Why Daniel Kahneman Is Really Excited About AI
Happy Wednesday. Those of you who read the Fortune CEO Daily—penned each morning at an ungodly hour by my boss, Alan Murray—got a taste of what seems like a fascinating panel on artificial intelligence at Davos. (If you don’t already subscribe, you really should—Here’s the sign-up link.)
The panel included Mustafa Suleyman, co-founder of DeepMind (gobbled up for a song by Google in 2014), Microsoft CEO Satya Nadella, Dow CEO Andrew Liveris, and the cognitively impressive head of IBM Watson, David Kenny, whom I interviewed at length in October. You can find Alan’s full post from this morning here.
As it happens, I too was planning to write about AI today—after hearing the brilliant Daniel Kahneman speak at an informal salon arranged by Zachary Todd and Saumitra Thakur, who have set up a forward thinking network called College-100.
Kahneman, who won the Nobel Prize in economic sciences in 2002 and who wrote the runaway bestseller and book-club fave Thinking, Fast and Slow, has elucidated a masterful theory of human decision-making and judgment—which appears to operate under two interlocking systems. “System 1,” in Kahneman’s model, is the gut-reactor of the mind, twitch-quick and intuitive. This is our “fast thinking” machine. As the psychologist explains in the intro to the book above: “Fast thinking includes both variants of intuitive thought—the expert and the heuristic—as well as the entirely automatic mental activities of perception and memory, the operations that enable you to know there is a lamp on your desk or retrieve the name of the capital of Russia.”
System 2 is slower, more deliberative, more analytical, more, well… thinking-like.
On Sunday, Kahneman recalled one of the recent milestones in AI achievement—the stunning defeat of the world’s foremost Go champion last March by DeepMind’s AlphaGo. (My former colleague Roger Parloff wrote a truly marvelous feature on the machine-learning science that led up to this seminal event, which I urge you all to read.)
What made Kahneman marvel was that the feat was a victory, in some respects, in the more mysterious half of the bifurcated kingdom of cognition. The ancient Go is not just infinitely more complicated than chess, he said, it is also “a game of intuition. It’s far less a game of rules than chess is.”
“In a way, Go is a game of pure skill in the way that I use System 1 to describe—and artificial intelligence seems to be mimicking intuitive skills in System 1,” said Kahneman. And it’s doing that not just in games, but also in an increasing number of real-world venues of decision-making. We’re already seeing that in medical diagnosis. And more and more, Kahneman predicted, we’re going to see that in business across-the-board. So, for example, “when you present the proposition for a merger or for an acquisition, you will have software” to help you make the judgment.
“And it’s System 1,” he repeated in emphasis, “not System 2—the deliberative system, which works more on logic. That is much easier to program than programming intuition. So the direction of AI is remarkable.”
At the time of AlphaGo’s victory, Google cofounder Sergey Brin said the DeepMind team had instilled “a level of beauty inside a computer.”
Hmm. Beauty in judgment. What a wonderful notion.
Penn Medicine used Apple's ResearchKit to build a rare disease app. A team of researchers from the University of Pennsylvania School of Medicine have launched a new app using Apple's special medical tech platform that they hope can provider critical new insights into patients suffering from the rare disease sarcoidosis, an inflammatory condition that creates lumps of cells across the body. The iPhone app will send participating patients information about dealing with sarcoidosis and can even direct them to local medical specialists; but the broader purpose is to glean more precise data about patients' day-to-day lives, such as when the disease flares up and how the condition affects them personally. For instance, using the iPhone's sensors, the ResearchKit app could compile data such as geographic location and weather patterns to see whether or not these may be related to flare-ups. "There’s a great opportunity that has never been done," said Penn Medicine assistant professor Dr. Misha Rosenbach in a statement. "In traditional research, you can’t see patients every day, but in app-based research you can suddenly get all this information about the disease in real-time and over time, from many different patients all over the world. It gives us the power to do sarcoidosis research in a way that has never been done."
Stanford is launching a new Center for Digital Health. The Stanford University School of Medicine is taking advantage of its Silicon Valley proximity to launch a major new digital health center that aims to connect tech companies with Stanford staff. "Digital health is a space where Stanford should be leading the way," said Dr. Sumbul Desai, clinical associate professor of medicine and executive director of the Stanford Center for Digital Health, in a statement. "The new center will be focused on leveraging our resources and encouraging collaborations that will lead to better health care through digital technology." Desai went on to note that tech companies often contact Stanford faculty in search of partnerships and it simply made sense to create a dedicated center to set up these connections. "We can help connect interested faculty with industry, or vice versa. Say, for example, there’s a faculty member interested in pulmonary digital health research. We may know a company with the same interest. We can help connect them," Desai said.
Theranos reportedly fails another round of inspections. Another day, another Wall Street Journal scoop on the continued trials and tribulations of Theranos. The Journal cites sources claiming that Elizabeth Holmes' blood testing upstart failed a second inspection of a laboratory, this one in Scottsdale, Arizona, in September, just before the company announced that it would be departing the blood testing business altogether. The sources also said that the Centers for Medicare and Medicaid Services (CMS) inspection found "deficiencies" that may lead to additional sanctions beyond those levied on Theranos after an earlier failed inspection at a California lab. "After months of careful consideration, and prior to CMS’s unannounced inspection in Arizona, Theranos decided to close its laboratories," the company's legal counsel told the Journal in a statement. (Wall Street Journal)
Lilly scoops up a late-stage migraine drug in $960 million deal for CoLucid. Fresh new Eli Lilly CEO David Ricks has made his first purchase, snapping up pain drug maker CoLucid for $960 million. Lilly is trying to elbow its way into a growing field of next-gen experimental migraine therapies, including CoLucid's late-stage candidate lasmiditan (which has shown promising results in phase 3 trials). From a pipeline perspective, this makes sense for Lilly after a year which saw the high-profile late-stage failure of Alzheimer's hopeful solanezumab.
Study: Trial researchers who get biopharma cash tend to produce more positive studies. A new study published in the BMJ finds that financial ties between clinical trial investigators and may have a correlation with more positive results in those trials. The authors warn that there have been conflicting analyses of this very issue; however, in a random sampling of 195 drug trials, the investigators found that 58% received some sort of compensation from biopharma companies, and that "trials authored by principal investigators with financial ties to drug manufacturers were more likely than other trials to report favourable results." (BMJ)
The biosimilar regulatory tussle is headed to SCOTUS. Tuesday featured two big pieces of news on the biosimilars front: the FDA finally issued guidance on what would be necessary to establish interchangeability between a branded biologic drug and its biosimilar copycat; and the Supreme Court agreed to hear the legal "patent dance" spat between Novartis generic unit Sandoz and biotech giant Amgen. The FDA recommendations include recommendations that drug makers developing biosimilars conduct at least one or more "switching" studies that show their experimental products are sufficiently similar to the original branded versions. As for the SCOTUS case? At issue is when, exactly, biosimilar manufacturers have to give branded drug makers a six-month notice for getting a biosim to market. Companies like Amgen have argued that this waiting period should only be invoked once a biosimilar has been approved; the generic copycat makers think they should be able to do so earlier in the process.
THE BIG PICTURE
Tom Price begins Senate confirmation hearing beneath a cloud of controversy. Congressman Tom Price's HHS Secretary confirmation hearing before the Senate Health, Education, Labor, and Pensions (HELP) committee began at 10 a.m. this morning. And the pointed questions facing the nominee about his staunchly conservative views on health policy and a number of potentially problematic investments underscore that the confirmation may not be a 100% guarantee. Price has already gotten pointed questions on his various stock investments in biopharma companies like Zimmer Biomet and Innate Immunotherapeutics, which seemed to occur awfully close to when he was pushing various pieces of legislation that would benefit the industry and have raised the specter of potential STOCK Act violations. Several senators also (naturally) pressed him on the Affordable Care Act and health entitlements like Medicare and Medicaid - programs that Price has advocated for privatizing or gutting in various ways. Stay tuned for my writeup on the biggest moments from the hearing. (Fortune)
Repealing Obamacare without a replacement would explode premiums and the uninsurance rate. Speaking of Obamacare: a new report from the independent Congressional Budget Office (CBO) underscores the costs of dismantling the law without coming up with a proper replacement. CBO estimates that nixing Obamacare's mandates and subsidies for buying health insurance while also preserving its protections for Americans with pre-existing conditions would lead to 32 million more uninsured Americans by 2026 and 18 million more uninsured in just one year. Furthermore, individual insurance plan sellers would double their premiums over the next decade thanks to adverse selection problems in the market, according to the analysis. GOP leaders dismissed the report, arguing that it fails to take an eventual replacement plan into account. (Fortune)
A Senator Wants Almond Milk to Stop Calling Itself Milk, by Kate Samuelson
Donald Trump to GOP on Corporate Taxes: Keep It Simple, Stupid, by Shawn Tully
Will This Be the Year VR Finally Goes Mainstream?, by Erin Griffith
Sheryl Sandberg Says This Is the Single Greatest Threat to Gender Equality, by Claire Zillman
|Produced by Sy Mukherjee|
Find past coverage. Sign up for other Fortune newsletters.