Front Pages and Complex Problems
On Friday, June 9, 2006, the front page of the Financial Times had a picture of a lifeless terrorist, al-Zarqawi (see here). I remember the newspaper folded at the end of the bed. Juan Santiago and I were watching the television when the doctor entered the room and delivered the news. A pathologist had just extracted a piece of bone from Santi’s knee and inspected it quickly. Even without the full review of a lab, it was possible to see it was malignant. Santi had cancer. I began crying while he was still in my arms, sitting in between my legs, both looking in the direction of the television. I hugged him tight in an attempt to protect him. He asked me why I was laughing. He couldn’t see my tears. He felt I was shaking. He couldn’t feel the dagger the news plunged into my heart.
Machine Learning, Computations and Cancer
Santi’s cancer most likely originated in a tumor above his right lung, a tumor that was detected a year earlier in a routine X-ray. The tumor was left untouched by the decision of a doctor based on literature, best practices, and his experience in the field. Many medical decisions are made based on best practices, publications, and expertise. Many decisions are correct. Others are wrong. You can use statistics to understand if the system works, and the percentages will likely show it does. But a child is not the system. Your child doesn’t experience the average of the outcomes. Your child either has or doesn’t have cancer. Your child survives cancer or dies. My son had cancer, and three years and nine months later, he died.
In recent years machine learning has made tremendous advances using images to detect breast cancer (read here, and here). The algorithms can detect small spots in images that humans cannot due to an array of image analyses that can only be accomplished with computers. Not just higher definition than the naked eye, but mathematical computations of many different kinds are deployed to analyze thousands of images that result in a later diagnosis of breast cancer. The improvement is substantial.
As the father of a child who died of cancer, I will always wonder if that technology could have been used to make a better decision in my son’s case. Perhaps not only an algorithm could find a spot of malignant cells in that tumor a year earlier, perhaps the case could have helped in the early diagnosis of other children and saved their lives as well. I wonder if the next generation of algorithms will have a systemic and collaborative aspect to them, and if coupled with the right incentives they could change not only cancer diagnosis but many other areas to improve the quality of the human experience.
Collaboration vs. Second Opinion
Imagine that while one algorithm inspects images and detects patterns that humans cannot, another algorithm consumes the contextual information, the meta-data if you will, and finds that there are other professionals that should be involved. If someone would have shown the images of my son’s scans to doctors familiar with the tumor that ended up metastasizing, they could have offered another point of view. It goes beyond second opinions because usually, second opinions are with similar subject matter experts. What is missing is a type of logic and analysis only a machine can do finding relationships across fields and silos, finding patterns beyond pre-established human structures, and bureaucracy.
Eventually, my son’s tumor was extracted by a childhood cancer specialist that simply grabbed it and pulled softly. That specialist operated in a different hospital, only that type of tumor. Finding him required arriving at that hospital with stage four, fully metastatic cancer, many rounds of chemo later. We had to travel through hell, spend weeks at the hospital, and experience emotional and physical devastation similar to torture. My son’s chances of survival were dismal at stage four compared to stage three, which likely was the stage he was a year earlier without metastasis. The doctors, knowledge, skill, literature, and understanding were all there, in a different hospital, in another city, reachable with one phone call, one email, still far away in another galaxy of knowledge, another silo of information.
Additionally, imagine my wife and I would have been involved in the process and understood the whole picture, not just presented with one expert’s opinion as to the only option, but educated using a graph of related subjects on tumors like my son’s. Imagine doctors in all fields are incentivized to share, expose, tell, and spread their information on a case. Or, perhaps patients who have the tools at their disposal share the knowledge and are assisted by technology that offers different routes. And other people who could take a look, again, not a second opinion, but a collaborative view that takes a different look at the same case.
Funding the Right Innovation
Years later, I went through a stash of old magazines in a house my family and I rented to spend the first vacation without Santi. I found an MIT Technology Review with Buzz Aldrin on its cover, saying, “You promised me Mars colonies. Instead, I got Facebook.” The subtitle is even sharper, “We’ve stopped solving big problems. Meet the technologists who refuse to give up.” The author, Jason Pontin, has a very interesting analysis of the economic incentives behind funding innovation and states what has been obvious to many of us for a long time. If we build economic systems that prize the fast, superfluous, and trivial, such as short term profits, big wins for a few, tax breaks for some, and so on, we shouldn’t be surprised we get bubbles, increasing wealth gaps, and investment in some places and not others.
Today, the case for deploying capital and technology behind the hardest of problems remains as relevant as a decade ago. Aldrin’s criticism of social media platforms is stronger after an attempt against America’s democracy was organized on Facebook and Twitter. But, we also have electric cars with better driving ranges and assisted driving that is beginning to prevent serious accidents. Perhaps the development of vaccines to tame a global pandemic in record time may have been the biggest win of all. New technologies will continue to emerge and we can’t expect to only have won. What I consider trivial, is hilarious to someone else. What I consider a step in the right direction on safe driving, others consider insufficient.
But, what is very clear is that under the right incentives and applications of technology, more people can have access to the tools and knowledge that my son -and countless others- didn’t have. That day in June of 2006, several other families got similar news. Statistics say one every three minutes. Yet, not one of the lives devastated by a cancer diagnosis rises to the level of the death of a known terrorist. Individual suffering is less newsworthy than the fear that a few individuals can impose on all of us. Rightly so. Moreover, when we improve the mix of incentives and technologies, there will be more front pages with technological breakthroughs that solve complex problems for all humanity and fewer pages with dead terrorists.
Originally published at https://www.linkedin.com.
