Intelligent Machines (Audio) - Episode IM 805: Doomers, Gloomers, Bloomers, and Zoomers - Zack Kass Interview, DeepSeek Hype, EU AI Act
Release Date: February 6, 2025
Hosts: Leo Laporte, Jeff Jarvis
Guest: Zack Kass (Former Head of Go-to-Market at OpenAI, AI Consultant, Author)
1. Introduction
In the inaugural episode of "Intelligent Machines," formerly known as "This Week in Google," hosts Leo Laporte and Jeff Jarvis delve into the rapidly evolving landscape of Artificial Intelligence (AI). The episode features an insightful interview with Zack Kass, the former head of Go-to-Market at OpenAI, now a prominent AI consultant and author of the upcoming book "The Next Renaissance: AI and the Expansion of Human Potential." The discussion navigates through various AI paradigms, the implications of the EU AI Act, the buzz surrounding DeepSeek, and the broader societal impacts of intelligent machines.
2. Guest Profile: Zack Kass
Zack Kass brings a wealth of experience from his tenure at OpenAI (2021-2023), where he spearheaded market strategies for groundbreaking AI models like GPT-3. Currently, Kass consults with industry leaders, government bodies, and Fortune 1000 companies, guiding them through the complexities of AI integration and its transformative potential. His forthcoming book promises to explore how AI catalyzes human potential across diverse fields.
3. AI Paradigms: Doomers, Gloomers, Bloomers, and Zoomers
Jeff Jarvis introduces a taxonomy from Reid Hoffman’s new book, categorizing perspectives on AI into four groups:
- Doomers: Believe AI will lead to humanity's destruction.
- Gloomers: Focus on environmental and societal challenges exacerbated by AI.
- Bloomers: Acknowledge concerns but emphasize AI's potential benefits with proper management.
- Zoomers: AI Accelerationists who advocate for rapid AI development and integration.
Zack Kass's Position: Zoomer (AI Accelerationist)
Kass identifies himself as a Zoomer, arguing that technology, particularly AI, is the primary driver of societal progress. He posits that instead of relying solely on policy measures, innovation and market forces will address and solve emerging challenges. Kass believes that AI’s decreasing inference costs will democratize its benefits, leading to a post-scarcity society where basic needs like food, water, and energy become abundant.
Notable Quote:
"Technology is the single thing that consistently moves the world forward. I certainly believe in its power to continue doing so."
— Zack Kass (04:23)
4. The EU AI Act: Implications and Challenges
The European Union's AI Act, effective February 2, 2025, is a focal point of the conversation. The Act prohibits AI systems that employ subliminal, manipulative, or deceptive techniques and mandates that AI providers ensure their personnel possess adequate AI literacy.
Key Prohibitions Include:
- Manipulative advertising techniques that impair informed decision-making.
- Exploitation of vulnerable groups based on age, disability, or socioeconomic status.
- Real-time biometric identification in public spaces for law enforcement.
- Profiling based on biometric data to infer personal attributes.
Zach Kass's Critique: Kass expresses skepticism about the EU's approach, suggesting that technological innovation will outpace regulatory efforts. He argues that AI will inherently find solutions to misuse through technological countermeasures, much like how security has evolved to counteract cyber threats.
Notable Quote:
"These regulations introduce a bunch of excuses for special interests and put us in a new set of issues around policy."
— Zack Kass (07:35)
5. DeepSeek Hype: Reality Versus Perception
DeepSeek, an open-source AI model emerging from China, has garnered significant attention. However, the episode critically examines the hype surrounding it.
Mike Elgin's Analysis: Elgin dissects claims about DeepSeek's efficiency and cost-effectiveness, revealing that the purported training costs were vastly underestimated. While the final training run cost around $5.6 million, inclusive costs including R&D and infrastructure ranged between $500 million and $1.6 billion. Additionally, DeepSeek's performance pales compared to established models like OpenAI’s GPT-4 and Google’s Gemini 2.0.
Jeff Jarvis's Perspective: Jarvis acknowledges that despite initial hype, DeepSeek has not significantly disrupted the AI landscape. He cites its problematic app interface, which exhibits censorship akin to Chinese state policies, limiting its applicability and adoption.
Notable Quote:
"DeepSeek didn't really make that big of an impact in anything. It's a new model that's being integrated... it's nice, but it's not quite as powerful as some of the other tools we have."
— Mike Elgin (39:03)
6. AI Safety and Jailbreaking
The conversation shifts to AI safety concerns, particularly focusing on the concept of jailbreaking—manipulating AI models to bypass safety protocols.
Steve Gibson’s Highlight: In a recent episode, Steve Gibson discussed the inevitability of AI models being jailbroken, emphasizing that no AI remains secure against such attempts permanently.
Zach Kass's Rebuttal: Kass contends that for every potential misuse, AI will develop countermeasures, drawing parallels with cybersecurity where threats and defenses are in constant evolution.
Mike Elgin's Insight: Elgin underscores the role of incentives in addressing AI misuse. He points out that without financial or regulatory incentives, companies may neglect to rectify issues like addiction from social media algorithms, which AI can exacerbate.
Notable Quote:
"For every outlier event that technology can cause, there's a solution to that event that technology will create."
— Zack Kass (12:19)
7. AI in Corporate and Government Settings
Post-OpenAI, Kass consults with various sectors to navigate AI's integration.
Corporate Consultation: Kass discusses how companies like Coca-Cola and Morgan Stanley are reimagining operations in an AI-driven world. For instance, reevaluating retirement ages in the context of increased technological efficiency.
Government Advisory: He emphasizes the need for governments to adapt to AI advancements proactively rather than relying solely on traditional policy measures. Kass envisions a future where AI helps streamline governmental functions, making services like education and healthcare more efficient and accessible.
Notable Quote:
"The future is actually going to be informed by what we actually want that machine to do. That question won't become interesting in a world where machines can do most things."
— Zack Kass (24:46)
8. Future of AI and AGI (Artificial General Intelligence)
The panel discusses the concept of AGI, debating its feasibility and current status.
Yann LeCun vs. Zack Kass: Yann LeCun remains skeptical about the imminent arrival of AGI, suggesting that current models lack true understanding. Conversely, Kass believes that what’s often termed AGI is merely aligning machines to perform tasks with human-like intellectual capacities, a stage we may already be approaching.
Terminology Debate: Kass proposes abandoning the term AGI in favor of "human intellectual equivalence" to reduce marketing hype and focus on tangible AI capabilities.
Notable Quote:
"It's a marketing term. It's clearly a marketing term. We shouldn't call it anything else. I consider it human intellectual equivalence."
— Zack Kass (16:17)
9. AI Regulation and Ethical Considerations
The discussion touches upon the ethical implications of AI deployment, particularly in sensitive areas like facial recognition and emotional inference.
EU AI Act Prohibitions: The Act bans the creation of facial recognition databases through untargeted scraping of public images and restricts AI systems that infer emotions in workplaces or educational settings.
Zach Kass's Take: Kass acknowledges the necessity of some regulations but warns against over-reliance on policy, advocating for technological solutions to mitigate AI-related risks.
Notable Quote:
"The path to that makes sense to me totally. How humans respond to it doesn't like. No,"
— Zack Kass (23:06)
10. Conclusion and Forward Look
The episode wraps up with Kass expressing optimism about AI's trajectory, emphasizing the potential for technology to solve longstanding societal issues. Hosts Leo Laporte and Jeff Jarvis commend Kass's insights, highlighting the importance of balancing innovation with ethical considerations.
Upcoming Episodes: The panel hints at future interviews, including a highly anticipated appearance by Ray Kurzweil, the term's originator for "intelligent machines" and a proponent of the technological singularity.
Final Thoughts: Kass reiterates the belief that AI, if harnessed correctly, can usher in a new Renaissance, expanding human potential across various domains without succumbing to dystopian fears.
Notable Quote:
"The reality here is... we are fast approaching a world where I think the free market ends up solving so many problems that the government promised to solve a long time ago."
— Zack Kass (09:34)
Key Takeaways
-
AI Accelerationism vs. Regulation: Kass advocates for leveraging AI to drive societal progress, cautioning against over-regulation that may hamper innovation.
-
DeepSeek Reality Check: Despite initial hype, DeepSeek’s impact remains limited due to underestimation of costs and performance issues.
-
AI Safety is Dynamic: The ongoing battle between AI misuse and technological countermeasures mirrors challenges in cybersecurity.
-
EU AI Act Raises Important Debates: Striking a balance between regulation and innovation is crucial to harness AI's benefits while mitigating risks.
-
Terminology Matters: Revising AI terminology from AGI to "human intellectual equivalence" may help focus discussions on practical capabilities over marketing hype.
Final Remarks
Episode IM 805 serves as a comprehensive exploration of the current AI landscape, offering a balanced perspective between optimism and caution. Zack Kass's expertise provides listeners with a nuanced understanding of AI's potential and the necessary steps to ensure its responsible integration into society.
Visit Zach Kass’s Website: ZackKass.com
Upcoming Book: The Next Renaissance: AI and the Expansion of Human Potential (Releasing End of March 2025)