Verified by 
Artificial Intelligence
Can AI Predict Humanity's Future Events?
Diagnostic Robotics’ Kira Radinsky on artificial general intelligence and more.
Posted November 23, 2019
Can artificial intelligence (AI) predict future events? Successful serial entrepreneur, award-winning inventor, scientist, and technology innovator Kira Radinsky, Ph.D., has an expert’s point of view and first-hand experience to answer that question. She is a member of the United Nations Secretary-General’s high-level panel on digital cooperation that is chaired by Melinda Gates, Co-Chair of The Bill and Melinda Gates Foundation, and Jack Ma, the Executive Chairman of Alibaba Group.
Radinsky is the Co-Founder, Chairwoman, and Chief Technology Officer of Diagnostic Robotics, a health care artificial intelligence (AI) system with predictive analytics, with locations in Tel-Aviv and New York City. In early November 2019, she successfully raised $24 million in Series A venture capital financing led by Accelmed Ventures with other investors for the two-year-old technology startup.
Radinsky was named to the MIT Technology Review’s “35 Innovators Under 35” in 2013—one of the many international accolades she has received. Other awards that she has received include Forbes’ “30 under 30 Rising Stars of Enterprise Technology,” Times of Israel’s “Technology Pioneer Award,” Forbes’ “10 Female Founders To Watch,” and Technion’s “Innovation in Technology Prize.”
As a scientist, Radinsky has co-authored over 10 patents and has been a featured speaker at events such as TEDMed, TEDx, Google Research Talks, GE Global Research, IBM Research, MIT Innovator Talks, Microsoft ThinkNext, and the Wall Street Journal Executive conference. Radinsky has been featured in many international media, including WIRED, The Wall Street Journal, MIT Technology Review, Forbes, Time Magazine, TechCrunch, The New Yorker, IEEE Spectrum, The Jerusalem Post, Business Insider, Der Spiegel, and the BBC.
In November 2019, Radinsky participated in a panel discussion on the future of clinical care at the Exponential Medicine conference in San Diego, California, where she met up with The Future Brain on Psychology Today. The conversation has been edited for clarity and length.
“Everybody is talking about how you cannot predict the black swans, and etcetera,” said Radinsky. She views it as a philosophical question. “My personal belief is that everything is a pattern; we just don’t have the data to identify this pattern yet. That’s why I’m a big believer in getting more data and more sensory information.”
Radinsky was only 15 years old when she started her Bachelor of Science in the gifted program at Technion—Israel Institute of Technology, where she graduated Summa Cum Laude. While earning her Ph.D. in computer science at Technion, Radinsky created an artificial intelligence software that predicted major events with Eric Horvitz, Co-Director at Microsoft Research in Redmond, Washington. Their AI system predicted a cholera outbreak in Cuba, riots, the pricing of electronic products, and more high-impact global events.
In 2012, Radinsky cofounded SalesPredict, an AI analytics company, and served as its Chief Technology Officer. American multinational e-commerce giant eBay acquired SalesPredict in 2016, and Radinsky became eBay’s Chief Scientist in Israel and Director of Data Science.
Radinsky co-founded Diagnostic Robotics in 2017. She believes in designing AI systems to work together with health care professionals in order to enhance human capabilities.
“Let’s talk about digital health care,” said Radinsky. “A lot of the things that people are trying to do, I think, are overly futuristic. I didn’t see EKG’s replacing cardiologists.”
At Diagnostic Robotics, Radinsky is focused on using AI algorithms for health care predictions. She views AI as a technology tool to assist, not replace, doctors. “Especially in health care, people don’t want to be diagnosed by systems,” said Radinsky. “People want to be diagnosed by people. They want people who understand. And they want empathy. A system is not going to give this empathy—at least not now.”
Radinsky takes a grounded stance on what AI can and cannot do. “I think we’re not actually ready to do a completely automated system—now,” she said. “I’m not saying that we’re not going to be, but I just don’t see it happening in the different regulations that we have today and the different liabilities that we have to take.”
Regarding the current hype on artificial intelligence, she advocates a balanced view of artificial intelligence. “Too many warnings are going to prevent the many advances we need to make in AI,” cautions Radinsky.
She views AI as having gone through previous cycles of hype over the years. “The hype before deep learning was Bayesian learning,” said Radinsky. “And I think that everything is coming back. So, the new one is reinforcement learning. And I think deep learning is going to disappear a little bit. Again, it made huge breakthroughs right now in the computer vision space. But if you look, for example, how we are doing at natural language processing, which is what I focus on right now, I didn’t see huge breakthroughs right now.”
“At this point with computer vision systems, you can show them a banana, and they’re going to think it’s a car,” quipped Radinsky.
Radinsky points out that while deep learning has created progress, there are still areas for improvement, such as natural language processing (NLP). “Natural language is going through some of the revolutions that computer vision went through,” she said. “There were a lot of advances in the questioning and answering space. I think the next breakthrough will be a full understanding with word knowledge. Until today, most of the models looked for historical correlations, and then tried to predict—for example, given a question, generate an answer.”
Radinsky, an expert in the field of AI predictive data mining, advocates the need for systems to go beyond pattern recognition, towards identifying underlying causes. “What we’re currently very interested about is how to actually mimic causality, she said. “Because what a lot of machine learning models are doing today is identifying correlation, not causation.”
“My question is, how do we identify causalities?” posits Radinsky. “Because we’ve also identified a lot of spurious correlations—I’ll give you an example. We identified that people who were using liquid soap, as compared to those using regular soap, have lower blood pressure. And we all know that liquid soap is not going to treat hypertension. And the reason is, there is a correlation with age. People who are using liquid soap are just younger in that population.”
Radinsky is a proponent of causality networks. “I think it’s untapped,” she shared. “A lot of the things that I’m doing right now is building the causality graphs. But we’re just scratching the surface. We’re only at the beginning. We’re building causality networks for diseases, and diagnostics robotics we’re building this for symptoms and diseases.”
As for the future of artificial intelligence itself, Radinsky believes that artificial general intelligence (AGI) is achievable. “I’m a big believer that we’re going to get there,” she said. “I’m a big believer also; it’s not going to happen in the next couple of years. I had a bet with one of the professors who told me that we would have AGI in five years. It’s been 15 years, and I’m still waiting.”
Radinsky finds zero-shot learning (ZSL) interesting. ZSL is a machine learning process where algorithms can predict what an object is based solely on its description. It’s somewhat analogous to human intelligence when a child sees a Betta for the first time and correctly predicts that it is a fish by observing that it has gills and fins and is living underwater.
When asked if AI can predict the future, Radinsky replied, “I think the best way to predict the future is to create it. One of the things that we are doing right now is identifying the patterns, and when the patterns start, try to predict the next step. So, it can predict things that have a pattern. Random things? It’s a philosophical question. Do we even have random things? Or is it part of a pattern that we don’t have data for? So, if you believe there is no random thing and everything has a pattern, then AI can predict the future. We just need more data for that.”
Copyright © 2019 Cami Rosso All rights reserved.
