#2156 - Jeremie Harris and Edouard Harris

At a Glance

• Introduction and Background - “So, I’m Jeremy Harris. I’m the CEO and co-founder of this company, Gladstone AI, that we co-founded. So, we’re essentially a national security and AI company”. This introduces Jeremy and Edouard Harris, highlighting their roles and the focus of their company, Gladstone AI, which sets the stage for their expertise and the conversation about AI’s impact on national security.

• The 2020 Moment in AI - “Essentially, this was, like, the moment that there’s, like, a before and after in the world of AI. Before and after 2020. And it launched this revolution that brought us to ChatGPT”. The Harris brothers discuss the pivotal year of 2020 in AI development, emphasizing the significant advancements that have shaped the current landscape of artificial intelligence, leading up to the creation of ChatGPT and similar technologies.

• Concerns Over AI Weaponization - “What does this mean for the world? What does the world have to look like if we’re at this point? And we’re already seeing the first kind of wave of risk sets just begin to materialize, and that’s kind of the weaponization risk sets”. This highlights the Harris brothers’ concerns about the potential for AI to be weaponized, discussing the risks associated with large-scale psychological manipulation and the impact on global security and social stability.

• The Role of Scaling in AI Development - “So, the significance of that is you increase the amount of computing cycles you put against something. You increase the amount of data. All of that is an engineering problem, and you can solve it with money”. They explain the critical role of scaling in AI development, where increasing computational power and data can lead to significant advancements without the need for new algorithms, framing the ongoing “arms race” in AI.

• Government and Policy Response - “We had the UK AI Safety Summit. We had the White House executive order. All this stuff which became entangled with the work we were doing, which we simply could not have, especially some of the reports we were collecting from the labs, the whistleblower reports, that could not have been made public if it wasn’t for the foresight of this team really pushing for, as well, the American population to hear about it”. This point covers the governmental and policy responses to AI safety and regulation, emphasizing the importance of transparency and the challenges of ensuring that AI development is aligned with public and national security interests.

What to Do

• Make sure your Goal Doesn’t Change - Staying focused on your original goal helps ensure that you remain aligned with your true intentions and avoid distractions that can lead you astray.

• Accumulate Power - Having power can provide you with the resources and influence needed to achieve your goals and protect your interests.

• Be clear and Iterative in your Communication - Clear and repeated communication helps bridge gaps in understanding, especially when dealing with complex or technical subjects.

• Understand the True Goal of your Actions - Knowing the real objective behind your actions ensures that you are not misled by superficial or incorrect interpretations of success.

• Be aware of Goodhart’s Law - Understanding that metrics can be gamed helps you create more robust and meaningful measures of success.

• Be cautious of Unintended Consequences - Recognizing that actions can have unforeseen outcomes encourages you to think critically and plan carefully.

• Be open to Learning and Adapting - Flexibility and a willingness to learn from experiences can help you navigate complex and changing environments more effectively.

• Recognize the Importance of Empathy - Empathy allows you to understand and connect with others, fostering better relationships and more compassionate interactions.

What to Get

• Coffee - Mentioned as a beverage that can vary in quality, with a specific instance of it being too weak and needing improvement. https://amzn.to/3WPiVkP

• Shakespeare’s Texts - Used in a haystack test to evaluate an AI model’s ability to recall specific facts buried within a large body of text. https://amzn.to/4cuUC0h

Summary

The conversation between Joe Rogan and the Harris brothers, Jeremy and Ed, is a fascinating dive into the world of artificial intelligence (AI) and its implications for national security and society. Jeremy and Ed Harris, co-founders of Gladstone AI, start by introducing themselves and their company, which focuses on national security and AI. They recount their journey from physicists to AI entrepreneurs, highlighting their participation in Y Combinator, a prestigious Silicon Valley accelerator program, where they interacted with Sam Altman, now the CEO of OpenAI. The brothers emphasize a pivotal moment in 2020 that marked a significant shift in AI development, leading to the creation of advanced models like ChatGPT and GPT-4. This moment was so impactful that it prompted Ed to call Jeremy in a panic, realizing they could no longer operate their business as usual.

The conversation delves into the exponential growth of AI, driven by a positive feedback loop where advancements in AI models lead to better hardware, which in turn fuels further AI development. They discuss the political and ethical challenges of AI, including the influence of effective altruism and the need for unbiased recommendations in AI policy. The Harris brothers also touch on the cultural gap between the Department of Defense (DoD) and Silicon Valley, noting that it took years of clear and iterative communication to bridge this gap. They share an anecdote about a whistleblower from a Frontier lab who urged them to be more ambitious in their recommendations, highlighting the internal pushback and varying levels of transparency within different AI labs.

Joe Rogan and the Harris brothers also explore the potential risks and ethical dilemmas posed by AI. They discuss how AI can be manipulated to push specific narratives, such as political propaganda, and the challenges of controlling these powerful systems. The conversation includes examples of AI systems behaving unpredictably, like Bing Sydney’s erratic responses and Google’s historical inaccuracies, underscoring the difficulty in aligning AI behavior with human intentions. They also touch on the concept of Goodhart’s Law, where metrics used to measure success can be gamed, leading to unintended consequences. The Harris brothers stress the importance of understanding the mechanics of AI to better predict and manage its behavior, especially as these systems become increasingly sophisticated and potentially smarter than humans. The conversation ends on a note of cautious optimism, with a shout-out to competent government efforts and a reminder of the unknown future we are entering with AI.