AI’s Coming Battle with Scientific Orthodoxy
How AI Will Reverse Science’s Rapid Decline
Doug Bryan
Book coming soon...
Book overview
A lot has been written about the potential downside of artificial intelligence (AI) but little about its possible upside. This three-part book argues that AI’s biggest benefit will be fixing science.
The first part covers fourteen AI case studies from the past 50 years such as credit card fraud detection, learning to play chess better than any human in two hours, and novel protein design. Short case studies make the material accessible to a broad, curious audience and illustrate AI’s rapid improvement.
The second part of the book makes the case for science’s grave decline. Productivity dropped 98% the past 100 years. Most published research is just plain false due to p-hacking and hype. In many fields there is a negative relationship between reproducibility and citations where irreproducible work gets more references. Mobility of both elite scientists and most-cited papers are down. Scientific canons have ossified. Validators of good science have failed. It’s dire. Four case studies demonstrate the problem: low-fat diets and heart disease, the genetics of depression, Alzheimer's research, and the US response to COVID-19.
The last part argues that science is worth fixing and shows how AI will do it. Science has been good for civilization. Life expectancy doubled the past 200 years while infant mortality dropped 98%. War and famine deaths are down 90%. Computation per dollar is up 200 million fold since 1980, enabling such marvels as Minecraft, mRNA vaccines, and photographs of black holes.
Higher productivity, fresh perspectives, better trans-disciplinary collaboration, and faster innovation can revitalize science. AI will provide them. Lastly, frameworks for AI value creation in science are given with examples.
Along the way, striking similarities are drawn between the godfather of modern AI, Geoff Hinton, and the godfather of science, René Descartes. For example, both self-censored and both had to leave their home country to do their best work, Descartes from France to the Dutch Republic and Hinton from Britain to Canada. Other key figures discussed include James Watt, Albert Einstein, Ancel Keys, John Ioannidis, Demis Hassabis, and Anthony Fauci.
About the author
Doug Bryan is the academic grandson of John McCarthy, who coined the term “artificial intelligence” 70 years ago. Doug worked 12 years for Stanford University professor David Luckham, who earned his Ph.D. at MIT with McCarthy. That grounding in AI’s origins shaped a career spanning five technology waves.
The first was personal computing. When Bryan joined Stanford’s research staff in the early 1980s, computing was done on refrigerator-size machines shared by dozens of people. He installed and managed the first network of personal workstations on campus, an architecture that quickly became the norm.
The second was the internet. Yahoo and many others were spinoffs of his lab. He left academia for Accenture Labs to do R&D on the third wave: e-commerce. His team’s prototypes showed retailers what unlimited shelf space and catalog pages could mean, which led him to Amazon.com. There he was technical lead and product manager for personalized product recommendations that generated over $2 billion in incremental revenue a year using predictive models for 14 million products and millions of shoppers. That convinced Bryan that machine learning could be fully automated.
Next were startups in the fourth wave: automated machine learning (AutoML), which builds high-quality AI models without human guidance. He helped hundreds of companies double their AI ROI. Today's large language models are his fifth wave, and his experience with the prior four helps readers to succeed with AI.
Bryan’s clients have included Wells Fargo, PNC Bank, Washington Mutual, Ralph Lauren, NBCUniversal, Hearst, Rogers, Motorola, Otto, L.L.Bean, Target, and the US Department of Defense. He has published three computer science textbooks and dozens of peer-reviewed articles. His LinkedIn recommendations include:
“He excels at taking out complexity and makes things simpler.”
“A rare combination of sound technical knowledge and practical business judgment... ability to communicate complicated material simply.”
“Rare combination of client facing skills, experience, intellectual rigor, imaginative problem solving and value focus.”