Survey: 6 in 10 managers use AI to make personnel decisions

Online career services company ResumeBuilder has released a new survey of 1,342 US managers, exploring their use of AI tools at work. Almost two-thirds of those surveyed said they did use AI in their jobs, and not just for help in drafting memos, training materials, or reports. Almost all of the managers who used AI tools used them to help make high-stakes decisions about hiring, firing, raises, and promotions. More than 1 in 5 said they frequently let the AI make the final decision, without human input. Nearly half of managers reported that they had been tasked by their company to assess whether AI could replace some or all of their direct reports. And yet, two-thirds of managers using AI in these ways said that they had received no formal training in how to use AI.

Managers are using AI to make decisions on hiring, firing, and promotions

The AI talent-poaching war rages on

As we reported last week, Meta/Facebook CEO Mark Zuckerberg, in a desperate ploy to claw his way back into the front ranks of AI companies, launched a full-scale assault to recruit away top-flight engineers from rivals OpenAI, Google, and others. Zuck offered signing bonuses of up to $100 million, and 4-year contracts worth $300 million to top prospects. By last week, he had bagged at least 14 big names from rivals, 8 of them from OpenAI. This week, the talent war rages on. Zuck poached Apple’s head of AI models, Ruoming Pang, leaving the House That Steve Built in an even more hapless and rudderless state than before. Siri may stay dumb as a rock forever. Not one to turn the other cheek, OpenAI CEO Sam Altman ended last week with 4 new hires from Meta, Tesla, and Elon’s xAI, in a tit-for-tat poach-a-thon. Having made his point, Altman called for talks with Zuck, aiming for a cease fire. In this war, no doubt both Sam and Zuck think that they are Zelensky, when most observers think that they are both Putin.

Zuck poaches on, snagging Apple’s head of AI models

The Empire Strikes Back: OpenAI poaches from Zuck and Elon

Companies pledge $23 million to train 400,000 teachers to use AI

Yesterday OpenAI announced that they were joining with Microsoft and rival AI startup Anthropic to fund a National Academy of AI Instruction with $23 million in cash and in-kind contributions. The 3 AI giants are allying with 2 large teachers unions, the American Federation of Teachers and its New York-based affiliate, the United Federation of Teachers, to train 400,000 teachers in how to use AI in classroom education. Although the Academy is a timely and welcome boost for teachers, the AI partners hope that it will blunt criticism that AI is being used mainly for cheating on homework, and will serve as an entry point for making schools, students, and parents into loyal users.

Companies pledge $23 million to train 400,000 teachers to use AI

Google to buy 200 MW of FUSION (?!?) power from MIT spinoff

AI needs massive amounts of computer power, which means massive amounts of electricity. So massive, that AI companies see access to electrical power as perhaps the most critical barrier to growth. This power imperative has prompted AI companies to pay for brand new conventional gas-fired electrical generators, and to refurbish old nuclear power plants (even the notorious Three Mile Island plant) or contract to build new ones. Now, in an even more speculative move, Google has inked a deal to buy 200 Mega-watts (enough to power a small city) of electrical power from Commonwealth Fusion Systems, a company that spun off from MIT in 2018. Atomic fusion, in which 2 atoms of hydrogen are squeezed together under such tremendous heat and pressure that they fuse into a single helium atom, is the process that powers the sun. Theoretically, fusion offers almost unlimited power, but after decades of trying, scientists have to never yet gotten a sustainable fusion reaction. But that day may be coming sooner rather than later. Commonwealth has raised almost $2 billion in its 7 years of existence, and this deal hints that both Google and the company are feeling bullish about the near-term future of fusion power.

Google to buy 200 MW of fusion power from MIT spinoff

AI avatars are earning millions on YouTube

YouTube has long been known as a platform that helps top content creators, called “influencers” for the power they wield over their fan base, to monetize their fame. The current top earner is 27-year-old Jimmy Donaldson, known online as MrBeast, whose whacky stunts on his YouTube channel have made him a billionaire (with a “B”.) Now a new generation of online personalities is emerging, who are entirely fictional: AI generated avatars are growing in popularity. One of the most successful of the new breed is called Bloo, a blue-haired gamer boy with over 2 million subscribers to his YouTube channel, earning over 1 million euros over the past 2 years. Bloo is the brainchild of Jordi Van Den Bussche, a 29-year-old resident of Amsterdam, who used to create content on YouTube the old-fashioned way - as himself. The daily grind of constantly scripting, performing, recording, and posting videos wore him out, and he decided to use AI for as much of the process as possible. The result was his blue-haired alter ego, the AI avatar Bloo, whose appearance and voice (in multiple languages simultaneously) are AI generated. Bloo is not autonomous, but is puppeteered by Van Den Bussche. However, the 29-year old creator is looking forward to the day that AI improves to the point that he can fully retire from the process and just cash those monthly checks from YouTube.

AI avatars are earning millions on YouTube

Wharton professor: the 2 paths to getting good at using AI

AI expert and Wharton Business School professor Ethan Mollick recently posted a succinct guide to the 2 paths he has seen for getting good at using AI in your work. The first path is technical, and involves deeply understanding how the current AI models work. However there is a second, equally successful path, open to non-technical people. That way involves deeply understanding how to give people instructions, and the information they need to accomplish the task you have given them. Mollick reports that in his experience, many coders struggle with non-deterministic systems in a way that managers and teachers do not. He recommends that you treat the AI like a student or an eager junior staffer, who needs both clear directions and actionable information.

Wharton professor explains the 2 paths to getting good at using AI

Impostor clones Secretary of State Rubio’s voice to spy

You knew it was going to happen. Someone created an AI-generated voice clone of Secretary of State Marco Rubio, and used it on the encrypted communication app Signal (beloved by President Trump’s top staffers, including Secretary of Defense Pete Hegseth) to try to uncover sensitive information. According to the Washington Post, the impostor contacted at least 5 government officials, including 3 foreign ministers, a US governor, and a member of Congress. Apparently Rubio caught wind of the impersonation before damage was done, and sent a cable to all State Department employees revealing the ruse and advising caution with any unusual messages.

Impostor clones Secretary of State Rubio’s voice, tries to spy

Amazon on track to employ more robots than people

Amazon recently announced that it had deployed its 1 millionth robot, and simultaneously revealed that it has upgraded its internal robot fleet management software, a move which is projected to increase efficiency of operations by 10%. Robots at Amazon are used primarily in warehouse operations, helping fulfill the nearly 13 million orders the company receives each day. Amazon currently employs approximately 1.5 million people, also mostly in their warehouse fulfillment centers. As the robots become more capable, the need for human pickers and sorters will lessen, and the day will likely come that robots outnumber humans at Amazon.

Amazon on track to employ more robots than humans

Surprise 2nd in robot marathon lifts maker from flop to hot startup

Chinese startup Noetix Robotics was struggling to find its first customer just a few months ago. But then, its hobbit-sized N2 humanoid robot won a surprising second place in the recent robot half marathon in Beijing, outpacing many much better funded competitors. Now the company is awash in orders, and is raising $35 million in new funding for expansion. This Cinderella story is validation of China’s strategy of staging publicity stunts and contests to generate buzz about robotics. Next up: China is staging the World Humanoid Robot Olympic Games, in which robots from multiple different makers compete in track and fields events, gymnastics, and either athletic challenges.

Google AI drug discovery company launching drugs into clinical trials

Google’s DeepMind AI research lab won a 2024 Nobel Prize for its AlphaFold AI system, which cracked a decades-old problem of how to predict the precise folding structure of a protein. This information is crucial to understanding or predicting how the protein will function in the body. DeepMind’s CEO, Demis Hassabis (since knighted by the King), has predicted that AI will be able to cure all human disease within 10 years. Now another arm of Google, Isomorphic Labs, is using AI to discover new drugs. Isomorphic Labs has recently announced that it is staffing up to launch several new AI-discovered cancer drugs into human trials. Boston-based Insilico Medicine was the first AI-powered drug discovery startup to launch human trials, but Isomorphic Labs, with deep pockets behind it and Nobel Prize-winning AI systems, is likely to make its mark as well. The FDA has also gone on record with its goal to use AI to shorten the time between drug discovery and approval.

Google AI-powered drug discovery company to launch drugs into clinical trials

AI allows couple to conceive after 18 years

A couple who had been trying to conceive for 18 years finally succeeded with the help of AI. Columbia University researchers developed the STAR AI system, which optically analyzes semen for the presence of viable sperm. The AI system is many times more effective than humans peering through microscopes, and in the case of this couple, where the husband had extremely low sperm counts, was able to find multiple viable sperm, when humans were unable to find any. Male infertility is implicated in around 40% of couples who are unable to conceive. Male infertility is also on the rise around the world, for reasons that are poorly understood. This couple, who wish to remain anonymous, are expecting their child in December.

AI allows couple to conceive after 18 years