🎥 In conversation with: John Larson
AI advisor John and I discuss the future of AI, its impact on work, and the evolving landscape of AI investments.
This is an edited transcript from a conversation featuring John Larson, a seasoned AI builder and advisor with extensive experience in AI/ML as a top management consultant to Fortune 500 executives. We delved deep into the transformative power of AI and its implications for the future of work and investments.
John Larson’s insights into the future of AI and its broader implications were invaluable, and his perspectives are essential reading for anyone interested in AI’s evolving landscape. We discussed:
How businesses can lead in AI by viewing their workforce as AI contributors
The flawed argument of AI replacing human jobs
The future of AI investments and the concept of value density in the AI market
The complex dynamics between open-source and proprietary AI models
Enjoy the Honest AI edit below.
Alberto (A): Today, we have the pleasure of speaking with John Larson, a seasoned AI consultant with over 15 years of experience in machine learning and AI. John has a rich background, including advising investors, builders and executives on AI at McKinsey and leading enterprise AI deployments at Accenture. He is an early leader of large language models (LLMs) enterprise deployments building and scaling AI products across the entire US as far back as 2019. Welcome, John. To kick things off, could you walk us through your journey into AI and your career trajectory so far?
John (J): Absolutely, Alberto. My journey into AI started with an interest in how technology could interface with the world in more tangible ways. I pursued my education in electrical engineering at UCLA, where I worked on embedded sensing systems for health applications. This included projects like using sensors on shoes to improve the running form of Olympic athletes. During this time, I also took MBA classes to better understand the business side of technology.
My professional journey began at Accenture, where I focused on workflow and business process management systems. As machine learning started integrating into these platforms, I raised my hand to do deeper in this emerging field. This led to roles where I became a senior architect for these systems. Eventually, I joined McKinsey to work on automation and machine learning applications for enterprises, which broadened my experience to include cloud operations and customer support systems. I am finalizing plans to lead growth and commercial at an AI research lab and, in the meantime, independently advising investors and startups on the AI research to commercialization journey, focusing on the intersection of product and GTM.Â
A: That's quite a diverse background. What excites you most about the current state of AI, and what are your biggest concerns?
J: What excites me most is the potential of AI to improve the standard of living for all of humanity. Even with the technology available today, we can significantly enhance various aspects of life. However, my biggest concern is the potential widening of the wealth gap, exacerbating socioeconomic stratification and diminishing the purchasing power of the middle class. While improving social good with AI is relatively straightforward, spreading these benefits equitably is incredibly challenging.
However, my biggest concern is the potential widening of the wealth gap, exacerbating socioeconomic stratification and diminishing the purchasing power of the middle class. While improving social good with AI is relatively straightforward, spreading these benefits equitably is incredibly challenging.
A: You mentioned the importance of businesses viewing their workforces as contributors to the AI feedback loop. Could you elaborate on that?
J: Certainly. The businesses that will lead in AI in the coming years will be those that see their employees as integral parts of the AI feedback loop. Workers have the context and judgment to act as the guiding world model for AI. Instead of using AI solely for automation and efficiency to reduce headcount, companies should deploy these technologies to enable workers to become contributors to AI. This approach not only enhances the AI systems but also allows employees to engage in more creative and meaningful work.
A: That's an interesting perspective. How do you view the AI replacement argument, which suggests AI will eliminate many jobs?
J: The AI replacement argument often fails to consider the future capabilities of humanity. It tends to focus on what will be eliminated rather than what new opportunities will arise. Historically, technological advancements have always led to new types of work and industries. The same will be true for AI. While certain tasks will be automated, this will free up people to focus on more creative and strategic activities that can drive business and societal improvements. In my years of consulting experience, I rarely walked into a situation where employees felt like they could get to every task in their backlog. The tasks which were typically deprioritized in backlogs were business intelligence and improvement. Whether it was individual contributors or management, repetitive tasks keeping the day-to-day operation running overrode these creative and strategic activities.
A: Let's talk about AI investment. There has been a surge in valuations for AI startups. Do you think these valuations are sustainable?
J: The surge in valuations reflects the venture capital strategy of placing numerous bets, expecting a few to yield massive returns while many will fail. The Internet bubble is the most prominent, recent reminder that not every high-valuation company will succeed. However, those that do can define new markets. The key is responsible investment and managing growth without inflating valuations prematurely.
A: You’ve mentioned the concept of value density in the AI market. Can you explain what you mean by that?
J: Value density refers to the concentration of significant value among a few players at the top segment of the AI market. The most familiar of these is what most people refer to as SOTA, or state-of-the-art, with dominant companies like OpenAI. However, there is also room for smaller, specialized firms to thrive if they clearly define their value proposition and target customer in a way the dominant companies will not pose a serious threat since it would divert focus from their critical strategy. These smaller companies might be acquired to build function-specific models or to create specialized products. It's not a winner-takes-all market; instead, there are opportunities across different segments for substantial business growth.
A: Given the rapid advancements in AI, what skills do you think are essential for the next generation?
J: Resiliency and curiosity are crucial. Resiliency helps individuals adapt to rapid changes and navigate through uncertainties, while curiosity drives continuous learning and experimentation. These traits are foundational for success in any future work environment shaped by AI and other emerging technologies.
A: Lastly, how do you see the balance between open-source and proprietary AI models evolving?
J: The term "open-source" in AI can be misleading. There's open-weight models like Meta’s Llama, open data from platforms like Hugging Face, and truly open initiatives like BLOOM, which provide data, training code, and weights. Anyone who claims one approach is better in every situation is misleading you. Both will play roles with different pros and cons to each. A hybrid approach may become the norm, where foundational models are developed by a few, while specialized applications are built by many. The market will dictate the balance between open and proprietary models.
A: Thank you for your insights, John. It's been a fascinating discussion, and I look forward to seeing how these themes play out in the future of AI.