One year later: Did expert AI predictions deliver?

This year marked a significant milestone as an AI-designed drug advanced deep into clinical trials, demonstrating substantial reductions in both time and cost compared to traditional development processes. While drug development can easily cost over $2 billion and take 10-15 years, all for a single drug, one team recently demonstrated that their developmental pace was significantly faster, taking the process of preclinical candidate nomination through phase 0/1 testing to a matter of not years, but months.

As more of these AI-driven therapeutics progress through clinical trials, expectations are mounting for this technology to deliver on its promise of faster, more efficient drug development. 

AI can serve what we truly believe is an aspirational goal which is to bring better drugs faster to the patients who can benefit most.

 Daphne Koller, insitro

Artificial intelligence tools excel when fed diverse data from multiple sources, revealing unexpected insights that would otherwise remain hidden. Multimodal AI is an exciting frontier filled with tools that can process different data types simultaneously, such as text, images and audio. These systems are attracting significant experimentation from practitioners eager to explore their potential. Of course, like all AI technologies, multimodal tools come with inherent concerns and drawbacks that must be carefully considered as the field advances.

When we think about modeling cells, compared to proteins, the scale is just enormously different. A protein is very small compared to the size of an entire cell. How can we possibly bridge those scales? There’s a lot of emergent properties in biology, and historically we’ve been very bad at modeling those. We need a lot of multimodal data captured and multimodal models – but also multi-scale models. Because if we want to understand biology, it’s about the spatial scales.

Emma Lundberg, Stanford University

The most significant barrier to adopting artificial intelligence tools, particularly in medical contexts, continues to be trust. Recent studies have shown that user and patient comfort and confidence represent one of the biggest hurdles to widespread AI adoption in healthcare. But even as trust issues persist, we’re seeing them integrated across all stages and in all aspects of the world of medicine, from research to patient interactions with providers, and even medical education.

What matters for the law, what matters for ethics is the deeds, the way in which it impacts patients' experience, whether it's an algorithm, predictive algorithm, predictive analytics, old school regression, it doesn't really matter as far as the law is concerned. What matters is, it right? Is it wrong? How much care did you take in doing it? Now, saying that that's true from the law's perspective is very different from saying that's true from the patient's perspective. When you talk to patients and you say, I'm using AI, in the best case scenario, they're thinking Star Trek. In the middle case scenario, they're thinking Battlestar Galactica. In the worst case scenario, they're thinking the Terminator. They care a lot about this, and they don't understand it very well.

Vardit Ravitsky, The Hastings Center

AI already has its share of evangelists in the provider community who praise the positive changes it has made to how they run their practice and communicate with patients. Tools like AI scribes can take detailed notes in the background, enabling providers to focus on speaking and listening to their patients. But so far, the patients who report a positive experience with AI tend to be those with a high level of comfort with technology, enabling them to understand how the tools work. More remains to be done to ensure that many more stakeholders across the healthcare ecosystem – from scientists to providers to patients – can embrace the benefits these tools offer.

The key next challenge is that the regulatory spectrum has not kept pace with how fast these developments are happening. Digital pathology adoption in the U.S. is around 4% for clinical use, which is really low. And I think a key reason is there hasn’t been a good, acceptable use case – an FDA-approved device – for primary diagnosis. That’s preventing us from getting to that adoption.

Faisal Mahmood, Harvard Medical School

Indeed, we’re still awaiting the first drugs that demonstrate AI's complete end-to-end value in translational medicine. But that’s because it’s a massive undertaking to change how the whole process works. We are seeing important changes in many parts of the process but linking all the parts of the process and accounting for all the parts of the process outside our control (e.g. regulators) is going to take dedicated time, collaboration and sustained investment from the entire industry.

But by any measure, quantitative or qualitative, we’ve seen fast movement toward that goal in the past year. Even though most experiments that use AI in general across enterprises don’t make it “out of the lab,” the ones that do are becoming indispensable.

In this period when we see a patchwork of these tools being used across many parts of the drug discovery and development process, it can be useful to think about the scale of impact when evaluating value. 80% to 90% of all biologics drugs are manufactured on instruments made by Danaher. The gains that have been realized by implementing tools even partially across the development process, have shown impactful value. 

The biological tools industry can learn so much from the semiconductor industry in terms of the scale, precision, instrumentation quality and reliability, as well as the ability to mass-fabricate these highly technical tools and solutions. Weave this in with the advances we’re seeing with data and AI for complex pattern recognition and insight generation, and I believe that in the not too-distant future we will come to a point where biology transitions to an applied science which is more deterministic.

Murali Venkatesan, Vice President of Science, Technology and Innovation and Head of Danaher Ventures LLC 

There’s no doubt AI is pushing life science forward at an unprecedented pace. It’s never been more important to reflect on what we can reasonably expect AI to do given the time and investment to date. But the gains we’ve made in the field, even over the course of just a year, continue to keep our experts enthusiastic for the future of solving the world’s problems at the speed of life.