AI Will Help Bring About The Bio Age Too

The explosive arrival of AI for all is accelerating the takeoff of bioengineering, a second world-historic technology with huge ramifications for the climate & healthcare

The arrival of artificial intelligence for everyone, in the form of Generative AI, is a world-historic event that will become increasingly clear as we figure out all the things this general purpose technology can do and watch its widening impact in solving our challenges and ultimately making a better world. That’s essentially the mission of The AI Age Begins series that systematically looks at how AI could change the game for the better in a wide range of fields.

But the remarkable thing about the 2020s is that we actually have another world-historic technology emerging at the same time in the form of what can be broadly called bioengineering. We are in the early stages of a range of breakthroughs in our understanding of the life sciences that are coming together to give us the power to engineer living things, to not just understand but manipulate biology, in a new field called synthetic biology.

This development could have huge implications in dealing with a range of challenges from mitigating climate change to dramatically improving healthcare. Humans’ ability to reengineer life, from plants and animals, to bacteria and viruses, could open up an age of biological production of sustainable materials much more in sync with nature that could eventually supersede the methods of industrial publication that have brought great prosperity but also much environmental damage, including climate change.

What we learn in synthetic biology also will be able to be applied to human life, with the possibility of healthcare that could become much more precise and effective, as well as much less costly— perhaps achieving the holy grail of personalized medicine at scale.

So the beginning of The AI Age is coinciding with the beginning of what we might call The Bio Age as well. We’ve now got two transformative general purpose technologies arriving at roughly the same time. (We actually have a third in the form of clean energy but let’s keep it simple for now.) The AI Age and The Bio Age are not just parallel developments but highly synergistic ones. Like we’ve seen in other fields, AI has the potential to supercharge progress in the Bio world, in bioengineering and healthcare.

Most people in the world have no idea that this is happening, but there is a growing number of people who do. So we gathered about 250 of them in our event this month at the Shack15 club at ground zero in San Francisco, and had a dozen experts try to answer our question of the night: How can AI accelerate progress in bioengineering and healthcare?

We had top minds from the Bio World who understood genetics and genetic engineering and stem cells and immunology and medicine and drug development and materials science, among other things. We had top minds from the AI world who understood the new capabilities of Generative AI and machine learning, including some from the networks of our partner BCG X and our sponsors Google, Adobe, Writer, Radium and B Capital. We also had key players who could fund breakthrough developments, from venture capitalists to representatives of billionaire philanthropists to U.S. government officials from the Defense Department.

This essay will lay out the many big ideas and key insights that emerged from our conversation that evening, which we also captured in highly produced videos of all the short talks. (Follow the links from each of the speakers quoted below to watch their full talks and get all the nuance of their ideas. Or watch the entire YouTube playlist.)

One of the key themes of the evening was that the convergence of the AI and Bio worlds is just beginning and practitioners in each realm have little understanding of the other realm. Biologists may be underestimating the potential of AI to supercharge progress in mastering complex challenges. But AI techies may be underestimating the incredible complexity of biological systems and how much we humans have yet to learn.

"When is there going to be the ChatGPT moment for biology?," said Anna Marie Wagner, the Head of AI at Ginkgo Bioworks, which plays in both realms. “If you talk to a traditional Bio AI CEO, they'll basically say: ‘Once I raise enough money to get enough compute to train the big model. There's tons of data, there are trillions of tokens, we have the data that we need, we just need to compute."

“And if you talk to a biologist and you ask them that question, they say: ‘Can I get back to the lab? This is stupid. Why are we talking about this?’ And so from my perspective, one of the biggest issues that we have today is that with very few exceptions, those communities are not talking to each other well. They're not respecting each other, and they're not respecting the magic of biology.”

“Part of the humility needs to come from recognizing that while we invented language and basically everything that we've ever thought or said or written is somewhere on the internet for AI to train on — biology invented us.”

That said, the conversation was upbeat with most speakers genuinely optimistic that we’re entering a new era of accelerated breakthroughs in the life sciences, as well as in the development of synthetic biology that can help humans deal with climate change. That new era probably will bring about a transformation in healthcare too.

“At the same time we're experiencing exponential leaps in AI, we're also benefiting from a multi-layered wave of innovation across a spectrum of categories converging to advance human health, said Unity Stoakes, President & Cofounder of Startup Health, which has invested in 500 health startups.

“It’s this mashup that gives me such hope: Wearable tech, affordable sensors, home diagnostics, neuro tech, brain-computer interfaces, robotics, nanotech, genomics, microbiome, personalized medicine, AI-powered drug discovery. We're in the midst of a new era for health innovation.”

To be sure, the possibilities of using AI to not just better understand but engineer the world of living things comes with risks. The group was cognizant of the fact that tinkering with life and reinventing healthcare for humans needs to be done with caution and a heavy sense of responsibility. No one knows that better than Micheal Koeris, the newly installed Director of the Biological Technologies Office of DARPA, or the Defense Advanced Research Projects Agency.

“Technology and science is fundamentally democratizing,” Koeris said. “That's great. It gives everybody the opportunity to do everything and anything that they can, which is good. If you want to develop new medicines, if you want to help the planet, if you want to actually help and develop solutions that affect us all.”

“Conversely, if you actually like to do something that's not quite as nice, you're also appropriately empowered nowadays.”

The complexity of the world of biology & the big-picture view ahead

AI was born of complexity to help humans deal with complexity. Generative AI emerged by applying very high levels of computer power to vast amounts of data far beyond the comprehension of individual human minds. Now that Gen AI is up and running, it’s starting to help us humans deal with many of the complexities of the modern world.

The complexity of biology, however, is in another league. The human body has somewhere between 50 to 100 trillion cells, said Mickey McManus, a Visiting Scholar at Tufts Bioengineering Lab, and a senior advisor to BCG. McManus is the coauthor of “Trillions: Thriving in the Emerging Information Ecology,” written about a decade ago to apply to what became known as the Internet of Things.

He’s one of those who created companies that dealt with the built world of inert things and mastered techniques that worked in the digital world of electronic computers. But now he’s trying to transition to the new Bio world and come to terms with the complexity of nature in comparison.

“We're a complex information system,” he said. “We don't reboot every month. We'll go 80, 90, 100 years without a catastrophic failure. We're mostly bottom-up, peer-to-peer, every single cell in our body, except for mature red blood cells, have a spare copy of us in the form of DNA.”

Andrew Hessel is a pioneer in synthetic biology who made that transition between fields early. He’s a cofounder of Genome Project-Write, and now has a company that makes viruses that kill cancer cells.

“I love biology because it's the only technology that we all have in common and it's also the only technology we didn't create, so there's always this mystery to it,” Hessel said. “It's been a black box for most of humanity and really, it's just been in the last hundred years that we've really been able to crack open that box and start to explore it.”

Hessel started his career excited by the world of electronics and simple digital computers, with hardware and software, inputs and outputs, mastering the art of programing. Now he wants to help figure out how to program biology, and he thinks AI can really propel us forward.

We still don’t fully understand how a single cell, one of those 100 trillion that make each of us up, really works. Hessel thinks AI will be able to help us model a single cell, perhaps starting with the single celled organism E. Coli bacterium. Then if we had what in effect is a digital twin of that real organism, we could rapidly experiment in simulations and only later move to real life.

“Anything you tweak in that virtual model, anything you change, you'll have full control over time, and space, and the molecular structure,” he said. “But when you make a change, the simulator will run it. And if you want to try it in the real world, you just will have to hit print, and that organism will get created.”

He thinks that engineering ability is now within reach and that his young children will master programing biology the way that he mastered programing electronics.

“When I was their age, I had access to electronics, and chemistry, and simple computers. And now I want them to have the same access to the new form of computing biology that's literally going to change our world and make it better, keep us safe, make us heal diseases, correct climate, and so much more.”

The roadblocks & possible breakthroughs ahead in the Life Sciences

Tom Kalil, the founder of the new Renaissance Philanthropy, has a vision that these world-historic breakthroughs in science and technology, like in AI and Bioengineering, could create the context for a modern day Renaissance.

“Our vision is that in the same way that wealthy Italian families supported the Italian Renaissance by supporting Michelangelo and da Vinci, today's philanthropists could support a 21st century Renaissance that is fueled by advances in science and technology,” said Kalil, who most recently worked as the Chief Innovation Officer of Schmidt Futures, the philanthropic arm of former Google CEO Eric Schmidt.

One opportunity Kalil is now focused on is to try to tap into the 2000 or so American families who have more than $500 million in wealth yet who only give away 1.2 percent of that each years in philanthropy— far less than they undoubtedly make up in the stock markets.

“One of the areas where they could be more active is the use of AI and machine learning to accelerate the pace of scientific research,” he said. “Many of the challenges that we face in the 21st century, whether it's allowing people around the world to live longer, healthier lives, accelerating the transition to a carbon-neutral economy, improving our ability to respond to future pandemics, can really take advantage of not only AI, but all these other AI-adjacent technologies like self-driving labs and modeling and simulation software.”

The most striking success to date of using AI to create a breakthrough in the life sciences was the recent solution to the protein structure prediction problem, which had stymied humans for 50 years. Demis Hassabis of Google’s DeepMind assigned a team of 18 people to create the AI Model AlphaFold, which was able to analyze the sequence of amino acids and predict the three-dimensional shapes of the proteins that get created from them.

Why did it work? The machine learning community has known for a while now that progress in AI often comes when the problem they are trying to solve meets what is called “the common task framework” and four criteria are met. One, the problem is well defined. Two, and this is extremely important — there is plenty of data. Three, there is a clear benchmark for evaluating progress. And four, you can have a leaderboard that can pit competing team efforts.

“I would submit that if we want to make more progress in applying AI and machine learning to healthcare and to biology, that that is a play that we need to run,” Kalil said. “And it is conventional wisdom within the machine learning community, but it is still relatively new to the biology and healthcare community.”

The protein structure prediction problem met all four criteria but the key criteria was having large amounts of good data. In this case, the biology community had spent decades collectively building up what was known as the protein data bank accessible to everyone. Unfortunately, that situation is unusual.

“We have a systemic issue in the way that science is done that exacerbates the lack of collaboration that we see right now at the intersection of Bio and AI,” said Wagner, of Ginkgo Bioworks.

R&D dollars that go to the life sciences go into one of two worlds that have very little cross-over, she said. A big chunk goes into academia where it’s divvied up into tiny grants which pay highly trained scientists to perform a lot of manual lab work. This is a highly inefficient process but the results are made public and shared with the broader community who can all learn.

The other big chunk goes into industry that does have the resources to build infrastructure and automate many processes that bring much more internal efficiency but they keep all their results secretive. So industry as a whole is inefficient because competitors need to keep reinventing the wheel and going down paths that their rivals know from their own experiments will end up in failure.

“The problem that I'm really interested in thinking about next is this systemic question of how do we create a third model?,” said Wagner, who said she was soon leaving Ginkgo Bioworks to try and solve this new challenge.

The near-term implications for Healthcare in systematically applying AI

Can Generative AI be trusted in human Healthcare? Some critics say that today’s Gen AI is too inconsistent, too unreliable, and potentially biased. There’s no place for such behavior when treating sick people. No way we should take such a risk.

“If you describe to me an entity that works really hard and really fast, has an enormous wealth of information, but doesn't really know how to apply it in context, what you're describing is a fresh medical school graduate,” said Zayed Yasin, an experienced physician who also now works as the Healthcare Industry Lead for the Gen AI company Writer AI.

“And we don't ban them, we don't say that we don't want them in healthcare, we have developed a set of regulations and processes so that we can use them effectively and safely in the moment, as well as multi-year-long training and validation processes to get them better and better and better in terms of their performance over time.”

Yasin and his colleagues at Writer AI have come up with a great analogy to how to handle the integration of artificial intelligence into healthcare. They point to the auto industry that also is a highly regulated industry confronting life-and-death situations that already has an enormous amount of legacy infrastructure to take into account. The auto industry also has a head start in creating a safe process that slowly but surely is developing artificial intelligence that can meet, and probably supersede, the performance of human drivers.

The Society of Automotive Engineers years back came up with a 5 level framework that set milestones going from a car that has zero automation to help the human driver to one that has zero involvement with humans but still can safely drive. The graphic above lays out the 5 levels and shows that new modern cars already are at least at level 2 with computer assistance in braking and steering. Many high-end new cars like Teslas today have the ability to drive by themselves on freeways as long as the human drivers are there to take over. In San Francisco and a couple other cities, Waymo now operates at level 4 with completely driverless taxis, but not in all conditions like on freeways. But level 5 autonomous vehicles that can drive everywhere are clearly within sight.

The graphic below shows the analogous 5 step process for healthcare. Yasin said Gen AI already is starting to make an impact on Level 1, which is focused on administrative tasks in the back-office with no involvement with patients. Level 2 does enter the clinical process but is all about lowering the amount of busy work bogging down the working professionals. In Level 3 AI is aiding doctors and nurses in clinical decision-making and making those decisions more accurate and efficient, but this is always done transparently and with human oversight. Level 4 is where we enter the realm of the “black box” where we don’t know exactly why the AI is recommending certain decisions but by that time we will have come to trust it. This level is farther away and would need to be highly regulated like medical devices are currently.

Level 5 is where the AI could operate as a nurse or doctor without human oversight. This is farther on the horizon but may come sooner than many think. Yasin made the argument that the risks that AI might occasionally make a less than perfect diagnosis might be weighed against all the good that could come from making healthcare cheaper and more accessible to people who go totally without it today.

“I think this makes intuitive sense for anyone who's experienced healthcare: people aren't dropping dead because their doctors are making really bad decisions, prescribing the wrong medications, misdiagnosing — it happens— but it's not the big problem,” Yasin said. “The big problem is that people are locked out of care because they can't afford it, because the system doesn't have capacity.”

The need to scale & the quest for the Holy Grail of personalized medicine

Aashima Gupta, Global Director for Healthcare for Google Cloud, picked up on Yasmin’s theme and gave concrete examples of some of the 5 Levels. She pointed out that Google has already built AI systems that detect breast cancer with the accuracy of human radiologists— operating in Level 3.

“And the pace astounds me on Generative AI,” Gupta said. “Just last year, we introduced our medically-tuned foundation model called MedLM. It was the first state-of-the-art model to pass the U.S. Medical Licensing Exam with the passing score of 60 percent. In the current iteration, the model is passing the exam with 91 percent. And it's also not just passed the U.S. Exam, it also passed the Indian medical licensing exam.”

Meanwhile, the World Health Organization (WHO) projects that there will be a global shortage of 10 million healthcare workers by 2030. And on a planet of 8 billion people, many billions have no interaction with any healthcare system. Wouldn’t they rather have some interaction with a largely accurate yet cheap AI at Level 5?

Or wouldn’t people in the developed world still want more personalized healthcare that goes beyond an annual checkup and uses AI to constantly evaluate all updated streams of data coming from wearables and the like, combined with the latest research on topics concerning them?

“Very, very few people in the world— 0.1 percent— have the kind of concierge health and personal coaches,” Gupta said. “We believe AI can be that coach for all of us. This is where the democratization of healthcare has to work.”

Google as a company is all about scale, and scaling into global markets. Gupta gave a good example of how Google is already wading into Level 2 of streamlining clinical processes. They did a project with a group of 188 hospitals that employ 100,000 nurses and figured out how to use Gen AI to speed up the process that goes into creating and consuming the “nurse shift summary” that happens 40,000 times in a given day. If you just shave 5 minutes off the average 75 minute process, then every day you get 200,000 minutes freed up to redirect to hands-on care delivery.

BCG, the Boston Consulting Group, also operates all over the world and is all about scale. Matt Kropp, the CTO of BCG X, the group’s technology division, gave an example of a project they are doing with a pharmaceutical company to use Gen AI to speed up the filing process to get a drug into clinical trials.

“Technical writers do this work,” Kropp said. “It takes five months or more for them to complete these filings. We've been able to work with them to build a Gen AI-based solution, which does research, creates all of the elements of the report, basically writes this filing in a matter of minutes or an hour.”

“Is this about cost savings, is this about reducing the number of technical writers? Absolutely not,” he said. “What it's about is if they can move a drug to market three months faster, then that’s worth hundreds of millions of dollars. It's not about the cost, it's about the efficacy.”

Bioengineering & the quest to build a sustainable world

Not all the breakthroughs are coming from the AI World and driving change. There are breakthroughs in the science of the Bio World that the AI builders need to wrap their heads around too.

“I'm sure we've all heard it said at one point or another that AI will give us a way to harness all of human intelligence,” said Ashlee Hutchinson, a stem cell scientist working as a Program Manager for Revive & Restore. “Well, I'm really excited to share that we're actually entering an almost parallel phase in our ability to harness all of developmental biology, because thanks to stem cells, we now have the power to grow a whole animal from the starting point of a skin sample.”

“What if we had the power to take any endangered species and make more whenever we wanted? Producing healthy new individuals full of genetic variation? What if we could incorporate genetic engineering to further induce that variation for resistance to disease or climate change and to restore the earth's biodiversity?

“I'm here to tell you that we can, this is possible, and we already have all these tools at our disposal through advances in stem cell technology, genomics, and bioengineering.”

Let’s step back and unpack all the pieces Hutchinson laid out here. Mounting climate change alongside expanding human development is damaging if not destroying the natural habitat of many species and driving some of them towards extinction. As these habitats get lost and the population of species shrink, so too does their gene pool, and they become susceptible to inbreeding. Without genetic variation they can’t adapt to the changing environment and are likely to become extinct.

Revive & Restore is a non-profit organization founded by Ryan Phelan and the legendary Stewart Brand that funds and supports pioneering projects aimed at the genetic rescue of endangered and even extinct species. Hutchinson oversees projects related to stem cells, which are the all-purpose cells that can grow into any kind of cell, and she said she is now focused on how we can use “induced pluripotent stem cells” to prevent further extinctions.

A little over 15 years ago scientists made a breakthrough that allowed them to take an ordinary skin cell, turn back the genetic clock, and turn those cells into the equivalent of embryonic stem cells, or what are now called induced pluripotent stem cells or iPSCs. Hutchinson said that since then, scientists have been able to convert these induced stem cells into functional sperm and eggs, and there are a number of ways they can also be used to make an embryo.

So it’s theoretically possible to take skin samples in a non-invasive way from all species heading toward extinction, and convert those cells into stem cells, which can then grow into the animals themselves. Along the way you could add genetic modifications that could improve the genetic health of a population and help them better adapt to climate change. This is all within reach.

“But to derive stem cells from all these different species and make offspring, that will require us to really unpack the regulation of pluripotency across all these distinct lineages,” Hutchinson said. “And here is where I see AI emerging as a way to unravel all that biological complexity and really speed things up.”

The opportunity now for Moon Shots into the 21st Century

We have a routine this season of closing each of our events by taking the pulse of all the 250 or so experts and innovators who gathered about their level of optimism on the topic of the night. In April for our event on how AI could accelerate progress in solving climate change, the mood was more sober than any of the previous ones. We’re really up against the odds when it comes to stopping global warming and that can bring down the spirits of even stalwart techno-optimists.

This month’s gathering was predominantly confident that AI will be a key tool in supercharging bioengineering and transforming healthcare by 2035. But further conversation with the group showed much less confidence about the level of access most people would have to the wonders of healthcare a decade hence.

And one of our last speakers was DARPA’s Michael Koeris who reminded all of us that the powers of these new general purpose technologies are formidable, but also dangerous.

“I'm going to be seeing a couple of folks later who have shown that it's actually significantly easier than we had previously imagined using Generative AI to actually recode toxins from snails, and they don't look like snail toxins,” Koeris said. “Okay, that's a problem. It exists. It's our job to now wrestle with the problem.”

Koeris had flown in from his headquarters in Washington D.C. because his responsibility now is to keep the U.S. military and more generally the U.S. government on the cutting edge of the latest technologies and stay steps ahead of the capabilities of our potential adversaries.

“Why I'm here today is to make sure that we, on the West Coast and the East Coast, keep working together,” he said.

Our closing speaker picked up on that theme of public-private collaboration but with a more positive twist. Startup Health’s Unity Stoakes made the case that now more than ever we need to be inspired by NASA’s Moonshot that brought the entire country together. That monumental effort involved more than 20,000 companies and 400,000 people all focused on a goal that many deemed impossible.

“Progress happened for three reasons,” Stoakes said. “The challenge was sufficiently hard to attract the world's best and brightest, number one. Two, the mission was inspiring and captivated the world's imagination. And three, success depended not just on technology but on collaboration.”

Today we have many of our own daunting challenges, from climate change, to the next big pandemic, to diseases that have haunted humanity from the beginning of time.

“How do we cure Alzheimer's faster? How do we cure diabetes faster? How do we cure cancer faster? How do we deliver care to billions more people with the tools we already have?,” said Stoakes.

“We need more collaboration, cross-pollination and multidisciplinary innovation. We need more technologists merging with doctor-preneurs and healthcare leaders to unlock the potential of the tools and technologies already available. We need designers, scientists, engineers, entrepreneurs, investors and patients working on these challenges together.

This time around the collaboration does not necessarily need to be catalyzed by the national government or any government agency. This gathering we just held a couple weeks ago shows how much pent up desire there is for people from the AI fields and the Bio fields to come together and learn from each other. We just opened up a conversation that could continue for months if not years ahead.

(In fact, given the success of this event, we at Reinvent Futures are considering launching a separate series tentatively called The Bio Age Begins that would hold 4 events a year, one per quarter. If you are interested in getting involved as a partner or sponsor contact us.)

“Imagine if we cross-pollinate today's most amazing tools and supercharge them with AI. Imagine the step change forward for humanity when we all start collaborating on the hardest challenges,” Stoakes said to end.

“This is how we go from a thousand songs in our pocket to safely delivering to billions of people— a million doctors in your pocket.”

Special thanks go to our key Partners & Sponsors who brought together the networks & resources to help make this event & The AI Begins series successful.