In 1845, dentist Horace Wells stood before Harvard medical students and faculty, eager to demonstrate the utility of nitrous oxide – laughing gas – as a general anesthetic. 

Wells tried it out on a patient who needed  a tooth extraction. The dose, it turned out, wasn’t enough. The patient screamed in agony. 

As described by Paul Offit in You Bet Your Life (my 2021 WSJ review here), the demonstration elicited “peals of laughter from the audience, some of whom shouted, ‘Humbug!’ Wells left the building in disgrace.”

About a year and a half later, another dentist, Charles Morton, conducted a similar demonstration, using ether as the anesthetic instead.  In front of an  audience at an auditorium at the Massachusetts General Hospital (MGH), Morton excised a large tumor from the jaw of a 20-year-old housepainter named Gilbert Abbott. Abbott slept through the entire procedure. 

When the operation was complete, John Collins Warren, a professor of surgery at MGH who had hosted both demonstrations, “looked at the audience and declared, ‘Gentleman – this is no humbug.’” 

Today, the smartest and most skeptical academic experts I know are floored by a different emerging technology: generative AI. There seems to be a rush among healthtech investors to back startups leveraging AI to solve specific problems in biomedicine and healthcare. Meanwhile, incumbent biopharma and healthcare stakeholders are (or soon will be) urgently contemplating how and where to leverage generative AI – and where their own troves of unique data might be utilized to fine tune AI models and generate distinctive insight.

It seems time to declare, “Generative AI is no humbug.” 

Hope Beyond The Hype

I’ve started with this 19^th century story to remind us that physicians and scientists have always struggled to assess the promise of emerging technologies.

Today, the hype around generative AI is off the charts. “A New Area of A.I. Booms, Even Amid the Tech Gloom,” reads a recent New York Times headline. It continues, “An investment frenzy over ‘generative artificial intelligence’ has gripped Silicon Valley, as tools that generate text, images and sounds in response to short prompts seize the imagination.”

It’s reasonable to wonder whether this is just the latest shiny tech object that arrives with dazzling promise only to fizzle out, never meaningfully impacting the way care is delivered and the way drugs are discovered and developed.

So far, AI hasn’t really moved the needle in healthcare, as a remarkably blunt recent post from the Stanford Institute for Human-Centered AI (HAI) acknowledges. 

But I believe generative AI offers something different, and profound — a perspective shared by the Stanford HAI authors. Generative AI is an area with which we should (and arguably, must) engage deeply, rather than merely follow with detached, bemused interest.

Generative AI and Chat-GPT

What is generative AI? You can ask the AI itself. According to Chat-GPT, openAI’s wildly popular demonstration model of the technology, generative AI, “refers to a type of artificial intelligence that generates new data, such as text, images, or sound, based on a set of training data.” 

That explanation is well and good, but if you really want to viscerally appreciate some of the power of the technology, you really need to — and owe it to yourself to — experience it. Go to chat.openai.com, sign up for free, and try chat-GPT for yourself. It’s unbelievable in a way that you need to engage with to really understand. The specific examples always seem trivial, but the range and fluidity of the responses the technology provides is extraordinary.

For example, I asked it to write a commentary about climate change from the perspective of Bernie Sanders, and then another one from the perspective of Donald Trump – the results were uncanny. One of my teenage daughters, not easily impressed, was blown away when I asked the technology to “write a 200-word essay from perspective of teenage daughter asking dad to approve [a particular app],” a highly topical subject in our household. The result was fantastic, even persuasive.

Of course, the technology isn’t perfect, and certainly not infallible – for example when I asked it about the line “Is it safe yet?” in a Dustin Hoffman movie, it correctly identified both the film (“Marathon Man”) and Hoffman’s character, but incorrectly thought it was Hoffman’s line, rather than that of his interrogator, portrayed by Laurence Olivier. 

Such errors are not unusual and reflect a well-described challenge known as “hallucinations,” where the model confidently provides inaccurate information, often in the context of other information that’s accurate. 

In another example, discussed by Ben Thompson at Stratechery, the model is asked about the views of Thomas Hobbes. It generates a response that Thompson describes as “a confident answer, complete with supporting evidence and a citation to Hobbes work, and it is completely wrong,” confusing the arguments of Hobbes with those of John Locke.

Not surprisingly, healthcare AI experts tend to emphasize the role of “human in the loop” systems for high stakes situations like providing diagnoses. One framing I’ve heard a lot from AI enthusiasts is “you’re not going to be replaced by a computer – you’re going to be replaced by a person with a computer.”

Large Language Models and Emergence

The capabilities behind chat-GPT are driven by a category of model known as “Large Language Models,” or LLMs. The models are trained on as much coherent text as they can find to hoover up, and are designed to recognize words found in proximity to each other. 

A remarkable property of LLMs and other generative AI models is emergence: an ability that isn’t present in smaller models, but is present (and often arises, seeming abruptly) in larger models. 

As two authors of a recent paper on emergence in the context of LLMs explain,

“This new paradigm represents a shift from task-specific models, trained to do a single task, to task-general models, which can perform many tasks. Task-general models can even perform new tasks that were not explicitly included in their training data. For instance, GPT-3 showed that language models could successfully multiply two-digit numbers, even though they were not explicitly trained to do so. However, this ability to perform new tasks only occurred for models that had a certain number of parameters and were trained on a large-enough dataset.”

(If you’re first thought is that of Skynet becoming sentient, I’m with you.)

Models in this category are often termed “foundation models,” since they may be adapted to many applications (see this exceptional write-up in The Economist, and this associated podcast episode). While the training of the underlying model is generally both time-consuming and expensive, the adaptation of the model to a range of specific applications can be done with relative ease, requiring only modest additional tuning.

Implications for Healthcare and Biopharma

Foundational models represent a particularly attractive opportunity in healthcare, where there’s a “need to retrain every model for the specific patient population and hospital where it will be used,” which “creates cost, complexity, and personnel barriers to using AI,” as the Stanford HAI authors observe.

They continue:

”This is where foundation models can provide a mechanism for rapidly and inexpensively adapting models for local use. Rather than specializing in a single task, foundation models capture a wide breadth of knowledge from unlabeled data. Then, instead of training models from scratch, practitioners can adapt an existing foundation model, a process that requires substantially less labeled training data.”

Foundation models also offer the ability to combine multiple modalities during training. As Eric Topol writes in a recent, essential review (see also the many excellent references within). “Foundation models for medicine provide the potential for a diverse, integration of medical data that includes electronic health records, images, lab values, biologic layers such as the genome and gut microbiome, and social determinants of health.” 

At the same time, Topol acknowledges that the path forward is “not exactly clear or rapid.” Even so, he says, the opportunity to apply generative AI to a range of tasks in healthcare “would come in handy (an understatement).” (Readers interested in keeping up with advances in healthcare-related AI should consider subscribing to “Doctor Penguin,” a weekly update produced by Topol and colleagues.)

The question, of course, is how to get from here to there — not to mention envisioning and describing the “there.” 

The journey won’t be easy. The allure of applying tech to healthcare and drug discovery has been repeatedly, maddeningly thwarted by a range of challenges, particularly involving data: comparatively limited data volume (vs text on the internet, say), inconsistent data quality, data accessibility, and data privacy. Other obstacles include healthcare’s notorious perverse incentives and the perennial difficulty of reinventing processes in legacy organizations (how’s your latest digital transformation working out?).

As the seasoned tech experts at the “All In” podcast recently discussed, it’s not yet clear how the enormous models underlying generative AI will find impactful expression in startups – though the interest in figuring this out is enormous. One of the hosts suggested that the underlying AI itself was likely to become commoditized, or nearly commoditized; hence,

“the real advantage will come from applications that are able to get a hold of proprietary data sets and then use those proprietary data sets to generate insights, and then layering on … reinforcement learning.  If you can be the first out there in a given vertical with a proprietary data set, then you get the advantage, the moat of reinforcement learning. That would be the way to create, I think, a sustainable business.”

When you think about promising proprietary data sets, those that are owned or managed by healthcare organizations and biopharmaceutical companies certainly come to mind.

Healthtech Investors See An Opportunity

Perhaps not surprisingly, many healthtech experts are keen jump on these emerging opportunities through investments in AI-driven startups.

A new VC, Dimension, was recently launched with $350M in the bank, led by Nan Li (formerly a healthtech investor at Obvious Ventures), Adam Goulburn and Zavain Dar (both experienced healthtech investors joining from Lux Capital).  They’re focused on companies at the “interface of technology and the life sciences,” and looking “is looking for platform technologies that marry elements of biotech with computing.”  (TR coverage).

Healthtech and the promise of AI has also captured the attention of established biotech investors — it’s a key thesis of Noubar Afeyan’s Flagship Pioneering – and prominent tech VCs, like Andreessen Horowitz.  Generative AI informs the thinking of Vijay Pande, who leads Andreessen’s Bio Fund. 

Also focused on this interface: five emerging VC investors who collaborate on an thoughtful Substack focused on the evidence-based evaluation of advances (or putative advances) in Tech Bio, with a particular emphasis on AI. The contributors include Amee Kapadia, a biomedical engineer (Cantos Ventures); Morgan Cheatham, a data scientist and physician-in-training (Bessemer Venture Partners); Pablo Lubroth, a biochemical engineer and neuropharmacologist (Hummingbird Ventures); Patrick Malone, a physician-scientist (KdT Ventures); and Ketan Yerneni, a physician (also KdT Ventures).

Meanwhile, physician Ronny Hashmonay, recently announced on LinkedIn that he “is leaving Novartis, after 11.5 years,” and “is founding a new VC fund to continue working and leading the tech revolution in healthcare.”

Concluding Thoughts

It’s enormously exciting, if frequently disorienting, to participate in the installation phase of a new technology, the stage of technology development where the promise is recognized but the path to realization is less clear. Our challenge and opportunity is to help figure out how to translate, responsibly, the power and possibility associated with generative AI into tangible, meaningful benefit for patients and for science.

One final note: despite the ignominious demonstration of Horace Wells, both ether and nitrous oxide ultimately found widespread use as general anesthetics, along with chloroform. Significantly improved reagents and processes were developed, often incrementally, in the first half of the 20th century, and continuing forward. The progress in anesthetics over the last 150 years has been nothing short of remarkable. 

And yet, as Offit reminds us more than 175 years later, “the exact mechanism by which they work remains unknown.”

Sounds familiar.

For today: topics relevant to many drug developers (and others):

• The initial impact of digital
• The dilemma of data collection in early clinical studies
• The elusive search for “new normal” ways of working

Initial impact of digital in biopharma

The sexy promise of digital/data/AI in biopharma was that emerging digital technologies were going to solve our most important and vexing problems. At the most breathless moments, some asked whether IBM’s Watson would cure cancer.  

AI, generally speaking, has continued to struggle in areas where the underlying data are a mess and/or where it’s difficult to aggregate adequate amounts of reasonably reliable data. One example where such insight is needed but not currently available: translational models.

As I discussed in a recent column, drug developers desperately need improved translational models, improved abilities to predict which candidate molecules are safe and effective. This would increase the success rate of large and costly Phase 2 and Phase 3 trials. While AI has proved remarkably helpful in domains such as predicting  protein folding, it’s currently difficult to imagine that in the near future, AI will be able to anticipate reliably what will happen — good and bad — when a molecule is introduced into the human body.

On the other hand – as I described in 2021 — emerging digital technologies are having an immediate impact in the “digital transformation” of how pharma functions, similar to the way other businesses have been transformed. The new direction tends to emphasize large platforms, repeatable processes, and real-time dashboards that enable teams, as well as the broader organization, to align around shared data. In short: business operations, including R&D operations.

(I’ve previously emphasized the value of dashboards as a valuable management tool for Pfizer in the development of their COVID-19 vaccine; separately, I have also discussed the downside of excessive reliance on dashboards and metrics.)

The emphasis on operations in pharma closely parallels areas of near-term progress in healthcare, where much of the initial effort, and low-hanging fruit, seem to be in what physician and venture capitalist Vineeta Agarwala calls “the workflow and administration category,” and highlights the intensive activity she’s seen in the “revenue cycle management and billing world.”  (The comments are from a recent podcast, featuring Agarwala, fellow VC Vijay Pande, and physician and digital champion Eric Topol; the episode, here, is highly recommended.)

Why operations? For one, these data tend to be plentiful, easily obtainable, and readily analyzable. For another, these data are consequential in a fashion business leaders intuitively understand, particularly when focused on parameters like efficiency.

A more interesting question is whether these efforts will ultimately enable higher-level inquiry, or will just enable drugs to fail slightly faster and cheaper. The optimistic view – which I favor — is that eventually the process of digitization will enable the thoughtful secondary analysis of large volumes of laboratory and clinical data. That sort of analysis could catalyze the development of novel and effective medicines. In the short term, though, most of the impact is likely to be on more prosaic measures of process efficiency.

One final point: a key aspect of the digital vision for biopharma is to maximize the opportunity for the organization to learn from every byte of data collected. This mirrors the much-discussed and often elusive (see here) vision for a learning healthcare system, where the goal is for each patient to benefit from the knowledge gleaned from each previous patient and aggregated for all previous institutional experience.  (Dr. Agarwala nicely discusses this aspiration in a recent McKinsey interview, here.)

Historically, pharma companies have struggled to comprehensively capture and leverage the totality of their data and experience, as both data and expertise tend to reside in hard-to-locate organizational silos.  The digitization of the enterprise offers the hope that more wisdom will be shared, allowing more insights to emerge.

Dilemma of data collection in early clinical studies

As much as drug developers try to understand a new molecule in pre-clinical studies and models, the introduction of a compound into a human being for the first time represents a profoundly important event, and contributes to a step-change increase in our understanding of the molecule.

The initial (Phase 1) clinical studies are focused on ensuring the new molecule is safe and well-tolerated; the product is given to healthy subjects (Phase 1a) and volunteer patients (Phase 1b) in carefully escalating doses in a controlled and monitored research environment.

Phase 2 studies, generally somewhat larger, are focused on starting to figure out whether the new medicine is likely to be acting the way you hoped in the body – does it seem like it might be working on treating the disease?

Broadly speaking, there are two schools of thought regarding how to approach early phase studies. Many researchers see these trials as an opportunity (some would say obligation) to learn as much about the medicine and the human body as possible. New molecules, they point out, are ultimately “perturbogens,” designed to poke certain parts of certain cells in certain ways. By analyzing in as comprehensive a way as possible how the body responds, there is the opportunity for extensive knowledge capture and insight generation. Thus, some scientists want to include in early phase studies a number of additional assays and evaluations.

The other school of thought – generally adopted by the folks responsible for the conduct and operation of clinical studies – emphasizes the value of parsimony. 

They argue clinical studies should be (and must be) as simple as possible, with the minimal number of tests required to achieve the trial’s primary objective. Reducing complexity, they point out, increases both study recruitment (a simpler study with fewer evaluations is generally more convenient and less burdensome for subjects) and study conduct, since there are fewer ways for things to go wrong. The much-lauded RECOVERY study in COVID is frequently cited as highlighting the value of parsimony in trial design. 

The truth is that it’s a very difficult balance to strike, and if (like me) you’ve been around long enough, you’ve lived through the ongoing struggles as well as the consequences. I’ve seen clinical teams lamenting the omission of data elements that they subsequently wished they had, but I’ve also seen clinical study protocols (often drafted by recent academic emigrees) that are incredibly clever and refined and complex – and then prove to be impossible to recruit and execute. 

Clearly, an elaborately designed scientific  study with zero subjects isn’t particularly useful – the phrase “succès d’estime” springs to mind. At the same time, an extremely simple study that fails to capture valuable data may represent a critical missed opportunity to advance the science and accelerate therapeutic development for the benefit of patients.

My increasing concern here is that just as we’re gaining the ability to leverage and contextualize increasing amounts of multimodal data, especially critical for translational science, drug development organizations find themselves under enormous pressure to make studies as lean as possible.  Short-term performance metrics (the vital signs intensively scrutinized by most companies) are relentlessly focused on efficiency – on how fast and cheaply a study is conducted. 

The consequences of this emphasis – the missed opportunity to capture meaningful insight – is far more difficult to measure, and is appreciated (if at all) only far later in development.

To summarize, the short-term corporate incentives to drive efficiency tend to be more palpable than the incentives to resist them in favor of more extensive data collection. Presumably, if or when the value of richer data collection – and the pragmatic utility of and necessity for translational research and data – becomes more readily apparent, the balance might change, and measures viewed as optional today might become essential in the future.

Elusive search for “new normal” ways of working

Finally, a few thoughts on the “new normal.” Despite the efforts of most corporate leaders across most sectors of the economy to return employees to the office, many are still choosing hybrid work, and are coming to the office only sporadically.

The surprise, perhaps, is just how effective groups comprised of many hybrid workers seem to be. Teams are functioning, the work is getting done, and the employees are on balance happy with the arrangement — even if management continues to grumble.

I was talking about this recently with a technology colleague, a long-time pharma industry veteran who only recently joined our company, and had actually started working remotely even before the pandemic.

My colleague emphasized – correctly, I think – how much everyone has learned about remote work during the pandemic. The combination of improved tools and extensive experience has enabled many employees to become phenomenally effective. Particularly when combined with occasional, high-value in-person team experiences, hybrid work seems like a promising alternative for many employees that seems here to stay.

Full disclosure — I remain a bit old-school here. As much as I dislike the commute and dread the Cambridge traffic each day, I still come to the office 4-5 days a week. I like running into people unexpectedly and serendipitously, and I appreciate the sense of place – however empty the place may often be.

When not driving my kids to and from school — moments I cherish — I use my time in the car to listen to podcasts and audiobooks, catch up with friends and family, and follow-up with colleagues at work. But could I use the hours I spend in bumper-to-bumper traffic more productively outside the confines of my automobile? I suspect the answer is yes.