Why AI Is 'Not Particularly Good' at Curing Disease (Plus: The Next GLP-1 Boom and Why America Hates Big Pharma)
A wide-ranging interview with Dave Ricks, the CEO of Eli Lilly, which makes the GLP-1 drugs Mounjaro and Zepbound and recently became the first trillion-dollar pharma company in history
One way to tell the story of human history is to watch how we die.
At the scale of millennia, the most deadly disease was likely smallpox. Some scientists estimate that it killed billions of people before its cure was summoned from a barn animal. In the 1790s, a British physician named Edward Jenner took material from a cow infected with cowpox and used it to inoculate a young boy. Since that fortunate but deeply unethical science experiment, vaccination—the word coming from the Latin for cow, vacca—has spared hundreds of millions of people from an early death.
By the early 1900s, in rich countries like the US where smallpox was on the wane, another killer emerged as the apex predator of humanity: bacterial infection. According to the best records we have, almost all of the top causes of death in the late 19th century were bacterial, whether it was gunky stuff in your lungs or bugs in your gut. But thanks to the accidental discovery of penicillin in the 1920s, and a heroic effort by the United States to scale up the drug during World War II, bacterial infections have plummeted.
Today the top causes of death in the U.S. are not poxes or bacteria. They are heart disease, cancer, stroke, and Alzheimer’s. While we have made some progress against this new class of villain—statins for our hearts, immunotherapy for cancer—they are still killing us with remarkable frequency.
There is an idea in progress studies called the burden of knowledge. Sometimes it looks like progress in a field like medical science is slowing down, not because we’re getting dumber over time, but ironically because we’re getting smarter. Every time we solve a problem in medicine, it’s like plucking a fruit from a tree. If you pick all the low-hanging fruit, what you’re left with is the taller challenge.
One of the tallest challenges for medicine in the last few decades has been dealing with the complications of obesity. Americans who overeat are at a higher risk of a number of illnesses and discomforts: inflammation, cancer, knee pain, sleep apnea. But for a long time, we didn’t have a reliable technology to help people eat less in an environment of food abundance. Science needed new fruits to pluck. But first we needed a taller ladder.
That ladder is the GLP-1 drug revolution, which shows no signs of slowing down. In December 2025, Eli Lilly, the most valuable pharmaceutical company in the world, released results for a late-stage trial of retatrutide, its next-generation GLP-1 drug. Unlike initial drugs in this class, retatrutide targets not one hormone (like Ozempic) or two (like Zepbound) but three. The weight-loss effects were historic: Patients lost an average of 29 percent of their body weight over roughly 68 weeks and saw huge improvements in knee pain. The anticipation of retatrutide borders on the rapturous: In a recent podcast interview, the scientist and health influencer Andrew Huberman called retatrutide the protein that will “change everything.”
But this is a strange moment for the pharmaceutical industry. In Gallup surveys, Americans rate the drug industry less positively than every other business sector. On the right, the “Make America Healthy Again” movement has gained traction with a broad suspicion of therapies and health authorities, along with a taste for alternative, “do your own research” medicine. Meanwhile, AI boosters often excuse the technology’s downsides by reminding doubters that it will one day “cure cancer.” But I’ve never heard a particularly detailed explanation of how that might even happen.
To understand more about the GLP-1 drug revolution, the promise (or over-promise) of AI in medicine, and the roots of public distrust of pharma, I recently spoke to Dave Ricks, the chief executive of Eli Lilly. The following transcript has been edited for clarity, brevity, and the goal of making both speakers sound a bit smarter1.
THE HISTORY OF GLP-1 DRUGS
Derek Thompson: Eli Lilly is the largest pharmaceutical company in the world, mostly because of your GLP-1 drugs. Twenty years ago, Lilly helped launch the first approved GLP-1 drug. I’d like you to tell that story from wherever you’d like to start: 2005, the 1970s, the dawn of man. How did this drug category get kicked off?
David Ricks: The real first scientific breakthrough that happened was in the early 1970s. There were several papers that looked at something called “the incretin effect.” Incretins are a class of hormones or proteins that your gut excretes, like GLP-1. They noticed that if you ingested food orally versus intravenously, your gut signaled to the rest of your body that you’d eaten something [and released insulin to reduce blood sugar]. But if you took the same food intravenously, blood glucose would spike and stay up. The hypothesis was that some protein in the gut was signaling to the body that you’d eaten.
The first protein we identified was actually GIP, which is now the backbone of tirzepatide or Zepbound, our best-selling drug. The companies that worked on this in the 1970s and 1980s included Eli Lilly and Novo Nordisk, the maker of Ozempic. You could ask, “Well, why in the ‘70s didn’t you just make this drug then if you knew about this?” The answer is that peptide chemistry wasn’t very mature. If you give humans native GLP-1 or GIP, they both have a half-life in your body of seven or eight minutes. So it wouldn’t be a very convenient drug. You’d have to walk around with an IV bag all day.
A key breakthrough came from the Gila monster study in the late 1980s. John Eng, a doctor in the Bronx, was interested in why certain animals didn’t need to eat very often. He identified that Gila monsters have a mimic of the human peptide GLP-1 that the lizard uses [to stay full]. The Gila monster protein could go about 8 to 10 hours. So you could make a twice-a-day shot. That barely cleared the hurdle of good enough to be a drug. We made it into a drug along with a partner, Amylin, which launched it in 2006.
Thompson: Your story stopped in 2006 with the first version of this drug, which was used for type-2 diabetes. How long did it take Lilly to recognize that what had begun as a type-2 diabetes drug was in fact an incredibly powerful weight loss drug?
Ricks: When you look on the cover of our annual report from that year, there’s a patient, Maria, and she’s talking about her experience with Byetta, the brand name for exenatide, and she says, “It’s controlling her blood sugar and that her friends notice she’s losing some weight.” There it was like sitting there in April 2007, when we put out that annual report. We were doing studies.
But we had a problem.
The problem was the side effects associated with GLP-1 drugs are what we call peak to trough, the difference between the highest amount of drug and the lowest. Twice a day dosing [was causing] a lot of GI distress. You’ve about nausea, diarrhea, vomiting. The more we push the dose up, the more we ran into these things with that particular drug. We stopped because it was very unpleasant to take, and the weight loss effects were there but more modest.
What started the weight loss part was our ability to get dosing levels up. The incretin system—GLP-1, GIP, and coming soon glucagon—is being boosted in people with regular levels. People aren’t deficient necessarily in these hormones, like GLP-1. We’re boosting them to a super-normal level and it suppresses appetite and kicks up metabolism. To do that, you need to get to higher levels. We produced a once weekly GLP-1 in 2014 called Trulicity. Anyone who was in the diabetes world would know that. Very few people in the lay media would remember that drug, although it was our best-selling drug at one time. With a trick of protein engineering, we made that drug last a week, and people lost more weight.
Our competitor, Novo Nordisk, then did a similar trick and made a weekly that launched in 2017 called Ozempic. That only got so famous when they took the chance to study that at an even higher dose in people who had obesity but not diabetes. That came out in 2021, and then this whole thing started running.
At Lilly, we invented this drug called tirzepatide [which targets two hormones: GLP-1 and GIP] in 2014. We started doing studies around 2019. We knew this was going to be the most effective obesity drug we’ve ever seen. Of course, since then, we’ve been improving upon that. It’s a case of leapfrogging and iterative improvement.
Thompson: Just a pit stop on the business here before I get into the science. What share of Lilly’s revenue and profit at the moment is related to this class of drugs right now?
Ricks: Last year, it was a little less than half the total company revenue. This year, we don’t give out product level guidance, but it’s projected to exceed half. It’s the fastest growing part of the company by far.
[Note: I think it’s worth pausing to reflect on the fact that the most valuable pharmaceutical company in the world is 149 years old and significantly more than half the company’s total revenue is from drugs that the FDA only approved four years ago. This is not a normal industry! More on that later in the interview.]
WHY THESE DRUGS SEEM TO DO EVERYTHING
Thompson: To me, the most mysterious and most even confounding aspect of these drugs is that they seem to do so many things at once. They increase feelings of fullness. They help people stop smoking or gambling. They reduce knee pain. They seem to reduce inflammation. They seem, in some cases, to protect against heart disease among people who aren’t even losing weight, which suggests some mechanism that is maybe slightly unrelated to obesity. They might even have brain benefits. They’re being studied for dementia. You are as close to the best scientists studying GLP-1’s as anybody in the world. What the hell is going on here?
Ricks: I’d add one more category, which is cancer reduction. There’s a lot of interesting studies about population-wide cancer risk reduction, maybe more directly associated with losing weight or maybe the inflammation story.
Okay, so just to take it all the way back, a core behavior to keep organisms alive is feeding. We can all understand that. Animals have grown up in an environment of scarcity. There are very few counter-regulatory processes that we’ve evolved with, because our ancestors evolved in scarcity. It was unnecessary to put in our genome something that says, “Okay, you’re full.”
Human beings have evolved into an environment we weren’t built for—of abundance, to maybe hit that note for you. What does that do? More weight, more type-2 diabetes. You’re too heavy, and your insulin, which helps you process sugar, no longer works effectively because essentially, there’s resistance from your body from having too much sugar for too long. If you reduce weight, you increase insulin sensitivity. We have a study where we watch people on tirzepatide for three years, and they had 92% less diabetes at the end of the study. That’s a remarkable finding.
These drugs also reduce inflammation in ways that may be separate to over-consumption of calories. I think our scientists debate that still. There’s a key signal called hs-CRP, a protein that measures general inflammation. On these drugs that drops 50 to 70%, and it drops a lot faster than your weight.
Speaking of inflammation, we did a study with retatrutide, the next generation product we’re working on that reduced knee pain. [Tirzepatide is a dual-agonist, working on GLP-1 and GIP; retatrutide is a triple-agonist, which also targets the gut hormone glucagon.] Half the people in the study had no knee pain at the end. It was the most effective osteoarthritis, knee pain drug I think ever studied. Here you have probably two effects: the inflammatory factors and the offloading of weight.
One last thing just on the brain, because I think it is the strangest and most difficult to explain. In addition to suppressing our appetite for sugary foods, it does also seem to suppress our appetite for vices that are maybe driven by anxiety or driven by a cycle in our brain of trying to scratch an itch or have a craving. Tobacco use and alcohol use are the most noted in the literature. We’re actually doing prospective studies to prove both those things. But no one would’ve hypothesized that at the beginning.
WHAT ARE THESE THINGS DOING TO OUR BRAINS?
Thompson: What’s spookiest to me is the degree to which these drugs almost act as a system-wide mechanism for increasing moderation. It moderates our impulse to eat. It moderates our impulse to gamble. In some ways, it even seems to give patients the urgency to align their behavior with intent. How does this work?
Ricks: There seems to be some other signaling going on. The human brain is very complicated and pretty poorly understood relative to other organ systems. So we’re still trying to figure out exactly the mechanisms of these. What we can say is empirically, people report it and we are doing prospective studies to take people without obesity who have these problems you mentioned—too much gambling, even opioid use disorder—just to see, well, can we help?
Thompson: There’s a theory of science that says we develop a perfect theory of how the world works, and then we develop technology that applies that theory to the physical world. But it almost always seems to be the opposite.
Ricks: We wish.
Thompson: The classic example I think from tech history is that theoretically you would want to understand the principles of thermodynamics to invent the steam engine. But we invented the steam engine and then from it derived the principles of thermodynamics. And so, maybe in a similar way, it would be lovely if we accidentally fell into a synthesized Gila monster molecule that taught us the secrets of the gut-brain axis rather than having some perfect theory of how our brains worked and then said, “Oh, we should be doing is targeting glucagon and GIP in the gut.” So it would be lovely if this technology opened up some big secrets to the brain.
But for all the good that these drugs can do, I believe according to the most recent data, more than half the folks who start on GLP-1 drugs discontinue them before two years is up. So there’s still a pretty significant adherence problem, made even more significant by the fact that people who discontinue these drugs often gain the weight back. Why do you think the discontinuation rate is so high?
Ricks: The data so far for Mounjaro or Zepbound, otherwise known as tirzepatide, is actually pretty similar to other chronic medications that we sell. There was a period of time where it was higher, but that’s when we had shortages going on, and so people were actually having to discontinue because they couldn’t find the medicine.
All prescription drugs have untoward effects for some people. I think these drugs have been used for 20 years, and we have a lot of data to profile that, but people really should talk to their doctor before initiating these. Somewhere between 5% and 10% of people across our drugs discontinue due to side effects.
But untoward effects are not the most common cause for discontinuation in our studies. The most common reason people discontinue any chronic medication is that life gets in the way: some other medical problem where they discontinue the drug to manage that. Financial costs: more people pay out of pocket than any other drug class in America. So they’re paying their health premium, but they’re not getting covered for this drug. And that is difficult because these drugs are not cheap. And then I think people have sometimes experienced so much success that they want to try going off, and then they bounce back on, and go off, and back on.
GLP-1 DRUGS FOR … ALCOHOLISM? ASTHMA? BIPOLAR?
Thompson: One more beat on the frontier of GLP-1 drugs. There’s a lot of different directions I imagine you could push this class of medication. You can push on the weight loss effects. You can hold steady on weight-loss and try to build drugs that reduce side effects. And maybe a third direction is to accentuate a certain non-weight loss benefit of this drug category: a Zepbound for inflammation, a Mounjaro for reducing Alzheimer’s plaque. What are you investing in the most here right now?
Ricks: You’ve hit on three key ones. All three of those ideas are in the latest phase of testing for us. So in the next three years or so, we’ll have some answers. The only one I would add is convenience. People are interested in oral medications. We have that coming very soon. People are interested in much longer acting, “Oh, could I just do this once every three months or maybe on the future horizon, even once a year type of therapy?” So it’s more like getting your annual flu shot. That’s obviously very appealing to consumers. The technology is harder and harder to do that, but that’s another vector that we’re working on.
Looping back to what’s in the pipeline, retatrutide is our triple-acting drug. So if you think Ozempic, GLP-1, is single acting. Tirzepatide improved on that, Zepbound with dual-acting, so why not three? So we have a triple-acting single protein that’s in late phase development. We’ll get most of the data this year and submit it and probably launch it in ‘27. This promises to get between maybe 28%, maybe up to 30% weight loss. That’s the equivalent to gastric bypass surgery. Unfortunately there’s, I think, 8% of Americans have a BMI over 40. So even if you lose 25% of your body weight and your BMI is 40, you’re still classified as obese. So more weight-loss effect matters.
We’re also studying this category for alcohol use disorder and asthma. Asthma is an inflammatory condition. It is very co-morbid with overweight and obesity and it’s a major problem in the US. And so that’s another indication along with tobacco use disorder and bipolar disorder. So here we’re taking a drug that appears to have a little more brain properties that might be ideally suited for these, not necessarily just weight loss, but other conditions.
ARE WE VASTLY OVERRATING AI IN MEDICINE?
Thompson: Something I hear a lot from the folks building AI is that the risks of this technology are acceptable because in the long run “these things will be smart enough to cure cancer.” And I always want to know exactly how does that work. Like, I see these companies automating software development. I don’t actually see them developing new druggable targets for pancreatic cancer. So let’s say, the labs come to you—Anthropic, OpenAI, Google—and they say: “Dave, how do we cure cancer?”
Ricks: It’s a great question.
