Nikolai Yakovenko: the $200 million AI engineer

On this episode of Unsupervised Learning, in the wake of Elon Musk's xAI Grok chatbot turning anti-Semitic following a recent update, Razib catches up with Nikolai Yakovenko about the state of AI in the summer of 2025. Nearly three years after their first conversations on the topic, the catch up, covering ChatGPT's release and the anticipation of massive macroeconomic transformations driven by automation of knowledge-work. Yakovenko is a former professional poker player and research scientist at Google, Twitter (now X) and Nvidia (now the first $4 trillion company). With more than a decade on the leading edge computer science, Yakovenko has been at the forefront of the large-language-model revolution that was a necessary precursor to the rise of companies like OpenAI, Anthropic and Perplexity, as well as hundreds of smaller startups. Currently, he is the CEO of DeepNewz, an AI-driven news startup that leverages the latest models to retrieve the ground-truth on news-stories. Disclosure: Razib actively uses and recommends the service and is an advisor to the company. Razib and Yakovenko first tackle why Mark Zuckerberg's Meta is offering individual pay packages north of $200 million, poaching some of OpenAI's top individual contributors. Yakovenko observes that it seems Meta is giving up on its open-source Llama project, their competitor to the models that underpin OpenAI and ChatGPT (he also comments that it seems that engineers at xAI are disappointed in the latest version of Grok). Overall, though the pay-packages of AI engineers and researchers are high; there is now a big shakeout as massive companies with the money and engineering researchers pull away from their competitors. Additionally, in terms of cutting-edge models, the US and China are the only two international players (Yakovenko notes parenthetically that Chinese engineers are also the primary labor base of American AI firms). They also discuss how it is notable that almost three years after the beginning of the current booming repeated hype-cycles of artificial intelligence began to crest, we are still no closer to "artificial general intelligence" and the "intelligence super-explosion" that Ray Kurzweil has been predicting for generations. AI is partially behind the rise of companies like Waymo that are on the verge of transforming the economy, but overall, even though AI is still casting around for its killer app, big-tech has fully bought in and believes that the next decade will determine who wins the future.