We’re back at FutureMed for Day 2. Yesterday felt packed with only three speakers, but Day 2 featured 12 experts spread out over three themes: Introduction to Exponentials, Data-Driven Health, and the Future of Oncology. As an aside and quick reminder, feel free to learn more about each of the speakers by visiting their faculty pages on FutureMed’s website: futuremed2020.com/faculty.
The day kicked off with Neil Jacobstein’s talk, “Artificial Intelligence and the Future of Medicine.” He began by sharing the story of Archimedes who said, to paraphrase, “give me a place to stand and a lever long enough and I can move the world.” Jacobstein compared AI to the fulcrum upon which most future technologies – synthetic biology, genomics, nanotechnology, etc – will accomplish their Earth-moving consequences. The majority of the talk focused on providing examples for this bold claim: from AI implementations that can identify tissue pathology to Google’s self-driving car to, and, as we would hear about later, IBM’s Watson. A great take-away quote:
“The brain hasn’t had a major upgrade in over 50,000 years. If your laptop hadn’t had an update in five years, you’d be upset about that.”
Will artificial intelligence be that upgrade? Many of those at Singularity University certainly believe so.
Next up was Dan Berry, who is probably SU’s favorite astronaut and for good reason. He gave an inspirational talk about how he dreamed about spaceflight and thought about it every day for over a decade, which motivated him to work non-stop to finally accomplish that dream. What is he up to nowadays? One word: robots. He spent much of his talk on the subject of when machines would truly be “intelligent,” and, though it was early in the morning, the audience seemed to appreciate his discussion of Descarte (“I think therefore I am”) and why he didn’t get it all right. He rephrased that famous statement: “I move, therefore I perceive. I perceive, therefore I emote. I emote, therefore I have intellect.” Now, he continued, we’ve reached the stage where ubiquitous perception is cheaply and ubiquitously available for machines, such as the “trivial” toy helicopters that contain gyroscopes that just four years ago would have cost too much to sell them as toys. Berry predicts that this transition of increased perception is a harbinger for the coming ability of truly intelligent robots.
Berry was followed by Brad Templeton, a software engineer at Google who focuses on the automated car project. Though Templeton was wearing a Google Glass, it was unclear whether that was serving as his teleprompter (probably not, because he gave what seemed to be an unscripted, though entertaining talk). The first half of his talk focused on innovation in the age of the Internet and how we now have a battle between six primary competitive platforms: Apple (controlled platform), Microsoft (less controlled monopoly platform), Google (giant gateway to the web), Facebook (one site to rule them all), Amazon (one big cloud/store), and Linux (the vast, usually interoperable bazaar). The latter part of his talk transitioned to showing the FutureMed participants cool footage from the automated car and providing commentary along the way. Not that medically-related (e.g. no discussion of automated surgical robots, though when we asked Intuitive’s CMO Catherine Mohr about this she said automated surgical robots would be more difficult because of the more complex logic trees), but cool nonetheless.
The final talk of the morning was by Salim Ismail, who drew an extended analogy between transformations we are seeing in society and the physical state changes that water goes through (ice to water to steam). This mirrored yesterday’s talk by Peter Diamandis who discussed the “dematerialization” that exponential growth makes possible. Some examples of ice-to-steam transitions: physical computers become cloud computing, physical goods and money become the attention economy, and clans and tribes in society become Facebook and LinkedIn groups. One caveat to be aware of is that the transition to the virtual state (too much steam) can result in instability, as we saw with movements such as the Arab Spring and Tea Party.
After the lunch break we switched gears to Data-Driven Heath and began with a talk by the person The Atlantic called “The Most Measured Man in the World” – Larry Smarr. His talk mirrored the article pretty closely, so we won’t spend much time describing it here. In brief Smarr is a computer scientist who has measured over 1 billion data points about himself, ranging from the usual indicators (weight, heart rate, etc) to less mainstream health signs (blood serum levels, stool composition). In doing so he was able to detect that, though he was asymptomatic, his inflammation markers were highly elevated. When he showed this to his physician he was dismissed repeatedly, but soon thereafter he became symptomatic and was discovered to have Crohn’s disease. The talk focused on how he leveraged everything from his SNP data from 23andMe to daily symptom trackers to even stool microbial diversity analysis to “hack” his disease. When an audience member asked about whether this type of tracking would be “information overload to the average patient,” he responded that the Quantified Self movement should be considered the early adopters.
“Don’t confuse early adopters with what the mainstream is going to look like….This year is going to be the year of the integrated portal, but with a user interface that allows consumers to make better lifestyle choices. Think of it as your body’s instrument panel.”
Next up was John Mattison, the Chief Medical Information Officer of Kaiser Permanente, who gave a talk titled “Strategic opportunities for transforming healthcare: Surfing (Taming) the tsunami of exponentiality.” He began by discussing some of the successes of Kaiser Permanente’s HealthConnect system which covers 20,000 MDs and 9 million patients and sends more than 25,000 e-mails each day. Most impressively, they estimate that they save 17,000 lives each decade simply due to improved preventive care. The example he gave was of a woman who visited her optometrist for an eye check-up and prescription. The optometrist noticed on the HealthConnect system that the patient had not had a her recommended mammogram and encouraged her to do so. She got the mammogram and was found to have cancer, though at an early enough stage that she could be fully treated. However, Mattison continued, according to a recent RAND study the implementation of health IT has not been happening quickly enough and still faces problems such as interoperability and data sharing. To date, IT has actually increased expenses along with five key drivers of healthcare costs:
- Disorders of lifestyle, such as obesity and diabetes (as an aside, Mattison said “information therapy” has failed us – consumers don’t change their behavior despite knowing about high calorie counts and dangers of smoking)
- Failure to implement evidence-based preventive care
- Fee-for-Volume payment incentives
- End-of-Life care (5% of inpatients account for 50% of costs)
- The pervasive influence of pharma
He then proceeded to discuss how big data could help lower these cost drivers by posing eight questions, such as what the role is of big data and multi-omics in addressing lifestyle disorders.
Mattison was followed by Christopher Longhurst, his CMIO counterpart at Stanford’s Packard Children’s Hospital. Longhurst gave an overview of how his hospital went digital, which included a failed stint trying to implement Google Health. He made a couple of predictions, including two that resonated:
- Personal health records will evolve into personal health advisors
- Evidence-based practice will become practice-based evidence. For this one he gave the example of a 13-year old girl with lupus and a coagulopathy. Her physician was deciding whether to transfuse her and consulted colleagues who were “experts” who had seen similar patients. She received polar opposite advice and thus decided to consult Packard’s aggregate health records, found all the lupus patients with coagulopathies who had been transfused, and based her (positive) decision off of the collective experience.
At this point we’ve heard from individuals and CMIOs who are leveraging data to improve their personal and patient health, respectively.The next speaker, Daniel Riskin (CEO of Health Fidelity), pointed out that while information technology has improved our access to healthcare data, our workflow has not changed much in over a century. One reason for this is that healthcare data is unstructured. Riskin discussed how his company is using natural language processing (NLP) to extract context for applications such as “subgroup analytics” – the automated extraction and analysis of data from, say, 60-70 year old white males discharged with heart attacks who were readmitted to understand what went wrong.
Following a brief talk by Lark’s founder, Julia Hu, and a break for demos (including a really cool Arduino-based lab run by Dan Berry – see image), we returned to the topic of data-driven healthcare. Marty Kohn, IBM’s Chief Medical Scientist for Care Delivery Systems (aka one of the Drs. behind “Dr. Watson”), gave a fascinating presentation on how they foresee Watson being used in healthcare. He began by discussing the essential challenge that Watson accomplished: language comprehension including subtleties, nuances, and idioms (“How do you teach a computer that nose can run and feet can smell?”). Watson has three fundamental features: (1) understanding natural language and human communication, (2) generating and evaluating evidence-based hypothesis, and (3) adapting and learning from user responses.
Are these characteristics enough to replace doctors? As you may recall, after Watson won Jeopardy and IBM said they were interested in healthcare applications, the media went wild and began claiming that Dr. Watson would obviate the need of physicians. Kohn clarified that IBM does not see Watson replacing physicians anytime soon. Unlike Jeopardy, healthcare is not deterministic; there is often not one right answer since many patients have multiple co-morbidities. Instead, Watson will be capable of “reading and digesting” hundreds of thousands of papers (IBM is partnering with medical journal publishers) to help physicians make decisions based off of the latest evidence-based guidelines. So, according to Kohn, medical students like this author will be safe from Dr. Watson and most likely improve their outcomes tremendously with the clinical-decision support.
The next speaker, however, begs to differ. As a successful entrepreneur and venture capitalist, Vinod Khosla made headlines recently by writing that technology will replace 80% of what doctors do. His presentation, entitled “2025: 20% doctors included?,” consisted of a series of statements that were humble in form (devoid of exclamatory punctuation or even capitalization) but massive in consequence. The first half consisted of statements describing the limitations of (human) doctors, namely cognitive bias and limitations (e.g. though Khosla admitted that this was not necessarily causal, the number of diagnoses of disorders such as ADHD is roughly correlated to the number of press mentions of that disorder – a phenomenon known as recency bias). Khosla then spent the next half discussing how machines would replace doctors starting with “Toddler MD” and “clumsy point innovations like alivecor, cellscope, adamant, ginger.io, neurotrek, consumer physics, jawbone, misfit, “insighted” by Ayasdi, leading us to discover things we never knew were right in front of us.” By 2025 these systems will be far more advanced. He concluded by admitting that he may be wrong about many specifics, but he is certain that directionally this prediction is correct. Furthermore, he said that people should not equate the failure of individual health tech start-ups with the failure of the trend as a whole:
“There were thousands of failed dot coms, but Google emerged to change how we live. No failure is consequential, but one success is consequential. The important thing is to take 1,000 shots on goal.”
His firm, Khosla Ventures, is certainly beginning to do so with its healthcare investments.
As you can tell, it was a packed day full of insights and brilliance. We were unable to stay to cover the final one hour session, but heard good things from the participants. The undeniable focus of day 2, however, was Data, Data, and more Data.
Join us tomorrow for coverage of Day 3!