Podcast

Full Episode Transcript:

Jon: Dr. Yael Gold-Zamir, the CEO of Embryonics, joins us today on the hello, Human podcast where we discuss the latest topics in artificial intelligence and how it’s being applied in the real world. I’m Jon Knisley, the host of hello, Human and a long-time technologist, helping companies navigate emerging digital solutions.

A big thanks to FortressIQ for sponsoring the program. Be sure to hit the subscribe button wherever you listen to podcasts. 

This episode is part of a special series on Women in AI that we are very excited about here at FortressIQ. Elizabeth Mitelman from our marketing team who’s been the key driver of this special series is participating in the session as well. 

In this episode, we are going to explore human intelligence versus machine intelligence. Computers have always been superior to humans when it comes to computational-heavy work. While managing and finding patterns in huge data sets is a cakewalk for a computer, humans can’t make it past the first few 100 rows manually before data overload sets in. Our brains are simply not wired to handle the data.

On the other side of the coin, expert decisions requiring judgment have traditionally been the primary domain for humans over computers, but that gap is narrowing. In finance, algo trading now drives 80% of the U.S. stock market. Even lawyers are not immune. A team of experienced contract lawyers were outclassed last year by an AI system. 

Healthcare is one of the next major frontiers for AI, and we have a true expert in the field to give us her insight today. Embryonics’ technology outperformed a panel of embryologists in predicting which embryos will result in pregnancy by nearly 20%. It also outperformed the human experts by nearly 30% in recommending which embryos not to use which can lead to significant cost savings and prevent miscarriages. Remarkable result in an industry that has essentially had the same success rate for 30 years. 

Welcome to the hello, Human, Yael. You’ve got an incredible story. This promises to be a great conversation. 

Yael: Hi. Thank you for having me. Thank you for inviting me. 

Elizabeth: Thank you so much for joining us for the final episode of hello, Human, Women in AI. We are so excited to have you here to show your experiences from all the way across the world. 

I’d like to begin with going back to the start. You grew up in an ultra-Orthodox community and broke the norms by being the first in your community to be accepted into medical school and later becoming a doctor. Would you be able to share more of your personal journey and how you found your interest in the medical field and later IVF?

Yael: Yes. As you mentioned, I grew up in a Jewish ultra-Orthodox community of Israel—the same part of the community, by the way. I went through a beautiful childhood. The meaning of growing up in our community is basically that you’re going through an independent and separate educational system. It is separate for girls and boys. 

I went through the traditional schooling system of our community. I would say that generally speaking, in the place I’m coming from, girls are mostly accepted to become mothers and housewives. Going to university is taboo especially for women. Just leaving the community and going to a place that is very far from the known and sort of protected environment of the community was really a taboo.

Then, I remember myself for some reason because it’s like asking someone why you fell in love with someone else. Not everything can be explained, but from a very young age I remember myself dreaming of becoming a doctor and admiring this field. 

My attraction to the medical field was because of a combination of two things and my personality. One part is really the fact that being a doctor, with all the scientific knowledge and generally all the knowledge that doctors have, is very attractive for me to learn and educate myself. All those topics—chemistry, physics, and biology—were very attractive for me. 

I also felt that it’s the best thing you can do because it’s not only knowledge and science, but also combining the human part. You’re actually using this knowledge. You’re not just acquiring this knowledge, but you are in a very direct way using this to help people and you work with people. 

For me, from a very young age, it felt like the best profession you can have and I really wanted to become a doctor. Of course, when I got to a certain age when you need to apply to university, I understood that I have a huge academic gap that I need to fulfill in order to get accepted to a medical school. 

I’ll make the long story short. About 11 years ago, I became the first woman or the first person in my community who got accepted to medical school in Israel, and I’m happy to say that I’m not the last one. 

For the last part of your question, my attraction to the fertility field, again, I really like the scientific part. It was very interesting. I was really fascinated about all these scientific advances and all the opportunities that are available for patients because of this amazing research and science that is behind fertility treatments. 

I think it’s also part of my personality and my culture. I grew up in a place and an environment where family means a lot and family really comes first. I also met many women who struggle with infertility, so I guess it’s a combination of nature and environment, I would say.

Jon: Thank you very much for that background. Just a fascinating story being the first one for your community to attend medical school. As a doctor, based on your training and experience in the hospital, you’ve obviously learned the various aspects that impact fertility, including understanding the processes and factors that go into the decision-making. 

My father was a physician and we talked a lot about medicine growing up. Once I started working in AI a bit, it really dawned on me that a lot of the diagnosis and treatment is essentially pattern recognition. Whereas an individual may see 30,000 patients in their career, your AI model is based on millions of anonymized patient records. How do you think about this issue of experience versus data? Is one more valuable than the other?

Yael: It’s a great question. I’ll just tell you my experience. Being a young doctor and researcher gave me an ability to look on the current workflow from a different angle and sometimes question it. 

On one hand, I was behind the scenes. I was there. I saw how decisions are made. I saw how the process is done from a practical point of view. I understood the science that stands behind it. In a few months, I understood that IVF was invented 40 years ago, but I understood that in many aspects, this industry is stuck in 1990. Things like the whole process are still manual. The way that decisions are made, there is a prominent factor of trial and error. 

The data that is available for each patient is a lot of data that’s accumulating for each patient, but practically, we’re still using the limited factors that are believed to be influential traditionally. We’re still relying on them though there are so many new factors that are added, so many other tests. 

Potentially, the right way is not just to look at one factor or a limited set of parameters. It’s better to be able to analyze all the parameters, not only in a temporal way, not only in one time point but track all of them. But still, everything is manual (100%) and the trial and error component is really a huge part of the process. 

In IVF, you have many decisions. You see it again and again. I understood that fertility treatments are based in a great science—there is no argument about this—but in terms of technology, the field is left behind. We need to acknowledge it, we need to accept it, and we need to see what we can do about this. 

This is what I ask myself, can I do something about this? Because it was just obvious that in terms of technology, there is a huge lag there. The fact that the success rate of fertility treatments are around 25%—the average success is around 25%—and the fact that this situation stagnated for many years—it’s both very low, but also stagnated for many years—it’s just a good proof that the current practice has reached its performance low.

All this information and thoughts together brought me to the point that I thought, what can I do about this, and started thinking in the direction of Embryonics. This is the way I saw it, working in a clinic. This is what brought me to take on Embryonics. 

Jon: Early in my technology career, I was involved in the online medical education program for a couple of years. I used to debate with the key opinion leaders who are brought into to present, whether you’d prefer to get diagnosed by a leader in the field, your average doctor, or an AI model. 

Now that I thought about it for years, maybe the answer is almost an augmented integrated approach that really couples human intelligence with machine intelligence. With that potentially more formal human-agent teaming, does that potentially give us the best outcome?

Yael: I fully agree about this. It’s interesting because we’re willing to accept human errors but we’re not willing to accept technological errors. It’s not the conversation today or not the issue today, but it’s an interesting point to think about. 

I agree with you. I think that AI or any other technologies are not going to replace doctors. I don’t think they need to replace doctors, except for analyzing huge data sets, and there are many other things the doctors can do. These tools can augment and optimize the performance of a doctor, but also gives him more time to talk to the patient, touch the patient, to organize his thoughts, and on other things that he needs to do. 

We need doctors to stay with us. The only thing is that they will have these AI systems that will enable them to perform better and to allocate more time for the human part of being a doctor not for only analyzing the information. I want to be in a world where every doctor that I meet is I know that I have to talk to your doctor because he uses all the new advanced technologies to help me but still, I want to talk with a doctor. I want to have a human to walk me through my condition. I think that this is the best combination of doctors and technology.

Jon: There’s obviously a good side of both humans and machines, but there’s the other side of the coin as well. Each of them have their negatives. I want to talk for a minute about bias. Obviously, there’s a fair amount in the technology media about fair and ethical AI. At the same time—and you could argue less reported—there are nearly 200 forms of cognitive biases that people, including doctors, are subjected to. It influences their decisions as well. As both a technologist and a physician, how do you reconcile the issue of bias?

Yael: As a doctor, I would say that I try to be aware of myself. If I have some biases on something or at someone, trying to be aware of it is the first and the most important step towards solving it. 

As a technologist, it’s even more important because whenever it comes to data, and obviously when you are building AI algorithms, you need to have high-quality data sets that you really can trust because you’re training your algorithm using these data sets. 

One of the things that we were quite lucky in Embryonics is that when we trained our algorithm, we didn’t use any intermediate endpoints. The way that we train our algorithms is always based on the real endpoint. If it’s pregnancy or actual implantation, that can be measured with blood test results, ultrasound tests, or things like this. We’re relying here on the labeling. Our labeling process is relying on actual endpoints that can be measured and not on human decisions. 

We are not trying to mimic human thinking. We’re not asking, what would the doctor decide on this patient, or what would the embryologist say on this embryo, and using it for the training set We’re not doing this. We’re looking at the data of the patient and then the endpoint is, what happened to this patient? Did she get pregnant, had any side effects, or things that are measurable? Only measurable endpoints are considered in becoming part of our labeling process.

Elizabeth: It’s great to see that you’ve been able to address bias within Embryonics. Throughout this series, we’ve seen how different individuals and organizations address and conquer bias within AI. You’ve never imagined yourself working at a startup, but here you are today, owning and running your own company. What was it that finally pushed you to step away and start Embryonics?

Yael: Yes, I never imagined myself building and running a company. It wasn’t part of my plans. The plan was completely different, but it was a combination of several things. 

First thing is when I understood the big pain of the current IVF industry, the real lack of technology, the potential of good technology, and the potential impact of good technology on the lives of patients. I thought to myself that being a doctor in a hospital or a clinic can potentially have impact in the lives of the patients that I’m seeing in my clinic or the patients that I’m meeting in the hospital, but it’s a quite a limited impact that you can have because it’s only the people you’re meeting.

But if I’m able to build a technology that can impact the lives of millions of women globally, my potential impact will become unlimited and a really significant influence in the lives of millions of women. I felt obliged to do this. I couldn’t just say, okay. It’s something that became an obligation. I knew I can have an impact in the lives of millions of women and I couldn’t just leave it to them to do it. 

It’s also a curiosity on the research side. I guess, also some courage to embark in a totally new journey and something you didn’t plan on doing. It came from a very inner place that again, I knew that I can help many people and I just felt I had to do it.

Elizabeth: That’s an amazing story to hear. It’s great that you thought that you needed to do such a thing. You use data and real numbers to better the lives of hundreds of women and families. What was once a 100% manual decision-making process was transformed by Embryonics to take into account influential moments and decisions. From the patient side, they now actually understand what their diagnosis and treatment is. 

My question is Embryonics is one example of using AI to better human lives. How else do you see AI better our daily lives? Do you have any examples of future projects you’d be interested in sharing?

Yael: Elizabeth, as you just mentioned, I fully agree that AI is taking over us. We’re not aware enough of how AI is basically everywhere today—in our transportation, in our shopping experiences, everywhere really, in the medical field. I think that the future is going to be more about collaboration between these different systems. And I’m seeing it as integrated AI systems that are currently more independent. AI is everywhere but is fragmented in a way because there is AI for every different industry and it’s fragmented. The future will be the integration of those different AI systems from different industries to make our lives even easier and things will be even more faster. 

If I’m thinking of the future of, specifically, the fertility industry, it’s going to be completely different from the current situation. I believe that in 5, 7, 10 years, this industry is going to be completely changed. The process is going to be done with robots controlled by AI systems. What we’re doing in Embryonics is a very simple definition of the future. I don’t have a more simple way to describe it.

Elizabeth: When we think of the future of AI, a robot is definitely something that many of us touch on, but it’ll be interesting to see how the future of AI transforms over the next few years.

For this series, we’ve asked every guest for a piece of advice for the next generation of women in AI and tech leaders. Is there any advice that you can share?

Yael: I don’t want to sound generic, but my advice would be don’t try to adjust yourself to the popular narratives of success stories. Just try to be yourself, keep your values with your dreams, and do it your own way. It doesn’t have to be the narrative of a success story that you read or heard about because there are many, many different ways to build a successful venture. Just try to be yourself and focus on yourself and your dreams.

Elizabeth: Be yourself, chase your own dreams, remember your values, and the rest will come. That is a fantastic piece of advice. Thank you so much for sharing and for being a part of our series.

Yael: Thank you so much, Elizabeth, for having me.

Elizabeth: Of course.