pacific” region (Korea and Japan, but definitely not China) to try to sell the countrywide investment strategy there. More countries, more dollars for data centers, more ChatGPT customers - life is good for Sam these days.

OpenAI announces Stargate UAE, its first whole-country deal

Meta/Facebook struggles in AI

In AI, life comes at you fast. Last year, Meta/Facebook’s efforts in AI were riding high, based on the company’s decision to release their very capable Llama AI models to the world as open source, free for anyone to use and modify. CEO Zuckerberg, the guy who founded a social media empire because he couldn’t get a date at Harvard, was hailed as a hero of the little guys who couldn’t afford to compete in AI with megacorps like Google, Microsoft, and OpenAI. Nowadays, the luster is gone, and Meta is struggling to regain its footing. The downfall began with release of the Llama 4 model last year, which was seen as lagging behind the releases of OpenAI, Google, and Anthropic. Then China’s DeepSeek AI model showed the world that there was more than one company that was willing and able to release cutting-edge open source AI models, and several of those companies were in China. Now Meta’s next-generation Llama model, nicknamed “Behemoth”, has had its release date pushed back - apparently because it, too, has been judged to be disappointing. Worst of all, there has been a brain drain from Meta’s AI development team, with 11 of the 14 original designers of the Llama architecture having left the company. Rumors abound, but one credible one is that Zuck is contemplating splitting the Meta AI team into 2, with one team focused on achieving the Holy Grail of Artificial General Intelligence (AI as smart as humans at everything), and the other more narrowly focused on getting AI consumer products out the door.

Meta/Facebook struggles in AI

Making AI Work: Leader, Lab, and Crowd

Wharton professor Ethan Mollick is one of the world’s deepest thinkers about the applications of AI to business. His most recent post on his Substack blog is titled “Making AI Work: Leader, Lab, and Crowd.” He makes the case that transforming businesses to take advantage of the capabilities of AI at this moment in time requires 3 critical elements:

1. Leadership from the top - Because AI will require transformations of roles and processes throughout the company, employees need to know that top management is fully behind the changes, and that AI is not a fad that will just “blow over.”

2. The Crowd of employees who are allowed to experiment in ways to use AI in their job, to find efficiencies, improve speed of execution, and enhance outcomes. Companies that try to roll out AI solutions in a centralized, top-down manner, similar to what is common in IT departments today, will inevitably fall behind, because the technology is too new and changes too fast.

3. A lab for development and standardization - There is a need for a team of multidisciplinary professionals who can take suggestions from the Crowd and make them into standard products that can be distributed across the company, and can also work in a cross-disciplinary way to develop new products beyond the capabilities of the company’s enthusiasts and early adopters.

   Mollick ends:

   “The key is treating AI adoption as an organizational learning challenge, not just a technical one. Successful companies are building feedback loops between Leadership, Lab, and Crowd that let them learn faster than their competitors. They are rethinking fundamental assumptions about how work gets done. And, critically, they’re not outsourcing or ignoring this challenge.”

Making AI Work: Leader, Lab, and Crowd

AI scientist “agents” discover new treatment for serious eye disease

AI research lab FutureHouse, backed by former Google CEO Eric Schmidt, has run a test of a multi-agent AI scientific team, and almost immediately the AI system found a new treatment for dry age-related macular degeneration (dAMD), a retinal disease that can threaten sight. The system is designed around separate AI agents that can analyze research papers, generate hypotheses, and assist with the planning of experiments. In the first major test of the system, it conducted a broad literature review, generated novel hypotheses on the mechanisms causing dAMD, proposed candidate molecules for treating the disease, and ultimately was able to identify an effective treatment in the form of a medication already used to treat a different disease of the eye. This mediation can now be repurposed to treat dAMD without having to invest years and enormous financial resources to prove safety and efficacy for FDA approval.

AI scientist “agents” discover new treatment for serious eye disease

OpenAI’s o3 model finds a bug in the Linux kernel

OpenAI’s flagship o3 model is highly skilled at programming, including debugging and looking for security vulnerabilities in software. Recently it was pointed at a highly sophisticated open-source code set - the core module, or “kernel”, of the world’s leading open-source operating system, Linux, and o3 found a vulnerability no one else had ever noticed. Linux powers vast swathes of the internet, since its open-source code is widely regarded as the safest and most reliable available because so many millions of eyes, constructive as well as malicious, have viewed it and tried to find its vulnerabilities, to either patch (helpful) or exploit (malicious.) To find a new and serious vulnerability in the heart of Linux is relatively rare these days, and for the o3 model to find it in the highly complex 12,000 lines of code in the Linux kernel is an impressive feat.

OpenAI’s o3 model finds a bug in the Linux kernel

LinkedIn exec says AI is breaking the bottom rung of the career ladder

LinkedIn’s chief economic opportunity officer, Aneesh Raman, warned in a New York Times OpEd last week that AI is “breaking” the bottom rung of the career ladder by automating tasks that entry level workers used to do. By having no on-ramp to get into companies, Gen Z applicants may have no way to acquire the experience needed for higher-paying positions later, he says. He warns that the ripple effects can cause social upheaval. He advises companies to find ways to give entry level workers access to AI tools that will enable them to perform higher-value tasks, even as the low-level tasks are automated away.

LinkedIn exec says AI is breaking the bottom rung of the career ladder

Teens design robot snakes to keep runways clear of critters

Students from 4 California high schools competed to design solutions faced by a local Air Force base. One team tackled the issue of how to protect endangered tiger salamanders from being crushed when they wandered onto the base’s busy jetways. Their solution? Robot snakes that would scare away, but not harm, the salamanders.

Teens design robot snake to keep runways clear of critters

Amazon expands delivery by drone

Amazon has been experimenting with package delivery by drones for a few years now, and recently the company has decided to expand the service area from its initial sites in Texas and Arizona. Customers in areas where drone delivery is enabled can receive their package within an hour of less of ordering. Amazon reports that over 60,000 items are available for drone delivery, and the list of items is constantly expanding.

Amazon expands delivery by drone

AI is “superhuman” at clinical reasoning and diagnosis

A final version of a preprint that we reported on previously confirms that OpenAI’s o1 model demonstrates “superhuman” abilities at clinical reasoning in complex cases. Teams from Harvard, Stanford, and other elite medical institutions tested OpenAI’s o1 model against board-certified physicians’ judgement for 5 different collections of complex case vignettes. They then tested the AI model against expert physicians in actual cases randomly selected from a tertiary care hospital emergency room. In all tests, vignettes or actual cases, the o1 model was judged to have performed at a level significantly higher than the physicians.

Summary post on Twitter/X

Full paper on arXiv

New CRISPR technology moves RNA in neurons for better self-repair

Researchers at Stanford have devised a new CRISPR technology to help repair damaged nerve cells. CRISPR is generally used to repair DNA, but in this new technology, it moves RNA around the neuron to where it is most needed to promote healing of damaged nerves. This offers new avenues to explore possible treatments for spinal cord injuries and for neurodegenerative diseases like ALS, commonly referred to as “Lou Gehrig’s disease”, or the disease of the late physicist Stephen Hawking.

New CRISPR technology moves RNA in neurons for better self-repair