While pharma C-suite executives find themselves increasingly seduced by the promise of "digital transformation," and especially by the idea of leveraging AI, the lived, on the ground reality within virtually all pharma R&D organizations couldn't be further removed. "It's about the data," argues Dr. Amy Abernethy, Principal Deputy Commissioner of the FDA and experienced health data wrangler (2017 photo). Novartis CEO Vas Narasimhan candidly alluded to this in January when he reflected on his company's digital transformation journey, and said, "The first thing we've learned is the importance of having outstanding data to actually base your ML on. In our own shop, we've been working on a few big projects, and we've had to spend most of the time just cleaning the data sets before you can even run the algorithm. That's taken us years just to clean the datasets. I think people underestimate how little clean data there is out there, and how hard it is to clean and link the data."

FDA Commissioner Dr. Scott Gottlieb's recently announced appointment of Dr. Amy Abernethy to serve as his first lieutenant perfectly exemplifies the sort of thinking that has enabled Gottlieb to earn nearly universal praise at a time of unprecedented partisanship. He has focused his attention on high-priority, immediate concerns like the targeting of e-cigarettes to children, the opioid epidemic, and the need for a more robust market in affordable generic medications, while also seeking to prepare the agency to engage with emerging technologies, from consumer wearables to applications of artificial intelligence, to expand the breadth of evidence generation and accelerate the speed of analysis. Amy Abernethy, MD, PhD, formerly Chief Medical Officer, Chief Scientific Officer, and Senior Vice President, Flatiron Health, and recently named Deputy Commissioner of the FDA.A. Abernethy Abernethy combines the pragmatic humanism of an oncologist (which she is) with the technological sophistication of someone who spent their teenage years attending math camp and programming computers for NASA – after originally learning math by helping her mother edit a nursing textbook, as she recently told Lisa Suennen and me on our Tech Tonics podcast. The key challenge that Dr. Abernethy has focused on throughout her career – first in academia, at Duke, and more recently in business – is how to close the gap between clinical practice and clinical research. While at Duke, it struck her as odd, if not absurd, that she'd see patients in her oncology practice on a Monday, and then they'd need to come back on Tuesday to participate in a clinical trial she was conducting, because the two activities were considered so distinct.

Decades in the making, Canadian research in fields such as neural networking and machine learning is now coming into its own, as centres like the Montreal Institute for Learning Algorithms and Toronto's Vector Institute for Artificial Intelligence are propelling vibrant ecosystems and startup communities, while government is doing its part through immigration policies geared towards attracting new talent and programs like the Pan-Canadian Artificial Intelligence Strategy, which will devote $125 million in federal dollars to AI research over the next five years. We've seen significant investment from venture capital, too, with the current boom in VC funding for Canadian tech companies having a lot to do with movements in the AI space. And as more and more companies start to incorporate AI technologies, the field keeps growing. But where are the AI investment winners, you ask? Here are five stocks that analysts say have significant upsides.

On Wednesday, Shopify announced it had closed a (U.S.) $500-million share offering. The company said the proceeds would be used to strengthen its balance sheet or support growth. Abernethy says he thinks this cash will be used for acquisitions. "We see this financing as providing Shopify with greatly increased flexibility to pursue its expanding range of growth opportunities both organically and, potentially, through acquisitions," he says. "In terms of acquisitions, we believe Shopify could accelerate its product roadmap through the acquisition of emerging technologies, particularly in the artificial intelligence (AI) space. We also see opportunities for Shopify to broaden its offering through the acquisition of growth platforms in adjacent markets, such as accounting software (for example, private companies such as Wave or Freshbooks with millions of small business users could be of interest), marketing automation solutions, and logistics for small businesses."

Assessing uncertainty is an important step towards ensuring the safety and reliability of machine learning systems. Existing uncertainty estimation techniques may fail when their modeling assumptions are not met, e.g. when the data distribution differs from the one seen at training time. Here, we propose techniques that assess a classification algorithm’s uncertainty via calibrated probabilities (i.e. probabilities that match empirical outcome frequencies in the long run) and which are guaranteed to be reliable (i.e. accurate and calibrated) on out-of-distribution input, including input generated by an adversary. This represents an extension of classical online learning that handles uncertainty in addition to guaranteeing accuracy under adversarial assumptions. We establish formal guarantees for our methods, and we validate them on two real-world problems: question answering and medical diagnosis from genomic data.

We consider the problem of identifying the most profitable product design from a finite set of candidates under unknown consumer preference. A standard approach to this problem follows a two-step strategy: First, estimate the preference of the consumer population, represented as a point in part-worth space, using an adaptive discrete-choice questionnaire. Second, integrate the estimated part-worth vector with engineering feasibility and cost models to determine the optimal design. In this work, we (1) demonstrate that accurate preference estimation is neither necessary nor sufficient for identifying the optimal design, (2) introduce a novel adaptive questionnaire that leverages knowledge about engineering feasibility and manufacturing costs to directly determine the optimal design, and (3) interpret product design in terms of a nonlinear segmentation of part-worth space, and use this interpretation to illuminate the intrinsic difficulty of optimal design in the presence of noisy questionnaire responses. We establish the superiority of the proposed approach using a well-documented optimal product design task. This study demonstrates how the identification of optimal product design can be accelerated by integrating marketing and manufacturing knowledge into the adaptive questionnaire.