Biden executive order opens federal land to AI data centers

As the days of his Presidency wind down, Joe Biden is still running full tilt to secure as much legacy as he can. Yesterday Biden issued an executive order that allows AI companies to lease federal land owned by the Department of Defense and the Department of Energy in order to build AI data centers. Companies that build data centers on these lands will be required to bring online enough clean energy to fully offset their large electricity demands, and to avoid increasing utility rates for local residents. Energy demands of data centers in the US is expected to double in the next 5 years.

Biden executive order opens federal land to AI data centers

Microsoft commits to $80 billion this year to build data centers

Microsoft recently announced that it is planning to spend $80 billion on building AI data centers - this year. To put this staggering sum into perspective, the company’s entire annual operating revenue (profit) is approximately $109 billion. The company is therefore planning to spend almost all of that profit on data center expansion this year. Microsoft President and Vice Chair Brad Smith has called AI “the electricity of our age,” and the company is clearly putting its money where its mouth is.

Microsoft commits to spending $80 billion this year on AI data centers

AI is overtaking traditional search

Navigating the vast reaches of the internet’s 2 billion websites is performed primarily by search. Google is so dominant in this space that approximately 90% of all searches use its search engine. So dominant, in fact, that we call search “googling.” But now users are turning more and more to a new way of finding information on the web, using AI to deliver actual answers rather than a list of links (many of which have paid Google a fee to be ranked near the top of the list, regardless of their actual relevance to your query.) The poster child for the new AI-powered query engines is Perplexity, an AI startup that delivers Wikipedia-style snippets of distilled information along with links to sources. Anyone who has ever struggled to find the right keyword to unlock some sought-for story or data, only to be confronted with a blizzard of links of dubious relevance, can see the appeal of the new way. The appeal is so strong, that even Google is supplementing its list of links with an informative summary called AI Overview. The danger for Google is that the vast majority of their revenue is derived from ads tied to clicks on links. No clicks on links, no revenue, and maybe even, no Google. An internet without Google? Hard to imagine, even 2 years ago. AI comes at you fast.

AI is overtaking traditional search

World Economic Forum: AI will eliminate 92 million jobs by 2030 - but will create 170 million

The World Economic Forum is best known for sponsoring Davos, the annual global schmoozefest for the rich and powerful. But they consider themselves Very Serious People, and so they periodically release reports on the future of the world economy. Their latest such report is a good news/bad news projection of AI’s impact on global employment. The bad news? They project that AI will eliminate 92 million jobs by 2030 - just 5 years away. The good news is that they foresee that AI will simultaneously create 170 new jobs in that time, for a net gain of 78 million jobs. If accurate, the WEF projections mean that the next half-decade will see seismic changes to the job market, creating a tremendous need for reskilling employees from manual processes to AI-enabled workflows.

WEF: AI will create 78 million more jobs than it eliminates by 2030

White House announces 3-tiered export controls on AI chips

The White House has announced rules on export controls of advanced AI chips. To the Biden White House, the world, like Caesar’s Gaul, is divided into 3 parts: friends, frenemies, and hostile competitors. The map below shows the breakdown. Friends (close allies like the EU and Japan) will face no restrictions on importing the most powerful chips from the US. Frenemies (most of Latin America, Africa and Asia) will be subject greater restrictions, and hostile competitors (such as China, Russia, and Iran) will be barred from importing the most powerful chips. How effective these export controls will be in maintaining US dominance in AI remains to be seen.

White House announces 3-tiered export controls on AI chips

Salesforce CEO says no more hiring coders - AI will do it

Business software giant Salesforce’s CEO Marc Benioff announced recently that the company will not hire any software engineers in 2025, due to a 30% increase in coding productivity from using AI code assistants. Of course, he earlier announced that he is hiring 2,000 new salespeople to flog his new AI agents products, which offer customers pre-built AI “employees” to perform various repetitive tasks throughout the enterprise. As previously reported here, Benioff believes AI will be either the core of his business in the future, or a threat to his company’s very existence.

Salesforce CEO says no need to hire more coders - AI will do it

OpenAI announces Tasks, for automating future actions

OpenAI has announced a new feature for paying customers of ChatGPT - Tasks, a first stab at trying to make AI that actually does stuff, rather than just talking about it. With Tasks, you can set up automated future actions, such as asking ChatGPT to search the web for news on topics that interest you, and deliver you a daily news digest designed specifically for you. Tasks is currently in its beta-test phase, so use at your own risk, but the company says a fully-fledged version is coming soon, and will eventually be offered to all users, free or paid.

OpenAI announces Tasks, for automating future actions

Hobbyist builds AI-assisted rifle robot in his garage

A viral TikTok video took the internet by storm last week. In the clip, a balding and bearded hobbyist with, apparently, lots of spare time and a fertile imagination, demonstrated his latest project: a robotic rifleman controlled by voice commands fed through ChatGPT. The hobbyist/inventor demonstrated his ability to control the rifle-bot by voice, commanding it to fire in certain directions or patterns, with near-instant response from the robot. The result is eerie, exciting, and disturbing in equal measures. Having the powerful intelligence of leading-edge AI models available to everyone (and anyone) is going to lead to unpredictable new creations - good and bad.

Hobbyist builds AI-assisted rifle robot in his garage

Mom says Grok AI diagnosed broken wrist missed by doctors

This past weekend, a mother tweeted a post on Twitter/X about how Elon Musk’s Grok AI model, available to Twitter subscribers, diagnosed her daughter’s fractured wrist, which had been missed by the physicians treating her. Apparently, the teenage daughter was injured in a car accident and complained of a sore wrist. The wrist was X-rayed at an Urgent Care clinic, and the physician on duty told the patient and the mother that there was no fracture. Mom asked for the Xray to be printed out for follow up with the daughter’s PCP. When the daughter’s wrist was still very painful the next day, Mom scanned the X-ray print and uploaded it to Grok for an AI second opinion. Grok diagnosed a wrist fracture. Mom then took her daughter to an Orthopedic physician, who ordered multiple X-rays with different views, and ultimately confirmed a displaced wrist fracture. The daughter was appropriately casted, potentially avoiding future surgery for an improperly healed fracture. Patients will increasingly turn to AI for diagnoses and second opinions, so physicians and health systems need to be prepared.

Mom says Grok AI diagnosed broken wrist missed by doctors

AI tool helps fight health plan denials

Public outrage in the US over health plan denials of payment for care ordered by their physicians is on the rise (as was recently highlighted by the gleeful reactions to the assassination of the United Healthcare CEO.) A number of companies are working to develop AI tools to help physicians and patients fight back. Healthcare-payments processor Waystar has just announced a new AI tool that automatically drafts appeal letters for denials, cutting the time it otherwise takes physicians and staff to produce these documents manually. We may be entering an AI arms race, where AI writes appeals for claims denied by the health plan’s AI, and the health plan will use AI to automatically respond to the appeal. It reminds us of this doggerel from the acerbic English writer Jonathan Swift: “Great fleas have little fleas upon their back to bite ‘em. And little fleas have lesser fleas, and so ad infinitum.”

AI tool helps fight health plan denials

Cancer-sniffing dogs use AI to detect tumors

Patients are often surprised that there is no good general test for cancer. Israeli researchers have now developed what they call a bio-hybrid multiple-cancer detection test. They first trained Labrador retriever puppies to detect 4 different cancers (lung, breast, prostate, and colorectal) from the smell of a patient’s breath, captured on a surgical mask that the patient breathed through for 5 minutes. The researchers then developed an AI model that scored the dog’s behavior after sniffing a patient’s mask, to produce the final result. Accuracy of the test was reported as 94%. Not quite the high-tech Tricorder used by the doctor on Star Trek to instantly diagnose any disease, but then, real life is often stranger than fiction.

Nature article on the Israeli cancer-sniffing dogs and AI