Big Technology Podcast: DeepSeek's Fallout For AI Companies, OpenAI’s Path Forward, Siri Somehow Got Worse Hosted by Alex Kantrowitz Release Date: January 31, 2025

Introduction to the Week's Major Themes

In this episode of the Big Technology Podcast, host Alex Kantrowitz and guest Ranjan Roy delve into the seismic shifts occurring in the AI landscape following the emergence of DeepSeek. The discussion spans the ramifications for major AI players like OpenAI, Anthropic, Nvidia, and Apple, particularly focusing on the decline of Siri and falling iPhone sales.

DeepSeek's Impact on AI Companies

DeepSeek's Disruption Across AI Margins

At [00:32], Alex Kantrowitz introduces the topic by outlining the four key areas where margins exist in the AI business:

1. Hardware: Dominated by Nvidia, which has been emblematic of the AI boom.
2. Data Centers: Controlled by giants like Microsoft’s Azure, Amazon’s AWS, and Google Cloud.
3. AI Model Building: Previously seen as a profitable область, but recent developments challenge this view.
4. Applications: Encompassing products like ChatGPT that leverage AI models.

The episode highlights how DeepSeek has drastically altered these margins by introducing a model that is significantly cheaper to run. This shift suggests that the commoditization of AI models is accelerating, potentially eroding the profitability of companies focused solely on model development.

Quotes:

• Alex Kantrowitz [01:25]: "The Deep Seek sell-off was only five days ago. Now these weeks are getting longer."
• Ranjan Roy [03:08]: "LLMs will be commoditized for I think about a year and a half now."

OpenAI's Product Strategy and Challenges

Reevaluating OpenAI's Business Model

Alex discusses how OpenAI’s revenue projections, which relied heavily on API sales, are now under scrutiny. With DeepSeek proving that models can be run at a fraction of the cost, OpenAI's focus on applications like ChatGPT becomes both a strength and a vulnerability.

OpenAI’s Response to DeepSeek

Sam Altman’s reaction to DeepSeek is dissected, revealing a mix of commendation and underlying panic:

• Sam Altman [07:01]: "DeepSeek R1 is an impressive model... invigorating to have a new competitor."

Despite acknowledging DeepSeek’s advancements, OpenAI appears unprepared for the competitive pressure, particularly in the model-building arena.

Product Challenges: Operator and Sora

Alex and Ranjan explore OpenAI’s recent product launches—Operator and Sora—which have underperformed expectations:

• Ranjan Roy [16:22]: "Operator... it's not a great product yet."
• Alex Kantrowitz [10:23]: "Operator is supposed to be like a browser takeover tool, but it falls short in practical use."

These products highlight the tension between OpenAI’s technological ambitions and the practical usability of their offerings.

The Emergence of Open Source in AI

DeepSeek as an Open Source Challenger

DeepSeek’s open-source approach has democratized access to powerful AI models, posing a significant challenge to proprietary model builders like OpenAI and Anthropic. The accessibility and cost-effectiveness of DeepSeek have disrupted the traditional AI hierarchy.

Industry Reactions and Opinions

• Dario Amodei, CEO of Anthropic [26:00]: Discusses the implications of DeepSeek’s innovations and export controls, emphasizing the need for continued efficiency in model training.
• Tom Wolf, Co-founder of Hugging Face [29:00]: "Having one of the models freely available... renders the whole closed source arms race competition argument with AI unconvincing."

These perspectives underscore a growing consensus that open-source models are reshaping the AI ecosystem, fostering innovation and competition.

Meta and Other Players in AI

Meta's Strategic Response

Meta faces internal concerns as DeepSeek surpasses their open-source models. Although initially worried, Meta executives like Yann LeCun and Mark Zuckerberg advocate for integrating DeepSeek’s advancements into their own offerings to maintain a competitive edge.

Application vs. Model Leadership

Alex and Ranjan debate whether leading in model development is essential or if excelling in application development suffices. Meta’s vast user base offers a strategic advantage in deploying AI applications, potentially mitigating the need for proprietary model supremacy.

Nvidia’s Position

Shifting Narratives Around Nvidia

Traditionally seen as the bedrock of AI hardware, Nvidia now faces uncertainty as DeepSeek demonstrates that effective AI can be achieved with less advanced hardware. Ranjan remarks:

• Ranjan Roy [38:35]: "The invincibility of Nvidia for me is gone."

This shift challenges Nvidia’s valuation and its previously uncontested position in the AI hardware market.

The AI Application Market and Jevons Paradox

AI Becoming a Commodity

Satya Nadella’s assertion that increased efficiency and accessibility of AI will lead to its ubiquitous adoption aligns with the concept of Jevons Paradox, where improvements in resource efficiency result in higher overall consumption.

Surge in AI Applications

With cheaper and more accessible AI models, the application layer is poised for explosive growth. Ranjan emphasizes:

• Ranjan Roy [23:01]: "This is like Andreessen's time to build... hundreds of new startups."

This trend is expected to foster a rich ecosystem of AI-driven applications, democratizing AI utility across various sectors.

Security Concerns with DeepSeek

Data Exfiltration Allegations

Recent reports suggest that DeepSeek may have improperly accessed OpenAI’s data via APIs:

• Bloomberg [45:26]: "Microsoft probing if DeepSeek improperly obtained OpenAI data."

This incident raises serious concerns about data security and ethical practices within the AI community, exacerbating tensions between proprietary and open-source AI developers.

Apple's Siri Issues and Falling iPhone Sales

Siri’s Decline in Performance

John Gruber’s critique highlights significant declines in Siri’s accuracy:

• John Gruber [50:06]: "Siri got 34% of those right which is truly disastrous."

Despite integrating generative AI into Siri, Apple faces backlash as user satisfaction dwindles, questioning the efficacy of their AI initiatives.

Impact on iPhone Sales and Market Perception

Apple reported a decrease in iPhone sales this quarter, coupled with criticisms of their AI strategies:

• Alex Kantrowitz [51:01]: "Siri is garbage and iPhone sales are falling."

Despite these setbacks, Apple’s stock paradoxically rose by 6%, potentially driven by robust performance in other areas like services revenue.

Concluding Thoughts and Future Outlook

Open Source's Ascendancy and Proprietary Challenges

The episode underscores a pivotal moment where open-source AI models like DeepSeek are redefining industry standards, challenging established players to innovate or risk obsolescence. The future landscape suggests a shift towards application-focused AI development, fostering a more diverse and competitive ecosystem.

Anticipated Developments

Looking ahead, the podcast anticipates:

• Continued rise of AI-driven startups focusing on applications.
• Nvidia re-evaluating its role and strategy in the AI hardware market.
• Apple needing to overhaul its AI strategies to regain user trust and satisfaction.

Notable Quotes:

• Alex Kantrowitz [24:40]: "This has just been the week where it's been the open source week."
• Ranjan Roy [38:35]: "The invincibility of Nvidia for me is gone."

Join the Conversation: The hosts invite listeners to join their new Discord server for ongoing discussions about AI developments, DeepSeek, and more. Visit bigtechnology.com to participate.

This episode of the Big Technology Podcast offers a comprehensive analysis of the dynamic shifts within the AI industry, highlighting the disruptive impact of open-source models like DeepSeek and the ensuing challenges faced by established players such as OpenAI, Anthropic, Nvidia, and Apple. The nuanced conversation provides valuable insights for listeners keen on understanding the future trajectory of AI technology and its broader economic implications.