Michael I. Jordan: Machine Learning, Recommender Systems, and Future of AI | Lex Fridman Podcast #74 | Summary and Q&A
Michael I. Jordan discusses the need to broaden the scope of artificial intelligence to focus on decision-making and creating meaningful connections between producers and consumers.
Questions & Answers
Q: How does Michael I. Jordan characterize the current state of artificial intelligence?
He believes that the current focus on prediction and pattern recognition is limited and neglects the importance of decision-making and creating meaningful connections between producers and consumers.
Q: What example does Jordan use to highlight the potential of AI in decision-making and connecting producers and consumers?
He uses the example of creating a music market, where independent musicians can connect directly with listeners and monetize their work, bypassing traditional record companies.
Q: Why does Jordan criticize the advertising model in current social media platforms?
Jordan believes that the advertising model prioritizes companies' monetization over establishing genuine connections between producers and consumers, leading to a lack of transparency and trust.
Q: According to Jordan, what is needed for AI to facilitate better decision-making and connections between producers and consumers?
Jordan suggests a shift in the business model, where companies prioritize direct connections between producers and consumers and offer transparency, allowing consumers to trust and value the recommendations and connections facilitated by AI systems.
In this video, Lex Fridman interviews Michael I. Jordan, a professor at Berkeley and an influential figure in the fields of machine learning, statistics, and artificial intelligence. They discuss the history of AI, the limitations and challenges in understanding the human brain, the role of machine learning in decision-making systems, and the need for creating markets that connect producers and consumers in the digital realm.
Questions & Answers
Q: How does Michael I. Jordan see the development of AI in comparison to other engineering fields like chemical engineering or electrical engineering?
Michael I. Jordan sees the development of AI as similar to the development of chemical engineering from chemistry or electrical engineering from electromagnetism. Just as those fields emerged when there was a need to build factories that produce chemicals and electrical systems, AI is emerging as a proto field that combines statistical and computational ideas to build systems that bring value to human beings.
Q: Does Michael I. Jordan believe that there is something deeper about AI dreams and aspirations compared to other engineering fields?
While there may be dreams and aspirations for AI in the future, Michael I. Jordan believes that the current understanding of intelligence is limited. He explains that scientists are clueless about how the brain actually does computation and that there is still much to discover and understand. He emphasizes that we are not close to achieving true artificial intelligence and that current developments in AI are more focused on building systems that serve human needs rather than understanding intelligence at a deep level.
Q: Could breakthroughs in neuroscience, such as brain-computer interfaces, lead to deeper understanding of the brain and advancements in AI?
Michael I. Jordan does not believe that breakthroughs in neuroscience will lead to deep understanding of the brain or advancements in AI in the near future. He emphasizes that scientists are still far from understanding the fundamental principles of how the brain works and that attempts to build brain-computer interfaces or mimic human intelligence are more like speculation rather than science. While he acknowledges that there may be interesting developments in areas like brain-related diseases, the notion of achieving true artificial intelligence is still beyond our current understanding.
Q: What are some possible breakthroughs that could occur in the next five to ten years?
Michael I. Jordan believes that there may be breakthroughs in the field of AI in the coming years, but he clarifies that they may not be impressive in terms of mimicking human intelligence or achieving deep understanding. He mentions that breakthroughs in areas like chemical engineering, electrical engineering, or even companies like Google, Amazon, and Uber, which bring value to human life through large-scale systems based on data and market forces, can be considered as breakthroughs. He suggests that new fields and disciplines may emerge to address the challenges and opportunities of the AI era.
Q: How does Michael I. Jordan differentiate between AI and machine learning?
While acknowledging that there is no perfect terminology, Michael I. Jordan prefers to distinguish between AI and machine learning. He describes AI as an intellectual aspiration proposed by John McCarthy to engineer algorithms that mimic human intelligence. On the other hand, machine learning is a field that focuses on systems that learn and make decisions based on data. He emphasizes that the field of AI that exists today is more about building systems that use statistical and computational ideas to make decisions, rather than achieving true artificial intelligence.
Q: What is one of the most interesting disagreements between Michael I. Jordan and Yann LeCun?
Michael I. Jordan explains that he and Yann LeCun, both respected figures in the field of machine learning, do not have fundamental disagreements. While they emphasize different aspects like pattern recognition and prediction (LeCun) or decision-making and value creation (Jordan), they both share a vision of building systems that work effectively and create value. He states that they have more in common than differences and that their shared goal is to advance the field of AI and create good working systems.
Q: What are some challenges in creating AI systems that scale decision-making in a distributed way?
Michael I. Jordan highlights the challenge of creating new markets that connect producers and consumers in the digital realm. He uses the example of the music market, where many talented musicians create music but struggle to monetize their work. He envisions a system where producers can directly connect with consumers and have access to data that shows where their music is being listened to. However, he acknowledges that creating such markets requires thinking through issues like trust, privacy, and economic incentives. He suggests that recommender systems and other algorithms play a crucial role in connecting consumers to creators, but cultural credibility and a blend of technology and economic ideas are also essential.
Q: How can companies like Spotify, YouTube, and Netflix create markets that allow creators to make a living?
Michael I. Jordan believes that companies like Spotify, YouTube, and Netflix can create markets for creators by blending technology with cultural credibility. He emphasizes the importance of connecting producers and consumers in an economic way, rather than relying solely on advertising revenue. He suggests that recommender systems play a crucial role in connecting consumers with creators, but companies need to think beyond advertising and consider new economic models that support creators and allow them to monetize their work.
Q: How do recommendation systems and AI algorithms contribute to connecting consumers with creators?
Michael I. Jordan acknowledges the importance of recommendation systems in connecting consumers with creators. He mentions Amazon's book recommendations as an example of a successful recommendation system that helps consumers discover new books based on their preferences. However, he notes that recommendation systems are most effective in certain domains, such as books, and are more challenging in other areas like restaurants or news. He believes that recommendation systems, combined with economic principles and matchmaking algorithms, have the potential to create new markets between producers and consumers.
Q: What are the challenges faced by platforms like YouTube and Twitter in terms of recommendation systems and preventing controversial or harmful content?
Michael I. Jordan believes that the challenges faced by platforms like YouTube and Twitter are primarily related to their monetization models, which heavily rely on advertising revenue. He suggests that these platforms have not prioritized creating markets between producers and consumers, leading to a lack of economic incentives for positive, high-quality content. He argues that a shift towards more direct economic connections and markets could help address the issues of controversial and harmful content, as well as the challenges of recommendation systems. He emphasizes the need to think beyond advertising and consider cultural credibility, trust, and economic value creation.
Summary & Key Takeaways
Michael I. Jordan argues for the importance of expanding artificial intelligence to include decision-making and connecting producers and consumers.
He suggests that the current focus on prediction and pattern recognition neglects the complex realm of decision-making in real-world contexts.
Jordan uses the example of creating a music market to illustrate the potential of AI in facilitating connections between musicians and listeners.