Elon Musk's Chief AI Brain Battles Lyme Disease! Former Student of Shing-Tung Yau, Yang Ge Announces Resignation

  Synced Report 

Musk has lost a major member!

xAI co-founder and Grok chief architect Greg Yang suddenly revealed that he has been suffering from Lyme disease.

Ultimately, he decided to step down as xAI founder and move to a behind-the-scenes, informal “advisor” role.

All along, Greg Yang has referred to himself as “xAI’s mathematician” on his personal profile.

What’s more, he is regarded as the key mathematical architect behind Musk’s vision of building an “AI that understands the universe.”

For xAI, this means losing another core member. However, the role of informal advisor means the team can still seek Greg’s help on major issues.

In the comments section, Musk and his colleagues wished Greg Yang a speedy recovery.

Lyme disease is notoriously difficult to diagnose, but fortunately, Greg Yang found a great doctor.

It all started back in 2025, when Greg fell ill and then discovered—

Even several weeks after “recovering,” he could clearly feel a drop in energy, decreased creativity, and less drive (agentic) to get things done.

Afterwards, his physical condition fluctuated, but every downturn was worse than the last.

Just accidentally eating the wrong thing would leave him extremely fatigued for days, and even working out would leave his whole body weak for days.

According to him, there was a week when even sleeping 12 hours a day wasn’t enough to recover.

After seeing a doctor, it was firmly believed these symptoms were not psychological but rather a problem with the immune system.

After several rounds of detective-like tests, he was finally diagnosed with Lyme disease.

Currently, both the medical literature and public awareness of chronic Lyme disease are lacking. For those with this illness, it’s a lonely battle.

This is a bacterial infection transmitted by tick bites, with wounds swelling red like a bullseye. Fortunately, it was discovered early—if it’s only found in your fifties or sixties, you might lose all mobility or even become bedridden.

Greg was very surprised when he found out, as he doesn’t even remember being bitten.

He said it’s likely the infection was already there, but working at xAI recently weakened his immunity, triggering the symptoms.

It’s the same for almost everyone—you only realize the importance of health after your body collapses.

Recalling this illness, Greg remembered someone once told him, “You shouldn’t push yourself so hard.” He’s grateful he did, so the problem was discovered early.

What matters now is to let his body recover gradually. Once he’s fully recharged, he’ll return with even greater drive.

Greg Yang is a classic math prodigy and exactly the kind of “hardcore geek” Musk favors.

He was born in Hunan and moved to the United States with his parents during middle school.

As for his academic background, he earned a bachelor’s degree in mathematics from Harvard University, mentored by the renowned mathematician Shing-Tung Yau.

He received one of the highest honors for undergraduate mathematics, the Morgan Prize, as well as being a Putnam Fellow.

Interestingly, Greg Yang revealed in an XSpaces chat that he once took a break from his undergraduate studies at Harvard.

During that time, he became a DJ and a drum and bass music producer. Eventually, he realized his ultimate passion wasn’t DJing but mathematics.

Afterwards, he also obtained a master’s degree in computer science from Harvard University.

After graduation, instead of following the conventional path of completing a Ph.D., he directly joined Microsoft Research (MSR), a top global AI institute, as a senior researcher.

During this time, Greg completed the Tensor Programs series of papers, establishing his theoretical status in the AI field.

Put simply, Greg Yang is best known in AI as the “father of Tensor Programs theory.” His research aims to transform deep learning from “alchemy” into a rigorous “science.”

Previously, training large models like GPT-3 was extremely expensive, so people dared not experiment repeatedly with large models.

He therefore proposed a technique called mμP (Maximal Update Parameterization).

This allows researchers to tune parameters on small models (millions of parameters) and then zero-shot transfer them directly to large models with billions or even hundreds of billions of parameters, achieving optimal results.

After founding xAI, Greg Yang continued his work on Tensor Programs. In October 2023, the fourth series of papers was released.

This paper accomplished several things—

It discusses how deep residual networks classify different deep parameterizations in the “first infinitely wide, then infinitely deep” limit.

He proposed and validated Depth-μP parameterization, allowing hyperparameter transfer along the depth dimension as well.

In short, one practical significance of Tensor Programs is that “hyperparameters of extremely large models can be transferred from small models,” directly linking theory to engineering tuning costs.

This has saved xAI astronomical computational costs.

Without this technology, training a model like Grok would cost many times more.

On his personal profile, Greg Yang makes no secret of his ambition—to build a unified theory of large-scale deep learning that can:

1) Tell us the best way to scale neural networks;

2) Provide strong theoretical understanding of these models, to guide research on safety and alignment.

His departure means Musk loses a brain that can understand his “cosmic ambitions” and turn them into executable code.

According to foreign media, two other xAI co-founders also left the company in the past year.

Amidst this top-level shakeup, Musk has also promised the launch of Grok 5 in the first half of this year—something to look forward to.