Podcast Summary: In the City — What the DeepSeek Freak Out Means for the UK's AI Future
Date: January 30, 2025
Host(s): Francine Lacqua, Allegra Stratton (Bloomberg)
Guest: Parmi Olson (Bloomberg Opinion columnist, author of "Supremacy: AI, ChatGPT and the Race that Will Change the World")
Overview
This episode dives into the shockwaves sent through the global technology and financial sectors by the rise of DeepSeek, a young Chinese AI company that released a high-quality generative AI model built on lower-cost, less advanced chips compared to Western tech giants. The discussion focuses on the wider implications for the US, China, and UK's roles in the ongoing AI race, and what it means for markets, investment, and the City of London.
Key Discussion Points & Insights
1. The "DeepSeek Freak Out": Why Now?
-
Massive Market Ripples:
DeepSeek unveiled an open-source AI model that rivals the best from OpenAI, Google, and others, using a fraction of the resources. This resulted in dramatic market responses, including Nvidia’s largest stock drop ever.- Allegra Stratton: “On Monday… every hour seemed to be a sort of end of the world headline.” [01:54]
- Francine Lacqua: “It’s the fact that DeepSeek over the weekend came in with this lower cost AI model. So…if you can do the same for cheaper, then China has a massive advantage compared to the US companies.” [02:09]
-
The "Sputnik Moment":
The DeepSeek announcement was repeatedly described as a ‘Sputnik moment’—a wake-up call and potential paradigm shift in global technology leadership.- Allegra Stratton: "It was powerfully called a Sputnik moment… this huge game changer in technology, but also geopolitics." [02:26]
2. Technical & Economic Shifts: Cheap AI and Diminishing Returns
-
Scaling Laws Hit a Wall:
Previously, growth in AI was achieved by scaling up (more data, bigger models, more powerful chips), but industry leaders recently admitted those returns are now diminishing.- Parmi Olson: “Just a few weeks ago… the formula we’ve been using … is starting to have diminishing returns.” [04:01]
- “What DeepSeek has shown is… there is a workaround… as good as ChatGPT but for a fraction of the cost.” [04:40]
-
Silicon Valley Jolted:
The sudden success of a cheaper, open alternative is forcing tech giants to reconsider both their technical roadmaps and massive investments.- Parmi Olson: “In one way that’s very distressing for people in Silicon Valley because… it makes it look like all the money they’ve poured into AI has just been a big old waste.” [04:58]
- Allegra Stratton: “What is going on right now? What are they having huge meetings saying everything we thought we understood, we have to revise the numbers?” [05:29]
3. China’s Innovation Through Constraint
-
From Disadvantage to Leapfrogging:
US sanctions denied Chinese firms access to leading Nvidia chips, forcing them to innovate within constraints.- Parmi Olson: “They started with a disadvantage which was sanctions from the US… but when you limit yourself it forces you to be more creative, to think out of the box. That seems to be just what’s happened in China.” [06:19]
-
Cost Advantage:
DeepSeek claims to have developed their breakthrough model with a budget of just $6 million—less than the annual compensation of a Google executive.- Parmi Olson: “They’ve said that the budget they used for this new model was $6 million. That’s about what a Google executive gets paid every year.” [07:20]
- Allegra Stratton: “Yesterday, such a killer fact.” [07:53]
4. Open Source Disruption and Western Response
-
Industry Panic & Market Correction:
Questions arise about the value and sustainability of Silicon Valley’s huge investments, with open-source models leveling the playing field.- Parmi Olson: “What happened to the NASDAQ yesterday, you could call it a healthy correction to the market. The hype has been way overblown.” [09:36]
- “DeepSeek is open source… there’s no reason why Google and Facebook can’t copy their methods.” [09:54]
-
Skepticism About Real “Openness":
The hosts probe whether DeepSeek is truly open-sourcing all its methods and data, with guests affirming—at least on the surface.- Francine Lacqua: “…Why would Deep Seek have an open… I mean, is it really open or are they keeping parts for themselves?” [10:50]
- Parmi Olson: “As far as I know, it’s open, it’s out there. You can use this.” [11:08]
5. Geopolitical and Regulatory Fallout
-
US: Fear and Double-Down Strategy:
Discussion on how the Trump administration might respond—likely through renewed investment and deregulation to stay ahead of China.- Parmi Olson: “Trump is the kind of person that doubles down… his mandate… America first and let’s make sure we’re ahead of China…” [12:12]
-
Tech Giants’ Business Models Under Threat:
OpenAI and rivals need urgent new strategies to justify their business models.- Parmi Olson: “The most stressed person right now is Sam Altman because his entire company’s business model is based on selling a foundation model.” [14:29]
-
Profitability Still Unclear:
Despite multimillion-dollar revenues, most generative AI products are not yet profitable, fueling fears of an AI investment bubble.- Parmi Olson: “OpenAI made something like 5 billion in revenue, but it also lost 7 billion. Its path to profit is unclear.” [13:17]
6. The UK’s AI Ambitions: Niche, Regulator, or Innovator?
-
UK as AI Safety Regulator:
The UK could carve out a “market niche” by leading in AI safety and smart, pragmatic regulation—with more specialists in the government’s AI Safety Institute than in all of Brussels.- Parmi Olson: “I think he [Keir Starmer] should really play up the UK’s role as a safety regulator… there’s like about 150 people working at the AI Safety Institute in London… a lot of expertise.” [15:57]
-
Out-Innovating the Giants — UK Optimism:
DeepSeek’s story inspires hope that outsized innovation isn’t just about resources—it’s about thinking creatively. The UK could blend nimble regulation, open innovation, and start-up energy.- Parmi Olson: “DeepSeek has shown you don’t need that [big money]. You just need innovation. You need smart ideas. You need to think out of the box, think creatively and open source…” [17:21]
-
Less Regulation than EU — A Distinct Model:
The UK, by regulating less than the EU, aims to be the more attractive jurisdiction for innovation.- Allegra Stratton: “…the UK point or position has always been that we will regulate less than the EU. And so being that kind of more palatable version…” [18:09]
Notable Quotes & Memorable Moments
-
On the Disruptive Power of DeepSeek:
Parmi Olson: “It literally is a fraction of what Google and OpenAI are paying for their own equivalent models. …I think people are just stumped and stunned and we’re grasping for answers.” [07:55] -
On Market Shocks:
Allegra Stratton: “The amount they [Nvidia] lost yesterday was bigger than all of the money that Donald Trump managed to get all the AI companies to put in last week… So where does this go… It feels like it’s hard to see that those huge numbers that Silicon Valley kick around are sustainable medium term.” [08:54] -
On the Future of UK AI:
Parmi Olson: “…DeepSeek has shown you don’t need that. You just need innovation. You need smart ideas… and open source might be…” [17:21]
Timestamps for Important Segments
- DeepSeek as a "Sputnik Moment" and Market Reaction: [01:52–02:43]
- The Technology Behind DeepSeek, and Why It Upended Assumptions: [04:01–07:55]
- Silicon Valley's Response and Panic: [05:29–06:54], [09:36–10:24]
- Open Source, Copying, and the Future of AI R&D: [09:36–11:12]
- Geo-political Implications and the Trump Administration's likely response: [11:53–13:09]
- Profitability Questions for Tech Giants: [13:09–14:29]
- Who Is Most Stressed—Sam Altman & OpenAI: [14:15–14:29]
- Can the UK Carve Out A Regulatory Niche? [15:18–17:11]
- UK’s Model, Optimism, and Distinctiveness from EU & US: [17:11–18:27]
Tone & Style
The conversation is urgent, dynamic, and sometimes humorous, reflecting both the anxiety and excitement coursing through tech and finance circles in London amid unexpected disruption. Expert insight is tempered by candid uncertainty as the guests repeatedly acknowledge how quickly long-held assumptions are being overturned.
