Technology

➚ MOVE FAST: Chips money. Intel finally got some good news.The Biden administration is set to announce multi-billion dollars in funding for the chipmaker to boost production in the U.S. Samsung and others are also poised to benefit as the cash spigot from the CHIPS Act finally opens.

➘ BREAK THINGS: Chips orders. McDonald’s suffered a global tech outage that affected orders in the U.K., Australia, Japan, and elsewhere. The company ruled out a hack but disgruntled customers still took to social media to express their unhappiness. I’m not lovin it.

Anduril released a video yesterday showing off its latest autonomous weapon, a drone with a range of 100 miles that can fly for 75 minutes before taking out an armored vehicle, like a tank. Anduril says the ALTIUS-700M is comparable to the AGM-114 Hellfire, a rocket-powered, laser-guided missile that costs up to $100,000 each.

This improved weapon isn’t huge news. But I thought it was a good opportunity to look at how AI is already changing the nature of warfare.

The video is somewhat mesmerizing. The 700M, initially shot from a tube that might otherwise launch a missile, flies around looking for its target. It eventually finds an armored tank and destroys it with impressive force.

The most interesting angle, though, is the software that controls the drone. Anduril uses a system, which it calls Lattice, to orchestrate attacks. A swarm of these drones can fly by themselves, searching for enemy targets while a single human watches them on a faraway screen, ultimately approving the final target. Other Anduril drones can even be reused if they don’t hit a target.

It’s an example of how consumer tech knowhow (Anduril was founded by Palmer Luckey, creator of the Oculus VR headset) promises to massively reduce the cost of war for the U.S. and its allies — both in terms of taxpayer funds and human life.

The unexpected labor disruptions from AI-powered robotics

THE SCENE

Wellness startup Aescape’s pending launch of AI-powered robotic masseuses is one of the unexpected ways the technology could disrupt the labor market.

Founder Eric Litman has raised $80 million to develop “the first commercially available fully automated, customizable massage experience to give people more control of their self-care.” New York-based Aescape is partnering with 10 Equinox gyms for the rollout, with a half-hour session costing $60.

The machine looks like something you might see on the Starship Enterprise. It uses infrared scanners to map your body and sensors in the robotic hands to determine the pressure and motion of the arms.

With a target market of hotels, spas, and gyms, the robotic masseuses could put human ones out of work. Litman is quick to point out that there are currently 30,000 open masseuse jobs in the U.S. and that number is likely to go up.

REED’S VIEW

This potentially has dire implications for masseuses, who could be replaced by robotic arms. But then, what happens when a therapeutic massage is accessible to the masses? Will it enable people with back and muscle issues to remain more active, thus living healthier lives?

That’s probably a matter of debate, but the point is that with each new AI-enabled tool, the ripple effects reach far beyond a single profession.

The only thing I’m sure about is that, thanks to automation, the world is going to change drastically over the next five to 10 years in ways we cannot imagine.

Giant foundation models like GPT-4 have hit an inflection point and investment dollars are pouring in. But it’s the investments you don’t publicly see that are the most impactful right now. Those are going into the infrastructure layer of AI. At every point in the technology stack, teams of engineers are focusing on solving every little bottleneck. There is money to be made in clearing the way for AI innovation.

And something similar is happening in robotics. Robots today are somewhat rigid. They can be trained for specific tasks and useful in controlled environments. But researchers are figuring out how to train them in a way that’s similar to large language models, throwing enormous datasets at them and letting them learn on their own so that they can become more useful for general purposes. These two worlds of AI — the software and hardware — are colliding in ways that will likely lead to exponential growth in capability.

We often ask whether these nascent technologies will take human jobs. That will happen, but it will also create new ones, and even more importantly, it may change the concept of work itself.

Join me at the 2024 World Economy Summit in Washington, D.C. on April 17-18, where I’ll be moderating sessions on AI & Policy.

Over 1,000+ senior, forward-thinking leaders have registered to join us as well as a world-class line-up of speakers including Tom Lue, General Counsel, Google DeepMind; Kevin Scott, CTO, Microsoft; Max Fenkell, Head of Government Relations at ScaleAI; Alissa Starzak, Global Head of Policy, Cloudflare; Nandan Nilekani, Co-Founder & Chairman, Infosys and David A. Zapolsky, SVP, Global Public Policy & General Counsel, Amazon and 50+ more global economic leaders.

See all speakers, sessions & RSVP here.

The UAE is stepping up its already ambitious AI plans with a new investment firm focusing on deals involving the technology and semiconductors, Bloomberg reported. MGX, with sovereign wealth fund Mubadala, already a big investor in the tech industry, and AI firm G42 as founding partners, could have more than $100 billion in assets in just a few years.

In a reflection of the importance of MGX to the UAE government, the company will be chaired by the country’s National Security Advisor Sheikh Tahnoon bin Zayed Al Nahyan. OpenAI CEO Sam Altman recently met with him to discuss his push to expand chips production that can facilitate AI development, and the FT reported MGX is considering investing in that effort.

Kevin Xu, investor and founder of the Interconected blog, unpacked a fascinating presentation by the Beijing Academy of Artificial Intelligence on China’s challenges in developing large, homegrown AI models. They include an over-reliance on Meta’s open-source Llama technology and challenges in producing high-quality content and adherence to “facts,” a problem for LLMs in general but an especially complicated one in China, where censorship and other government controls color online information.

Aside from the political complexities of Congress’ effort to force a divestiture or ban of TikTok, there are fundamental tech and business problems. Because of a lack of jurisdiction, the plan carves out just the U.S. assets of the social media app. But only about 170 million of TikTok’s 1.3 billion users are in America; most of the growth opportunity is outside the country.

Let’s say someone did want to buy just the U.S. portion of TikTok. The company’s algorithm is the secret sauce that has made it so popular, turning the human brain into a malleable, content-consuming putty. That makes TikTok an amazing business with endless potential. But the algorithm depends on data. Lots and lots of data.

If TikTok U.S. can’t get the data from the vast majority of TikTok users, it’s not worth much. And if the international version of TikTok doesn’t rely on U.S. data, then U.S. users won’t get served the best content. Essentially, Instagram and YouTube would have a good shot at catching up to and surpassing TikTok’s addictiveness.

Because of its reach, interested buyers are circling. But some due diligence may find TikTok’s standalone U.S. business isn’t as viral as it appears to be.

• As inflation soars, Nigerians are raiding food warehouses.
• How Adobe’s Firefly repeats the same blunders as Google’s Gemini.
• Why Don Lemon’s deal with X just collapsed.