Study shows that 1 human + AI = 2 humans

A recent study of AI in the workplace found that one employee using AI produced the quantity and quality of work equivalent to that of 2-person teams. In a controlled trial at consumer products giant Procter & Gamble, 776 employees were given real product innovation challenges. The employees were randomly assigned either to work in a 2-person team, with or without using AI, or to work alone, with or without using AI. The results showed that the 2-person teams without AI and the AI-assisted employees produced roughly equivalent work, and both outperformed employees working alone without AI. The 2-person teams using AI did only slightly better overall, but were much more likely to produce a solution in the top 10% of quality. The authors end by writing, “Our results suggest that AI adoption at scale in knowledge work reshapes not only performance but also how expertise and social connectivity manifest within teams, compelling organizations to rethink the very structure of collaborative work.”

Study shows that 1 human + AI = 2 humans

SSRN paper

Google releases Gemini 2.5 Pro, and this new AI model is a banger

Companies producing leading AI models are in a frantic race for the lead in performance, with each new week featuring at least one new release that challenges the previous status quo. This week it’s Google’s turn to shine, and their new Gemini 2.5 Pro is smashing all prior records in benchmark after benchmark. It is one of the new-type “reasoning” models, and particularly strong in math, coding, and creative writing. Google promises more upgrades soon.

Google releases Gemini 2.5 Pro, which is topping all the leaderboards

CEO Sam Altman wants OpenAI to become a consumer tech company

As noted in the story above, the race to make the world’s most capable AI models is brutally competitive just now, and one of the leaders may be thinking about dropping out. Last week, OpenAI CEO Sam Altman stated publicly (in a tech-bro podcast, as is the fashion nowadays) that OpenAI was morphing into a consumer technology company, and that was going to be more valuable than being an AI model builder. The analogy used was Facebook (!). (An AI-powered Facebook?? Oh, lord, just shoot me now.)

Sam Altman wants OpenAI to become a consumer technology company

Apple sued over hype about Apple Intelligence

Apple has been hyping their “Apple Intelligence” AI since June of last year, and used that hype to suggest that consumers should buy one of their newer, more powerful, AI-ready iPhone 16s released last September. Well, now it’s almost April, and there is precious little AI available on the pricy new phones. In particular, an AI version of Siri is not projected until some time next year at the earliest, and possibly not until 2027. Now plaintiffs have filed a suit in federal court, essentially claiming deceptive advertising, and they are seeking class action status, to try to recover damages for all customers who bought the latest iPhones.

Apple sued over Apple Intelligence delays

Study shows mixed effects of chatbots on the mental health of users

Much has been written about the possible negative effects of chatbot use on the mental health of users, but there have been few rigorous studies. A paper in Nature a year ago reported on a study of 1006 Stanford students who used a personal chatbot service known as Replika. In this study, chatbot use was generally a positive factor: decreasing loneliness, unexpectedly increasing socialization with other humans (because users could practice social approach interactions in a low-stakes environment), and stunningly, being credited by 3% of users with halting their prior suicidal thoughts. Last week, MIT researchers published a paper examining the impact of chatbot use, both via text and via voice response, on users over a 4-week period. The results were complex, but in general, needy and socially vulnerable users were worse off emotionally after 4 weeks, whereas those who retained emotional distance from the chatbot had positive outcomes. It will likely take us some time to determine how to maximize positive effects of chatbots while limiting negative impacts.

Stanford’s Replika Study - chatbots prevent suicide

MIT’s chatbot study - chatbots make vulnerable people worse

Earth AI finds critical minerals in overlooked places

AI startup Earth AI uses artificial intelligence to detect the location of rare and valuable minerals from data produced in global mining operations for other metals. Based in Australia, the company has identified promising fields for the mining of copper, cobalt, gold, silver, molybdenum, and tin. Much like AI is discovering new uses for approved drugs in the medical field, Earth AI is repurposing data from existing mining operations around the world to find the locations of sites to mine rare metals.

Earth AI finds critical minerals in overlooked places

Nvidia launches quantum computing lab in Boston

Once a skeptic, Nvidia CEO Jensen Huang is now all in on quantum computing. The dominant AI chipmaker has announced that it’s opening a lab in Boston to advance applied research into how to meld quantum computing with AI supercomputers. Lab collaborators include MIT, Harvard, and several leading quantum computing startups. The lab is set to begin operations later this year. CEO Huang knows that now is the moment, while Nvidia retains chip dominance, to secure a foothold in the next big leap in computing technology.

Nvidia launches quantum computing lab in Boston

Mercedes partners with Apptronik to put humanoid robots in factories

Startup Apptronik has inked a deal with Mercedes-Benz to try out Apptronik’s humanoid robots in the automaker’s manufacturing plants. In addition, Mercedes-Benz has invested over $10 million in the robotics startup. As part of this first trial, the robots will be put to work bringing parts to the assembly areas, as well as inspecting the parts they bring. Over time, both companies will collaborate to expand the roles the robots fill as they prove their capabilities. Chinese manufacturers of electric vehicles have invested heavily in developing humanoid robots to help in manufacturing their cars. Now automakers around the globe are following suit.

Mercedes partners with Apptronik to put humanoid robots in factories

Lyft adds driverless cars to its fleet in Atlanta

Ridesharing company Lyft has announced that their customers will be able to hail driverless vehicles in Atlanta by this summer. The new initiative will be carried out in partnership with May Mobility, an autonomous vehicle company based in Ann Arbor, Michigan. Robotaxi industry leader Waymo, backed by Google, has already signed a partnership with Uber in Atlanta, so the streets of Hotlanta may be teeming with robo-cars. (Fun fact: Atlanta has been a transportation hub since at least 1837, when it was called “Terminus” because it was the endpoint of the Western and Atlantic Railroad.)

Lyft adds driverless cars to its fleet in Atlanta

AI saves life of man with rare disease that stumped his doctors

Joseph Coats, a 37-year-old man with POEMS, a rare and usually fatal blood disorder, suffered complications that made him ineligible for the only known lifesaving treatment, a stem cell transplant. His girlfriend reached out to a specialist in the disease, who then used ChatGPT to help him devise an unconventional combination of chemotherapy, immunotherapy, and steroids. It worked. Four months later, Coats was strong enough to undergo the stem cell transplant, and a year later, he is doing much better than before. The key insight from ChatGPT was to repurpose sirolimus, a medication typically used in kidney transplant patients, to treat Coats’ condition. Drug repurposing, using medications approved by the FDA for one disease to treat another disease, is a fruitful area where AI may be able to rapidly expand effective treatments for many rare conditions.

AI saves life of man with rare disease that stumped his doctors

Australian AI detects endometrial cancer with near-perfect accuracy

Researchers from Australia and Bangladesh have developed an AI model that can detect endometrial cancer with an accuracy of 99.26%. Current state of the art automated systems achieve only approximately 80% accuracy. The model analyzes images of microscope slides of tissue and annotates the slide with important findings, which helps the pathologist verify its final diagnosis. The research team behind the model say that different versions of the basic model have already been trained to diagnose other cancers. It reportedly has diagnosed colorectal cancer with 98.57% accuracy, breast cancer with 98.20% accuracy, and oral cancer with 97.34% accuracy.

Australian AI detects endometrial cancer with near-perfect accuracy