Odd Lots Podcast Summary: "The AI Model That Tanked the Stock Market" Bloomberg, Released January 28, 2025
Introduction
In the January 28, 2025 episode of Bloomberg's Odd Lots, hosts Joe Weisenthal and Tracy Alloway delve into a seismic event in the financial and technological sectors: the stock market's dramatic downturn triggered by the emergence of a new open-source AI model named Deepseek. The episode features an in-depth conversation with Zvi Moshowitz, a renowned AI and quantitative finance expert. This summary encapsulates the key discussions, insights, and conclusions drawn during the episode.
1. The Deepseek AI Model and Market Turmoil
The episode opens with a stark revelation by Joe Weisenthal about the massive market impact caused by Deepseek, an open-source AI model developed in China. He states:
"It seems to be catalyzed by the introduction of this high-performance open-source Chinese AI model called Deepseek. [...] there are now concerns about, of course, a cheap Chinese competitor." (01:31)
Tracy Alloway humorously attributes a significant portion of the market loss to Weisenthal's own enthusiasm for AI:
"Because last week you wrote that you were a deep seek AI bro. And look what you've done. You've wiped $560 billion off of Nvidia's market cap." (02:27)
This dramatic decline primarily affected tech giants like Nvidia, which experienced a notable 17% drop in stock value within a single day, underscoring the fragility of market sentiments tied to AI advancements.
2. Training Costs and Technical Insights
Zvi Moshowitz provides a comprehensive breakdown of Deepseek's development, emphasizing its cost-effectiveness:
"They did all these sorts of mini little optimizations... in order to train as cheaply as possible on 15 trillion tokens and get the same level of performance or close to the same level of performance as other companies have gotten with much, much more compute." (05:53)
He clarifies that while the initial training cost was reported as $5.5 million for Deepseek V3, the total investment, including data acquisition, engineering, and infrastructure, amounted to hundreds of millions of dollars. This distinction is crucial in understanding the true financial commitment behind developing competitive AI models.
3. Open Source AI and Its Broader Implications
The conversation shifts to the philosophical and practical ramifications of open-source AI. Moshowitz discusses Deepseek's ideological stance:
"They are essentially ideologically in favor of the idea that everyone should have access to the same AI... it doesn't mean that you are $5.5 million away from your own V3." (07:17)
He raises concerns about the potential risks of releasing highly capable AI models openly, citing the existential threats posed by Artificial General Intelligence (AGI). The democratization of such powerful technology could lead to unforeseen and possibly dangerous applications.
4. Competition Among AI Laboratories
The episode examines the competitive landscape of AI development, highlighting the position of major players like OpenAI, Meta, and Google. Moshowitz notes:
"OpenAI should be very nervous... it's a direct threat to their business model for obvious reasons." (15:10)
He underscores that Deepseek's advancements have narrowed the gap between established AI laboratories, making the race for AI supremacy more intense. The discussion also touches upon Google's overlooked contributions and the challenges faced by competitors like Anthropic in keeping pace with rapid AI advancements.
5. Implications for Nvidia and the Chip Industry
A significant portion of the discussion centers on Nvidia's role in the AI ecosystem. Moshowitz argues against the notion that Deepseek's efficiency lowers overall chip demand, invoking the Jevons Paradox:
"As we get further along towards AGI... there's basically unlimited demand to do that or to scale up the quality of the answers as the price drops." (09:50)
He contends that increased efficiency in AI models will drive greater demand for compute resources, thereby benefiting chip manufacturers like Nvidia. Despite Deepseek's ability to optimize chip usage, the overarching need for more powerful hardware remains unabated.
6. The Concept of Artificial General Intelligence (AGI)
Tracy Alloway probes the often-debated term "Artificial General Intelligence," seeking clarity on its definition. Moshowitz explains:
"Generally it is understood to mean you can do any task that can be done on a computer, that can be done cognitively only as well as a human." (08:49)
The hosts discuss the implications of reaching AGI, emphasizing the transformative and potentially disruptive impact it could have on various industries and society at large.
7. Google's Position in the AI Race
Towards the episode's conclusion, the hosts and Moshowitz reflect on Google's strategic positioning:
"Google's had its rhetorical lunch eaten over and over and over again." (15:40)
Despite being a powerhouse in AI research, Google's contributions often go underappreciated in public discourse. However, Moshowitz suggests that Google's substantial investments in AI infrastructure and model optimization could position them strongly against competitors like OpenAI and Deepseek.
Conclusion
The episode of Odd Lots provides a thorough examination of the immediate and far-reaching consequences of Deepseek's introduction to the AI landscape. Through expert analysis, the hosts highlight the intricate interplay between AI advancements, market dynamics, and the broader economic implications. The discussion underscores the pivotal role of AI in shaping future technological and financial paradigms, urging stakeholders to remain vigilant and adaptive in this rapidly evolving field.
Notable Quotes
- Joe Weisenthal (01:31): "It's pretty deep. Has anyone made that joke yet?"
- Tracy Alloway (02:31): "Because last week you wrote that you were a deep seek AI bro. And look what you've done."
- Zvi Moshowitz (05:53): "They did all these sorts of mini little optimizations... to train as cheaply as possible."
- Zvi Moshowitz (07:17): "It doesn't mean that you are $5.5 million away from your own V3."
- Zvi Moshowitz (09:50): "There's basically unlimited demand to do that or to scale up the quality of the answers as the price drops."
- Zvi Moshowitz (15:10): "OpenAI should be very nervous... it's a direct threat to their business model."
For more insights and detailed discussions, listeners are encouraged to tune into the full episode of Odd Lots on Bloomberg.