AI can already do 12% of work tasks, can double the growth of labor productivity

A new study out of MIT indicates that 11.7% of work tasks in the US labor force can be automated by today’s AI. This has been widely misinterpreted as saying that AI can replace 11.7% of US jobs. Instead, the study looked at specific work tasks that could be automated, and found that 11.7% of all tasks could be done with today’s AI technology. No job today consists of only those automatable tasks, so the impact on employment will be less , and will take a while to work through the economy, as jobs are redesigned to take advantage of AI efficiencies.

In related news, Anthropic, maker of the Claude chatbot, analyzed 100,000 conversations between users and Claude, and found that on average, AI reduced task completion time by up to 80% on the tasks that it was used for. They calculated that AI could increase the rate of labor productivity gains by 1.8% per year, essentially doubling the current rate of productivity growth. As always, the big question is who will benefit from this increase in productivity - workers, through increased pay, or companies, by reducing the number of employees.

MIT study finds that 11.7% of work tasks can be automated by today’s AI

Anthropic calculates AI can double the rate of growth in labor productivity

ChatGPT goes shopping - and will show ads

OpenAI’s massive investments in AI data centers and chips has created an imperative to begin generating revenue - a lot of it, and fast. This week, the company revealed 2 more avenues it will pursue to turn on the money spigot.

First, it announced Shopping Research, a feature in ChatGPT that turns it into your personal shopper. It will find products for you, in any category you ask, and will pull together reviews, prices, and product specifications into a side-by-side comparison. If you find a product you’d like to purchase, you can click through to the seller’s website. In the near future, you will be able to purchase items directly from within ChatGPT, using an Instant Checkout feature. Behind the scenes, OpenAI will be collecting a fee from the seller for each purchase. Early reviews of this feature have been generally favorable.

Second, it is credibly reported that OpenAI is preparing to display ads in their answers. Ad revenue is how Google makes its massive billions. Up to now, AI companies have been avoiding the use of ads, fearing user backlash. Each one is waiting, hoping that one of the others will blink first, so that they can start cashing in. OpenAI’s staggering commitments for infrastructure may force it to be the one to test the waters.

ChatGPT wants to be your personal shopper

OpenAI prepares to display ads in ChatGPT’s answers

Top music labels sign licensing deals with AI music startups

In a move reminiscent of the birth of legal online music with the creation of Apple’s iTunes store, major music labels Warner, Sony, and Universal have signed a licensing deal with AI music startup Klay. Warner simultaneously settled its lawsuit against AI music startup Suno in return for a licensing deal with that company. Evidently, both the tech bros and the labels have learned the lessons of the Napster/iTunes era of the internet - piracy bad, negotiated licenses good for business for both sides.

Top music labels sign licensing deals with AI music startups

China is outpacing the rest of the world in electricity generation

US tech giants such as Google, Microsoft, and Meta/Facebook are pouring hundreds of billions of dollars into building AI data centers. This is already starting to strain US electric power grids, as we have reported previously. In contrast, China began prioritizing building out its own electrical power generation system over a decade ago, covering deserts with solar farms and damming up large rivers for hydropower. China’s production of electricity is already greater than the US, EU, and India combined, and the rate of growth shows no sign of slacking. In the current race for AI dominance, China’s green and abundant electrical power is a strategic advantage.

China is outpacing the rest of the world in electricity generation

“The Thinking Game” documentary is now free on YouTube

Google has released its award-winning documentary film, “The Thinking Game”, for free on YouTube. The movie focuses on DeepMind, Google’s AI research lab, and its quest to solve a longstanding problem in biology - how do genes create proteins? DeepMind ended up developing AlphaFold, an AI system that can predict how a protein will fold into its functional form just from looking at the DNA sequence that codes for that protein. The leader of that effort, Demis Hassabis, was awarded the 2024 Nobel Prize in Chemistry, and a knighthood (he’s a Brit.) It’s an interesting look at recent AI history, although not surprisingly, it focuses heavily on the positives of AI, with less attention to controversies.

“The Thinking Game” documentary is now free on YouTube

Chinese parents use AI assistants as homework monitors

Chinese parents of schoolchildren have found a new use for AI assistants - as tireless monitors of their child’s diligence while doing homework. The parent will set up their phone with the camera facing the child at their desk, and then initiate a session with Doubao, an AI system from ByteDance, the developer of TikTok. Doubao is optimized for real time feedback from visual input. If the child’s gaze wanders from their work, Doubao will correct them in a firm tone of voice. Even posture is monitored - slumping gets another warning.

Chinese parents use AI assistants as homework monitors

Pompei’s pieces are being put together with robots

Pompei was destroyed by a volcanic eruption in 79 AD. Many ancient Roman frescoes (paintings on plaster walls) were shattered. More frescoes were damaged by the bombings in World War II. Restoring these ancient treasures is painfully slow work. Now a team of scientists is developing a robotic system to speed the restoration efforts up. The project combines advanced image recognition, AI-driven puzzle-solving, and ultra-precise robot hands. The professor from Venice’s Ca’ Foscari University who coordinated the robot project, Marcello Pelillo, has been quoted as saying “It’s like you buy four or five boxes of jigsaw puzzles. You mix everything together, then you throw away the boxes and try to solve four or five puzzles at the same time.” Without a reference picture for any of the puzzles.

Pompei’s pieces are being put together with robots

Harvard develops AI system that identifies rare genetic disorders

Harvard researchers have developed an AI system, called popEVE, that can identify variations in DNA that are likely to cause significant disease. When the genomes of 30,000 individuals with developmental disorders of unknown cause were run through the popEVE AI system, diagnosis of the causative abnormal gene was achieved for about one-third of them. The system found pathological variations in 123 genes that had never before been identified as causing disease. Subsequently, 25 of these variants have been confirmed by other labs as causing the patient’s malady.

Harvard develops AI system that identifies rare genetic disorders