Technology

➚ MOVE FAST: Alumni. Former researchers at Google are looking to help Japan catch up in the AI race. Their pitch using smaller, cheaper models will allow more companies there to tap into the technology. Meanwhile, OpenAI alum Elon Musk wants to raise up to $4 billion for his AI startup to better compete against his old venture.

➘ BREAK THINGS: Outsiders. The Chinese government is putting pressure on telecom companies there to stop using foreign chips like those made by Intel and AMD. China Mobile and others have been ordered to phase them out by 2027. It’s the latest tit-for-tat move in the tech cold war between the U.S. and China.

An old Japanese ceramics maker, Maruwa, is now projected to be worth $2.75 billion as its heat-dissipating materials are increasingly used to cool AI chips.

Maruwa builds circuit boards made up of ceramic composites containing metals that make it effective at absorbing heat. Keeping semiconductors cool is vital for their performance, and data centers spend a good chunk of their electricity bills on cooling. Shares in Maruwa have risen in line with demand for data center capacity.

“We expect that next-generation high-speed communications, including those related to generative AI, will be the biggest driver of our business growth over the next few years,” the company told the Financial Times.

Analysts from Goldman Sachs believe Maruwa has 60% of the global market share in high-temperature ceramics, and its deep expertise in manufacturing the materials, extending more than 200 years, provide a competitive advantage.

Symbolic is built around belief that OpenAI is wrong

THE SCENE

Venture capitalist Vinod Khosla made one of the shrewdest bets of his career in 2019, when he put $50 million into OpenAI, now the darling of the generative AI boom. In a sign of just how fast that industry is moving, Khosla recently made another big investment, this time in a company built around the belief that OpenAI has got it all wrong.

Symbolica AI, co-founded by former Tesla Autopilot engineer George Morgan, is building a new AI-assisted coding tool that it says uses a new method of machine learning that works completely differently from the cutting edge foundation models made by OpenAI, Google, and other major AI companies.

With this new approach, Morgan says Symbolica’s models won’t require the same massive, power-hungry compute that companies are now spending tens of billions of dollars to procure for the most advanced AI models.

In an interview with Semafor, Morgan said those investments are based on speculation that, if given enough data and enough compute resources, later versions of these models will become more intelligent.

Without mathematical proof showing how these things work, Morgan says the process is more like alchemy. “That’s exactly what AI models are today,” he said. “You mix a bunch of random stuff together, you test it, you see if it does the thing or not. If it doesn’t, you try something else. What Symbolic is doing is bringing this into the era of chemistry.”

REED’S VIEW

I haven’t seen Symbolica’s technology yet, so I can’t vouch for its capabilities. But tech leaders have told me that the next big breakthrough in AI is likely removing humans from the architecture step of its development.

There is a big market for what Symbolica is trying to build. While consumers may want to chat with AIs about any topic, businesses want the opposite, with language interfaces that are narrowly focused and totally reliable. Hallucination, in many enterprise contexts, is simply unacceptable.

In the near term, the large transformer models will get bigger and more capable. According to people who have used the still under wraps GPT-5, the next generation of OpenAI’s technology, it is much closer to reasoning abilities than GPT-4. It still hallucinates and is definitely not AGI, but it sounds like it is good enough that it will be more useful to a wider swath of customers.

Ten years from now, we may look back and realize that scaling transformer-based models only got us so far, and that new methods like Symbolica’s represented the path to AI with reasoning capabilities.

Even if that’s the case, today’s foundation models will have played a critical role. ChatGPT and the resulting AI craze has inspired a wave of investment and talent pouring into the field.

We’ve got some exciting additions to the list of speakers at the World Economy Summit, including Jeremy Hunt, UK Chancellor of the Exchequer; Gina Raimondo, U.S. Secretary of Commerce and Former U.S. Treasury Secretary Larry Summers. Join us in Washington, D.C. next week, on April 17-18, to hear from some of the world’s most influential economic and business decision-makers on the future of global economic growth, the rising middle class, digital infrastructure, AI, and much more.

Regulators are zeroing in on Big Tech’s links to AI companies. The UK’s Competition and Markets Authority said yesterday that it discovered an “interconnected web” of more than 90 partnerships and investments involving the same six firms: Google, Apple, Microsoft, Meta, Amazon, and Nvidia.

And it listed three main areas of concern: firms controlling crucial inputs for developing foundation models that could harm competition, big tech players distorting the market for such models, and partnerships with key AI-related firms that could strengthen the market power of major companies.

“Now, with a deeper understanding and having watched developments very closely, we have real concerns,” said Sarah Cardell, CMA’s CEO.

Meanwhile, her EU counterpart, Margrethe Vestager, told Bloomberg TV that Brussels plans to wrap up its probe of Microsoft’s investment in OpenAI and global watchdogs are “comparing notes” on these kinds of partnerships.

In the U.S., the Federal Trade Commission is also looking into these investments but framed its move as an inquiry rather than an investigation, and sought information from several companies. That means that once again, overseas regulators will likely be the first ones to bring enforcement actions against American companies.

OpenAI CEO Sam Altman is rallying government officials and industry leaders around the world to support building more infrastructure to increase chip manufacturing and expand data centers.

He traveled to the United Arab Emirates earlier this week to meet with government officials to see how OpenAI could help the country develop its network of data centers. He was then headed to Washington to hold similar meetings with Biden administration representatives, and has reportedly spoken to leaders in Europe too.

It’s reflective of the massive infrastructure needs of AI, and that companies can’t put those pieces together on their own. For Altman, being the emissary also means a better shot at controlling the resources needed to advance the technology. But dealing with governments looking out for their own national interests will test his skills as a global AI diplomat.

• How one company made bank off of the regional banking crisis.
• What China’s EV buyers’ remorse means for the world.
• Tougher EU migration rules spark anger across the political spectrum.