Knowledge

Between MACH and Magic: Max Welling shares his personal journey, motivation and outlook on the future of AI.

Prof. Dr. Max Welling, a heavyweight in Machine Learning, shares insights into his motivation and personal journey from physics to machine learning and academia to industry. Furthermore he and Dr. Ingmar Schuster discussed some of the emerging trends and application areas of AI.

A rundown of Max’s career reveals a diverse path – from a Chair in Machine Learning at the University of Amsterdam to a Scientist at Microsoft Research AI4Science. He’s been a fellow at CIFAR and ELLIS, served as VP at Qualcomm Technologies, and held academic roles at UC Irvine. Contributions like being an editor at IEEE TPAMI, Neurips advisory board member, and program chair at AISTATS and ECCV earned him accolades, including the ECCV Koenderink Prize in 2010 and the ICML Test of Time award in 2021.

Machines, Molecules and ? 

As in every podcast episode we invite our guests to contribute a third M-word that complements our theme. Max brought two options: MACH and Magic. At first glance, they seem disparate, but Max soon connects them through his personal journey and motivations. — MACH, denoting the speed at which an aircraft exceeds the sound barrier, traces back to Max’s childhood aspiration of becoming a fighter pilot. Due to difficulties with the exams this dream remained unfulfilled, however his fascination for the mechanisms and mysteries of the universe remained, leading to the second word – Magic. Max is fascinated by the magic inherent in scientific phenomena, from Artificial Intelligence to Quantum Computing. Quantum mechanics is more than mere computation; it embodies an enigmatic puzzle waiting to be solved. The challenge lies not in practical applications but in interpretation, a realm characterized by perpetual paradoxes that defy understanding, according to Max. Concepts like the Many-Worlds Interpretation, where the universe splinters with each quantum event, or the idea of massive superposition, seem fantastical.

Navigating Setbacks: From Fighter Pilot Dreams to Scientific Triumphs

In Max’s journey, he initially aspired to be a fighter pilot – a dream that didn’t materialize. Despite the disappointment, he views this setback as a crucial episode that shaped his character. It taught him the value of resilience in navigating life’s unexpected turns. According to Max, overcoming setbacks requires embracing them as opportunities. It involves recognizing numerous alternatives and choosing the one that resonates the most. Ultimately, he is very fulfilled with the path he chose – transitioning from the dream of a fighter pilot to a fulfilling career as a scientist.

“During my entire youth I was sort of working towards becoming a fighter pilot, I was dreaming of it, so I was very frustrated when this did not become a reality. […]  That’s fine, you just go to the next opportunity, and there’s plenty of opportunities out there for you to pick. […] But I feel good about being a scientist, because I think you need to dream of achieving something or understanding something bigger than you.”

From Intuition to Proof: How Max’s mentors approached scientific problems and how it influenced his own work

When asked about the qualities he admires in his advisors and aims to incorporate into his own work, Max reflects on the distinctive approaches of Jeff Hinton and Gerhard Hoft. Contrary to placing heavy emphasis on extensive mathematical formalism, both Hinton and Hoft were known for working intuitively. Max highlights their unconventional practice of erasing mathematical equations from the board and replacing them with tangible, intuitive examples. According to Max, this intuitive approach, coupled with a profound belief in their ideas, characterizes truly successful scientists. Rather than starting with axioms and deducing implications, they commence with a belief about how the world operates and then reverse-engineer to find proof and construct a compelling case. Max underscores the vital role of intuition in scientific exploration, a quality exemplified by his esteemed advisors.

“I believe that truly successful scientists begin with intuition. They navigate problems by feeling their way through, guided by a steadfast conviction that the answers lie ahead. Then, they skillfully work backward, employing mathematical tools or other methods to substantiate their points. Both Jeff Hinton and Gerhard Hoft, in my opinion, were remarkably intuitive scientists.”

Expanding Frontiers: Max’s Journey from Physics to Machine Learning

Max’s impressive academic career started with the completion of his Ph.D. in physics, where he specialized in quantum gravity under the guidance of Nobel laureate Gerard ‘t Hooft at Utrecht University in 1998. As an author of over 250 peer-reviewed articles spanning machine learning, computer vision, statistics, and physics, he has significantly advanced research in these domains. Particularly noteworthy is Max’s co-invention of variational autoencoders (VAEs) in collaboration with Diederik P Kingma. Despite his academic success, Max realized the limited practical impact of his work, causing his transition to Machine Learning in 1998. While he appreciates the autonomy within academia in setting one’s own research agenda, he recognizes the substantial influence and resources available at a company like Microsoft, where he joined two years ago. He worked for the AI4Science initiative, which aimed to apply artificial intelligence to various applications in physics and chemistry. Max’s exploration of machine learning is consistently influenced by his physics background, creating a symbiotic relationship between the two disciplines.

“I started in physics, intellectually exciting but lacking practical impact, working on quantum gravity in two dimensions. It boiled down to writing papers with no tangible outcomes. Joining Microsoft allowed me to utilize A.I. for impactful applications in science.”

The Delicate Dance: Juggling Freedom and Impact in Max’s Career Reflection

Despite Max’s positive experience at Microsoft, deeming it the right move, he is currently contemplating his next steps. While he sees the AI4Science initiative as fantastic, he grapples with finding the optimal equilibrium between the freedom and cutting-edge research offered by academia and the execution power and influence present in the corporate sphere. Max is currently taking time to reconnect with his academic community, as he engages with graduating students, visited CERN, and participated in NeurIPS. Additionally, he plans a two-month research stint at Caltech through an American fellowship. Post this period, he intends to make a definitive decision on his next endeavor.

“I personally find it very important to stay connected to technology and content. And sometimes you get promoted into places where that connection diminishes […] I want to stay with my feet in the mud when it comes to technology.”

What remains the same is his strong belief and premise in creating impact. He prioritizes addressing pressing issues like climate change and exploring the potential of machine learning to mitigate its challenges. To hear more about this and Max’ perspective on AI trends, tune into the entire podcast episode.

 

Spotify – Season 1, Episode 9, Max Welling

Youtube – Season 1, Episode 9, Max Welling