29 – Exclusive Interview w/ AI Healthcare Pioneer – Dr. Tarun Kapoor of Virtua Health | BTS Analysis of the OpenAI Value Destruction

Marcus: [00:00:00] Episode 29.

Vic: 29.

Marcus: Okay.

Vic: Yeah.

Marcus: I should ask that before we hit record. Um, well, we, uh, have a great interview for you all today. Um, we have Dr. Tarun Kapoor, who is the executive vice president and chief digital transformation officer at virtual health. Um, Vic, you brought the story to us two weeks ago from modern healthcare.

Yeah. The virtual health was embarking in the world of AI and behavioral health. And we had a pretty spirited conversation about it. Yeah. And then our, our buddy, Bruce Brandis listened to the show, connected us via email with Dr. Kapoor. And, uh, we just spoke to him.

Vic: Yeah. I mean, I think this, the podcast is working very well.

Like we, we found this, this really innovator [00:01:00] doing, um, interesting stuff with AI, talked about it. Kind of in the abstract. And now we have, have them on the show today. It should

Marcus: be great. Yeah. So, uh, we won’t take up too much time before we let you hear the interview, but I will say, uh, it was an absolute pleasure speaking with him.

We’ll debrief more after the interview, but, uh, what a guy. And actually, you know, when we talk to people actually in the industry who are willing to have these kinds of conversations, especially with investors and innovators like ourselves, it does boost me up a little bit and make me more excited about, you know, working in this industry.

Vic: Yeah, we can debrief about it, but I think he’s doing everything that he thinks is right. And it’s pretty exciting.

Marcus: Awesome. So without further ado, uh, here’s our interview with Dr. Tarun Kapoor. All right. Welcome to the show. Dr. Tarun Kapoor from Virtua Health. Thanks for being here with us. Uh, thank you so much, gentlemen.

Appreciate it. So the way that this all came about is we did a show a couple of weeks back where we were covering some things in the AI space and [00:02:00] thankfully, no, you know, we don’t have zero listeners, uh, our buddy, Bruce Brandis, uh, who’s, who’s a great healthcare entrepreneur sent us an email and said, Hey, you know, you did a story.

You were talking about, uh, my buddy. Who, who rolled out that program at virtual health with the robot and looped you right into the email and said, you guys should talk and you were gracious enough to, to offer the opportunity to speak with us today, which we really appreciate.

Dr. Tarun Kapoor: I was, I really appreciate the opportunity to talk to you all as well.

And, and thank you for the podcast session. I, I really enjoyed it. Bruce pinged me and said, uh, take a listen to the whole thing, but definitely listen around minute 45. So I listened in to the whole thing. I was like, Oh, this is really interesting. And then. I said, well, wait a minute, how is this going to pivot into the stuff we’re doing with Wobot?

And you pivoted to the stuff we were doing with Wobot. And I said to myself, Hmm, I like their framing of it, but that’s not exactly [00:03:00] what we’re doing. And so I reached out and I said, Hey, just wanted to talk through, this is how we’re thinking about it. Because some of the things that you were discussing as your concern and friction points.

Trust me, I’ve kept me plenty busy thinking about this. So I just wanted to say, let’s talk a little bit more because I think there’s such an opportunity to think about this collectively, both from a healthcare provider’s perspective, but an entrepreneurial space and how do we meet somewhere in between.

Vic: Yeah, I think it’s, it’s great that you can spend some time with us. I mean, we’re trying to sort of read the tea leaves from afar and try to give a balanced, uh, report, but, but it’s really great to talk to you right on the inside and sort of understand how it’s actually working for providers, for patients, you know, live.

In a live setting, which is incredible. So thanks for taking the time.

Dr. Tarun Kapoor: Of course. So maybe one of the things that we had chatted about [00:04:00] offline that I thought was important to clarify and exactly to a very important that you raised was Gen AI versus the other techniques in AI. Yes. And I think there’s tremendous promise, remarkable promise for Gen AI in healthcare that we’re not there yet.

Marcus: Yep.

Dr. Tarun Kapoor: Not at least for straight out clinical use cases. And so one of the things that we’re trying to use. Specifically with the robot for, and, and one step back for people who, uh, didn’t hear the initial episode when we talked about this robot, is what would be considered a, uh, relational agent or adjunctive agent that a patient who’s suffering from mild to moderate depression can use when they’re offline from their clinician.

And in the use cases that we’re [00:05:00] gonna be using it here, let’s say a primary care doctor, nurse practitioner. Et cetera would see the patient say, well, I see you’re experiencing these symptoms. I think you would benefit from maybe seeing one of our therapists, maybe possibly starting medication. And I would also like you to start doing some exercises around a technique called CBT, cognitive Behavioral Therapy and the beauty of BCBT.

Is it changes your mindset of how you react to things. That’s where we’re layering robots. So the clinician is still in the loop, but a lot of the work is being done offline because you spend 99. 9 percent of your life outside the clinician’s office, or at least hopefully you do. And. And so that’s where we’re using WoBi.

But it is really more of a knowledge graph that’s supplemented with machine learning. It is not generating new insights or new [00:06:00] responses back and forth. It’s, it’s much more controlled, scripted, uh, interaction, uh, for now. I do think though, there are going to be use cases for gen AI. So I think a really important takeaway for me was when I’m speaking about AI and healthcare, I make, need to make it very clear what is gen AI and what is not gen AI.

So I think that was one of the learning for me.

Marcus: Yeah, I think, I think that’s, that’s a really important distinction. A couple of things I just want to frame up for, for our listeners, you know, Uh, Vic and I, we get on this show and we’re just two investors reading the stories and trying to sort of process all this stuff.

Um, but we, you know, what we’re trying to do here as investors is we’re trying to collaborate with the industry, um, to advance the way that we operate our industry and the way that we care for, for patients and, you know, even outside of the work that we do in venture, both Vic and I are, are involved in a majority in a variety of [00:07:00] different ways.

I’m on the board of HFMA, for example, and trying to work with, Um, providers, especially, uh, on, on how to really develop a culture that is more. Embracing of innovation and can absorb innovation. So I think the first thing both Vic and I want to do is just, uh, you know, tip our hat to you and credit you for being, uh, you know, a pioneer here and, and, and working on ways to integrate artificial intelligence into the clinical setting.

Um, and then also, you know, Want to also say thank you for making the really important distinction between all the different forms of AI that are out there. Um, you know, remote patient, uh, you know, uh, robots that are, that are processing, you know, from an automation perspective, machine learning, that’s working on very prescriptive sets of data versus LLMs, right.

That are generating things in a way that we can’t really totally predict and are, uh, And can have hallucinations or, you know, [00:08:00] provide feedback that is not necessarily correct, as I think was the situation that happened, um, with the eating disorder organization. Um, I think that was what sort of happened there.

Is that, is that your understanding of that? That yes, I think you referenced

Dr. Tarun Kapoor: the Wall Street Journal article, which talked about a mental health app and of note, there are, I believe, like 10, 000 mental health apps. That are out there in the ecosystem, you know, in the ecosystem. And yes, it was very scripted responses that they were originally having.

They switched over to an LLM model and the LLM started doing LLM stuff. And, you know, I think that’s the other thing that is really also important conversation in healthcare. There are different temperatures. For LLMs, you can have warm LLMs who are meant to be creative. Problem with creativity is that sometimes you can confabulate or you can have very cold LLMs and they don’t do a great job of a demonstration at a dev day, right?

[00:09:00] Cause it’s not super creative. It’s not able to convert something into. IAM a pentameter or wrap, but it’s gonna be much more factual.

Marcus: Mm.

Dr. Tarun Kapoor: And, and so I think we need to be also thoughtful in asking these critical questions. Where’s the temperature on your LLM? What LLM should you be using? If you even should be using an LLM?

Those are the thought provoking questions. I think good investors and good clinicians should be asking and having that dialogue.

Vic: Yeah, and Dr. Kapoor, I want to just sort of re, re go back to what Marcus was saying that in the reference of behavioral health in the clinic. So, so you’re one of your titles is, uh, I think digital transformation officer or similar.

Sorry if I butchered that, but the, the situation health systems find themselves in around a lot of disease states, but let’s just take behavioral health because we’re talking about that now is there’s more patient demand for treatment then really can be serviced [00:10:00] by the clinicians. And it’s very challenging to find new doctors and nurses and psychiatrists and counselors.

And so what I want to really understand for you is in your spot as digital transformation officer, how do you balance that patient care patient safety? Uh, responsibility is where I think it would fall. With trying to bring more, more available treatments, more frequent treatments, more services to patients, there’s a balance there.

And so maybe talk about that and then how you thought about figuring out which of these 10, 000 potential apps could be useful. Maybe the origin story of how you Decided this even would be something to look into. And then what was that process like to evaluate?

Dr. Tarun Kapoor: Let me propose a model of how we look at it.

Now, the problem with models is that you go to first principles, models are [00:11:00] not good things to use, right? You’re supposed to use net new, but if you don’t start with something like you’re just grasping at straws. Here’s how. I generally try to think about it, and then I try to see if there are exceptions that require me to change the model.

So just think of a straight two by two, and I’m not the one who created this, lots of folks have talked about this. On the one axis is the consequences of getting it wrong. And then on the other one is the volume of decisions that need to be made. So if there’s low consequences of getting something wrong, From a clinical outcome, regardless of the volumes, automation is actually a place to go after, right?

So in health care, let’s say, for example, I want to put an automation tool around patient payments. I’m not saying that getting an incorrect bill is not painless, right? We can apologize and say, we made a mistake, but the clinical consequence [00:12:00] is relatively low.

Vic: Yeah, patient safety is not a risk, for instance, not a risk.

Dr. Tarun Kapoor: So go nuts on automation in non in straight administrative spaces. The downside of that is from an entrepreneurial perspective is like it’s a very crowded space. And that’s not where the big pain points are in terms of staffing. So the pain points are on the side where humans and specifically very highly trained and usually highly compensated clinicians.

Have to be involved with, but there’s there for consequences. And so,

Vic: and that’s where the impact is probably, I mean, if you can do

Dr. Tarun Kapoor: it, right.

Vic: Correctly. A hundred

Dr. Tarun Kapoor: percent. So the fallback is, well, we’re going to still have a human in the loop. And I think that’s going to be a reasonable approach. For the next couple of years, but we’re going to start running through some problems and we’re already in some of them right now, and they’re only going to extend and exacerbate over the next couple of years.

And let me give [00:13:00] an example that probably Safavi’s work from Accenture talks about. In the next five to 10 years, there’s Sweden will need 25 percent of its workforce to work in healthcare in order to care for its population.

Marcus: Wow. Wow.

Dr. Tarun Kapoor: Finland will need 100 percent of its workforce to work in healthcare to care for its population.

Marcus: Wow.

Dr. Tarun Kapoor: So, if you’re always going to have a human in the leaf, sorry, in the loop, you’re not going to solve the problem in the long run. Oh, and there’s one other conundrum. Automation tends to solve the easier things first. It replaces tasks. What does that leave behind? The hard stuff for an already diminishing workforce who’s already burnt out.

So at some point we’re going to have to get to automation anyway, even with very complex. And the model that I’ve been thinking through as an [00:14:00] analogy, it’s a little bit further of us in healthcare is. Autonomous cars, right? So in 2017, RAND Corporation, which is not a bunch of bleeding heart liberals, right, said, if we move to autonomous cars sooner rather than later, we will be able to decrease the death rate by 50%.

And we have about 35, 000 to 40, 000 motor vehicle accidents every year, 90 percent of the, sorry, motor vehicle deaths per year. 90 percent are human related accidents. If we could cut that down to 15, 000 a year, shouldn’t we do that? But here’s the problem. Are you going to be comfortable if your family member died in a motor vehicle accident where the machine made the mistake?

And that is the societal conundrum that we’re dealing with with autonomous vehicle, which is going to [00:15:00] also tie right in, I think, into medical technology and healthcare. We know our clinicians make mistakes. Medical errors is one of the leading top five causes of death in the United States. If we can decrease that number by 50%, but it was an algorithm that made a mistake, not a human, can we as a society live with that?

And I don’t have an answer to that.

Marcus: It’s the right question to ask though. Yeah,

Vic: I love your, your automated car example. Um, because I, I, I use a related example of analog brakes a lot, but I got, I consider myself a good above average driver, meaning like I, you know, I, I drive fast and I think I’m have good response time.

I’ve got a ticket once in a while. Hopefully no, no police are listening. Um, but I want analog brakes. It has nothing to do with the skill of my [00:16:00] driving, it has to do with the computer can assess the situation and know to not lock the brakes and pump them much more quickly than a human could ever do. And I, I use that example with doctors a lot, because doctors and nurses and clinicians, they are the experts, certainly.

But there may be situations where a computer could do something and not get tired and evaluate a bunch of things that the doc decided what the algorithm should be, but then it gets implemented with that. And I really think your question is exactly the right question, really for our society. Is how do we balance this between a empathetic caregiver, a doc, a nurse, another extender making a decision that’s best understanding the entire patient situation with helping them scale and deliver sort of more repeatable, uh, outcomes.

That’s a [00:17:00] balance that we have to figure out as a society. I think,

Dr. Tarun Kapoor: yeah, in Vic and Marcus, I’m going to throw another conundrum into there, right? And that is exponential growth of knowledge across the board, but especially in the medical field, beginning of the 19th century, medical knowledge was doubling medical knowledge, doubling every 50 years or so.

So within a career, a clinician retrained themselves. The current number for medical knowledge doubling is around 45 to 50 days. No human being, regardless of how bright, how well trained, can keep up. To the point where I think there’s going to be some situations where we can make an argument, maybe not today, but maybe in the next year, two, three years.

You’re actually dangerous. Not if you’re not using machine.

Marcus: Right,

Dr. Tarun Kapoor: right. And there will be case law that will be established at some point where someone is going to file a lawsuit [00:18:00] to say, I’m suing this health system or someone else because they didn’t even use the machine.

Vic: Yeah, they didn’t know about some, some clinical study that could have been available to me, um, because they didn’t read that particular journal article.

Except there are thousands of articles that come out every week.

Dr. Tarun Kapoor: Example I’ll give you, we’re using augmented reality to, to assist our GI doctors when doing a screening colonoscopy. It’s an extra set of eyes. What do you have to lose

Marcus: right

Dr. Tarun Kapoor: when we demoed it to the executive team, number one thing that the executive team came up and asked me, Do I need to repeat my colonoscopy that I had a couple of years ago?

So the other thing is, it’s like, well, the clinicians say, No, no, my eyes are really good. And I agree with you. You get it. 99. 99 percent right. But what if it’s my colon?

Marcus: Right?

Dr. Tarun Kapoor: And so we’re starting to arrange this, [00:19:00] this or trying to create an colonoscopy. An environment where our clinician can not be overburdened and not be questioned for using the technology, but also feeling that they’re protected while they’re using the technology.

So there’s societal governmental regulatory pieces of it, but I think this is at least a way of thinking about it.

Vic: Yeah. So maybe give us like 30 seconds of your background. I think you came up as a hospitalist. Before you entered this role, talk about that perspective and then how does your system or any health system?

How should they think about navigating these issues? Um, given the regulatory environment we’re in and the the economic situation we’re in, where they, they need to figure out a way to make the clinicians more effective, more efficient, deliver better, better, and more efficient care. Not lose

Dr. Tarun Kapoor: money.

Vic: Yeah.

Not lose money.

Dr. Tarun Kapoor: Yeah. Rule number one, don’t lose money. If you’re losing money, see rule number two.[00:20:00]

We’re not going to cut our way out of this. And that historically has been traditional healthcare delivery models to say, well, we’re going to consolidate. And then we’re going to find safety in numbers. I do think we are a very interesting inflection point in health care delivery in the United States, where there are actually, for the first time, winners and losers.

It’s starting to happen. Now, it tends to be a lot of the, a lot of the losers have been in the smaller rural communities. And that is not a good answer either. But now you’re even seeing those losers in the reason population deaths were historically well doing areas. So What we take approach here at Virtua is a model that came out of Silicon Valley, and that is we go ahead and we try to make a funding decision [00:21:00] internally to innovate with purpose.

It’s just not saying let’s run a whole bunch of experiments and see what sticks. What is the problem that we’re trying to solve? Our team, our digital transformation office, our team, we will partner with a traditional operator.

We try to help you figure out in six months, we bring the funding. If we can prove a concept that we can make this thing work after 12 months, 18 months, it goes into your operational budget and your new way of doing business. If you don’t like it after that window, we kill it. Something health systems aren’t really good at doing is stopping pilots.

Marcus: Yeah.

Dr. Tarun Kapoor: And so there have been pilots that I said, this is such a great idea. The timing wasn’t right. We were just. Too far ahead of the market on that. And so you come back to it again, again, and again, [00:22:00] and then maybe three years later, that was the right idea. If you are not doing that, you’re going to get left behind.

It’s just the reality.

Vic: And so what is the discussion like with the leader of the practice? Maybe it’s a behavioral health practice. Maybe it’s a primary care practice. How do they frame it for you? Before you know that’s going to be Wobot, we’re going to go find a tool. How has that discussion held? Or what, what are the, what are the benchmarks?

So how do you think about an RFP process or how do you evaluate these things?

Dr. Tarun Kapoor: First thing we look at is when I work with the clinicians, what is the problem you’re trying to solve? And then we ask them five more times when they tell the problem. I said, well, is that really the problem? But once we get to an idea of what the problem we’re trying to solve here, and with Wobot, for example, we’re saying, we’re not trying to cure Depression and anxiety without your help.[00:23:00]

You already have seen these patients. You want your patients to get better faster. That’s what the problem we’re trying to solve. Then why don’t we try to accelerate the healing process for all the time that they’re not with you in the office. So we change the argument as saying, it’s not here to replace you.

It’s actually here to make the next visit even more productive because the patient has progressed. The patient likes that, the clinician likes that, right? Everyone feels good about that. Then the other piece that we started working through is how fast can we get to market? And this is, I think, the key thing that I would tell the audience to take away.

Just something, just because something may not be perfect, still mean it can be safe. A lot of times in healthcare, we hide behind the Latin phrase, primo non nocere. First of all, do no harm. But [00:24:00] doing no harm and not doing something are not synonymous. And that’s where a lot of people hide. So as long as we can prove that it’s safe, then we can try to follow efficacy.

That’s just exactly how the FDA does it too. First, they look for safety. Then they look for efficacy, which is why I think you’re seeing so many parallels between digital therapeutics and pharmaceutical therapeutics, right? They’re trying to take that approach. We could talk about pair and where it went sideways with pair, but that’s how we try to look at it.

The number one thing I always talk about with physicians to understand What is their fear point? What about you? What about this problem? What about the solution worries you the most? And invariably, it’s fear that this problem will lead to a bad outcome. What if we prove it with five patients that we watch like hawks?

That we want to have a bad outcome. Then can we go to 10? Then can we go to 50? [00:25:00] That’s how I recommend working with health systems.

Marcus: I’d love to, I’d love to actually take you up on that and talk a little bit more about digital therapeutics. Uh, we, we did a whole episode on the pear implosion. And, uh, got a lot of feedback on that and ended up having a friend of ours come on the show, uh, Aaron Ghani, who’s the founder of, uh, Behave VR, um, and in a similar way to you help clarify sort of the situation in the digital therapeutic space and sort of what we left that that conversation.

Understanding is that there is a lack of framework between the FDA and CMS on how to actually position package and then ultimately code digital therapeutics. So I’m curious in the virtual health digital transformation framework, um, where you’re not necessarily. Looking just for, you know, getting a new new code where you can rebuild something, but you’re, you’re looking to sort of change the [00:26:00] overall care delivery model.

Um, does that open up a system like virtual health to digital therapeutics in the way that. An organization that doesn’t have that sort of digital transformation lens can’t really get there because they need a code attached to it. Um, or do you have the same sort of barriers to deploying digital therapeutics where you, you need the codes there to sort of really make a business case for it?

Dr. Tarun Kapoor: If you can’t get paid to your solution, this is not going to end well, right? That’s just the reality of it. What I’m most excited about the project that we’re doing with the robot is that this is the first time that a health system, a provider is saying, I’m going in on this to actually see if we can prove a return on an investment that we will then go to the insurers and say, we’re vouching for this.

We actually believe this is making an impact and we’ll decrease your total cost of care. [00:27:00] And so you’re bringing a clinician or a group of clinicians to the group and say, look, this is what we’re in here. That’s the partnership between us and Wobot that doesn’t get, I think, a lot of press on the innovation.

A lot of the work has already been done on building the therapeutic out, right? They’ve done 12 randomized clinical trials, you know, millions of minutes of patient interactions. The innovation here is on the business model. That, I think, is what is most exciting. And if we can prove this with the initial 500, maybe eventually 1000 patients, then get a code in a fee for service environment, in a capitated environment where you’re already managing the total cost of care.

If it works, go after it. Right. And if I knew I had more of these tools that I should manage total cost of care, well, I’m more open to taking risk. Right. I think a lot of problems with the health systems out there is like, oh, I’m going to go [00:28:00] take risk and then go figure out if I can manage it. Don’t go in a casino if you don’t know how to play blackjack.

It’s just not going to work at all.

Vic: So let’s, uh, double click maybe on that, um, use case or business model. You, I love your two by two, uh, kind of quadrants. Is it right? It was, it was our assessment from the outside. That this was going to be more of a low acuity, um, fairly educated patient that is not, for instance, not a suicide risk, or maybe they have low levels of assessed, uh, anxiety and depression on intake, but the clinician, the human puts them at a, at a minor, um, low triage, low acuity, low triage.

Is that fair? And is that a part of the calculus? Adverture, how do you think about, uh, sort of segmenting the population?

Dr. Tarun Kapoor: That’s exactly how we’re [00:29:00] starting with it. Like I told you, there was a lots of conversations around, wait a minute, the patient with severe depression. We mean, we’re going to exclude them.

Marcus: Yeah.

Dr. Tarun Kapoor: Aren’t the ones, aren’t the ones who need the tool the most.

Vic: And so, but the consequence could be much different with that patient. So that, that’s why I like your two by two model. Like it could really help that patient, but also there’s huge riding on, there’s a lot riding on that. On that interaction.

Dr. Tarun Kapoor: There is. But the other question though, I think about is how is that patient communicating for help today anyway? Maybe they’re posting on their social media network and who’s even watching that? Maybe they send us a message, they leave a call, phone call and say, Hey, I’m not feeling good. I’m having these interviews.

We may not get that message for a few hours. At least we have a relational agent that’s actively listening word by word and can accelerate and say, Hey, I’m noticing this. We need you to [00:30:00] accelerate help, but you have to start with somewhere. And then if you can rapidly iterate to expand, then you start to get that viral adoption.

Vic: Yeah, and might it, I mean, this is me being a VC, right? So VCs are, um, we quickly become experts even though we have no knowledge in things. So that’s a, but, um, is it fair to say that this could open the aperture of, the number of patients one clinician could bring in, and then it might highlight the subset that need, you know, twice weekly sessions.

And there might be another group that is, you could come in once every two weeks. And the tool could help, help modulate that or help make sure I’m in the right category. Is that fair? Is that a future world that could be?

Dr. Tarun Kapoor: Absolutely. And so I’m going to tell you this little dirty circumstance of medicine when you get hospitalized, what’s in your discharge instructions, see your primary care doctor in seven days, totally generic information.[00:31:00]

Maybe you should be seeing your primary care doctor in 24 hours. Maybe you should be seeing the cardiologist in 48. Maybe you don’t need to see me for a month. If we can start to really understand where you are at any given point in your journey and then get you the right level of care, we’re now making the usage of the very limited resources we have way more effective.

And in some of these cases, and this is a hard one for health systems, We may need to see you at all.

Vic: Yeah. The goal is you don’t want to see me at all, but in a, in a fee for service world, that’s a challenge.

Dr. Tarun Kapoor: Yeah, it is a challenge, but am I also going to, if I’m going to see you or something very basic that we could have used in a different environment, whether it’s in fee for service, whether you’re capitated.

You want to be [00:32:00] using the limited resource you have to the maximum of capability. If we’re not doing the stuff that only we can do, I don’t know if that makes a lot of sense for us to be the jack of all trades. I think focusing, I, when I was in my first startup, person who mentored me said to me over and over again, no one, one of it isn’t.

No one went out of business because they focus too much.

Vic: That’s a, that’s a great, uh, thing to talk through. So, um, I wanna, I wanna dig into your, uh, your interpretation of how for profits and non profits can work together in this loud environment. Of course, we’re all following the news of OpenAI, the board, um, dismissing the CEO.

Um, and it strikes me that, that you have a different, an interesting perspective on this because you’re in a nonprofit virtual health, working [00:33:00] with a for profit robot, trying to create a business model for your practices. But also for Wobot at some level to make a make a lot of money and make a big impact.

So where’s the, um, where’s the line? How do you think about nonprofits and for profits working together, particularly in health care? Um, we see it happening in sort of AI and whether it’s You know going to come back together or not. It’s it’s it’s showing me that the tension between those two

Dr. Tarun Kapoor: I don’t think they have to be mutually exclusive mindset even not for profits have to be Able to generate some type of contribution margin mark.

Vic: Yeah. You got to pay your people.

Dr. Tarun Kapoor: Yeah. Right. No margin, no mission as the statement goes. And so we also have [00:34:00] these constraints that completely align with the true entrepreneurial for profit world.

The big difference is that we don’t distribute earnings to our shareholders. We reinvest back in their community, but our communities are shareholders. So we’re doing our best bet there.

I actually look to see where’s their synergy or alignment on the problem we’re trying to solve and the ability to solve that problem. The other big thing I think that I spent a lot of time looking at is what is potential longevity of the startup versus the health system because the health systems, especially the not for profit health systems, typically the largest employers in many situations in their community.

If we make really big, bad, risky bets. To the point where we start to impact the employment status or ability to [00:35:00] employ in our local community, that can have a devastating impact. Not for profit SPAC didn’t work out so well. Right. So that’s where I look for some synergy of, or how to balance that out. The downside with that is, is that not for profit health systems were going to be drawn to the really large players.

And so I think there’s a balancing act of say, I’m willing to work with someone smaller, who’s venture backed. We may not have all the resources in the world, but I need to know they have a viable pathway forward. Are they reasonably capitalized in your world? You may say they may be over capitalized for the relative the size that gives me a tremendous reassurance that I’m not seeing something that they couldn’t make payroll.

Right. Yeah. And my product is [00:36:00] gone. Right. That’s how I would think about it.

Vic: Yeah. It’s, it’s interesting. It strikes me that it’s, it’s a, um, alignment. Maybe it’s all true in all business, getting alignment on what are we trying to achieve. And making sure that we’re aligned, whether it is two for profits or a nonprofit and a for profit, uh, at some level that can be the answer if you get that alignment, but there’s a lot of discussion that has to be talked through, but.

What are we trying to achieve? And then I think you’re absolutely right. If this becomes an important part of the way you treat behavioral health, you want to make sure that this company can, can support you and isn’t going to disappear. So I think that’s helpful.

Marcus: So, uh, I think we’re, we’re, we’re going to let you get back to work, but I didn’t want to give you an opportunity to sort of leave our audience with, um, You know, maybe some insight and also some optimism for, for 2024, um, big year ahead on, on, on a lot of different fronts, uh, the [00:37:00] technology landscape is going to totally explode.

We, we are, we can pretty safely predict that based on how the last, uh, 30 days have gone in the technology world. Um, but also the healthcare world, right. I mean, is, is under tremendous pressure. Um, and, and I think you’re, you’re definitely setting an example for how Uh, you know, quite frankly, nonprofit health systems can navigate these very, very challenging times ahead.

So what would, what would sort of be, you know, your, your message for our listeners?

Dr. Tarun Kapoor: So I should have done a better job and I had asked chat GBT to give me 10, five things that a good healthcare executive should give for their year end summary. Uh, personally, I’m very excited that we’re finally getting into a place in healthcare delivery.

We’re able to use some of these tools to meaningfully close some care gap. Give you a quick example. We found 12, 000 people in our database who needed a [00:38:00] colonoscopy based off of their age, just don’t want to get it right. So we reached out to them, use our internal data to say, here’s an alternative test that you can just do at home, send it in to us.

A couple of text messages, sent them a kit of those 12, 000 people. 3, 500 of them responded, nearly 300 of them have positive results that we are now actually getting in and helping them resolve and cure. This is a lot of fun now, right? All these tools are coming together to have meaningful outcomes and impact.

I’m also really excited about some of the tools to make our life. Our lives easier as clinician for each other, because all we’ve done for the last 15, 20 years, it seems [00:39:00] sometimes it’s just send it to the doctor, send it to the doctor. And now we can say, doctor, I can actually help you get through your day a little bit better and let doctors and nurse practitioners and clinicians do what they do best.

And that is the hyper technical or be hyper humanistic. And in some cases, both, those are some of the best things I had tremendous headwinds, no doubt. And so this is maybe where I’m, I guess, what are the two A’s electric? Uh, the altruist and the effect of this. So, yeah, yeah, yeah, yeah. So, uh, I don’t think it’s an either nor.

Okay. Well,

Vic: that’s positive. I think that’s great advice. And I just really want to. Say thanks for the work you’re doing really pioneering and blazing a trail of how can we bring? Technology tools to help clinicians and help patients. It’s incredible work [00:40:00] and there’s hundreds of Healthcare leaders that need to follow.

So, uh, hopefully they will follow your lead.

Dr. Tarun Kapoor: I’ve learned from others I just want to for continuing the conversation And that’s the only way we’re going to get better or for the title of your podcast, take health further. Yeah.

Marcus: Amen. Look, stay in touch. We’ll be checking in to see how everything is going and what you’re, uh, what you’re up to in 2024.

Vic: Yeah. When, when’s the update point? When should we check back? Is it a six month? Is it 12 months? How, how long are you thinking about running it? Like this first part?

Dr. Tarun Kapoor: Oh, we are really hoping to see results in the next three to six months. So we’re just launching now, but glad to, uh, check with you maybe later on Q2.

Marcus: Yeah,

Dr. Tarun Kapoor: we’ll circle back

Marcus: and see how it’s going to tell you what we have. Fantastic. All right. Have a good one. Happy thanksgiving. All right. We hope you enjoyed that Uh, we’re gonna take a quick break and let doug tell you about jumpstart foundry and then we’ll be back to debrief our conversation

Doug Edwards: Thanks guys for the opportunity to talk about our [00:41:00] precede fund jumpstart foundry My name is doug edwards ceo of jumpstart health investors the parent company of jumpstart foundry We’re so excited to be able to talk about, uh, early stage venture investing.

Certainly the need for us to change the crazy world of healthcare in the United States. We are spending 20 percent of our GDP north of 4 trillion a year on healthcare with suboptimal outcomes. Jumpstart Foundry exists to help us find and identify and invest in innovative companies that are going to make a difference in healthcare.

In our country, every year, Jumpstart Foundry invests a fund, raises a fund, and deploys that across 30, 40, 50 assets every year, allowing ease of access for our limited partners to invest to help us make something better in healthcare. Some of the benefits of Jumpstart Foundry is there’s no management fees.

We deploy all the capital that’s raised every year in the fund. We find the best and brightest, typically around [00:42:00] single digit percentage of companies that apply for funding from Dumpstart. And we invest in the most incredible, robust. Innovative solutions and founders in the United States. Over the last nine years, Jumpstart Foundry has invested in nearly 200 early stage, pre seed stage companies in the country.

Through those most innovative solutions that Jumpstart Foundry invests in, we also provide great returns and a great experience for our limited partners. We partner with AngelList to administer the fund, making that ease of access, not only with low minimums, but the ease of investing in venture much better.

We all know that healthcare is broken. Everyone deserves better. Come alongside us with Jumpstart Foundry. Invest in making the future of healthcare better. And make something better in healthcare. Thank you guys, now back to the show.

Vic: Alright, Vic, what’d you think? I liked him. You know, I think it’s um, Listen, being in charge of digital transformation at an academic medical center like [00:43:00] Virtual Health is a hard job.

There’s a lot of, um, incumbent thinking and decades of practicing medicine a certain way, and he’s tasked with transforming that. And I thought he was really open and, and great about like what, what he’s trying to accomplish and how he thinks about it. And seemed pretty, pretty open to trying to figure out what the right answer is.

Marcus: I was really impressed. I mean, a couple of things. Uh, first I was really impressed with his depth of knowledge on. AI and just the different versions of AI and being really clear and specific at the very beginning of our conversation that LLMs are not appropriate for the clinical, uh, theater today. I thought that just getting that off the table so we could have a deeper conversation was great.

Um, I thought the whole thing about the virtual approach to digital transformation with, in terms of their business model, how they’ll go in and try to take three years of, of change and accelerate it into six months. Yeah. And then [00:44:00] turn it off if it doesn’t work or, you know, keep it going and fully fund it as a new business model.

If it does work. Um, I thought that was really, really unique.

Vic: Yeah. Just having a leader that understands that we’re going to try maybe 10 things in the next two years and three will work three, we will have to do more work on and three just won’t work. And we’re going to shut that down and keep going.

That’s what we need as as VCs I think that’s what founders and management teams of startups need is clarity and a partner to work through the business model and figure it out even if it’s. Not the right answer. Clarity is what’s really needed.

Marcus: What, what do you think, um, what do you think is the likelihood that this pilot that they’re doing with Wobot, um, will actually be able to get real movement with payers?

I thought that was another really interesting part of what he shared was that they don’t yet have codes for this. [00:45:00] So they are partnering to demonstrate that in a, you know, two, three, four, Total cost of care model. This actually improves care for the patients, lowers the total cost of care and, and just even, you know, give me your feedback on what you think the likelihood of this actually making the payers, uh, sort of adjust their, their thinking around how they would pay for this form of care.

But then secondarily, I would love it. To know your thoughts on just the model of health systems partnering with digital therapeutic companies to prove efficacy and then make a case to payers. Like I thought that was, um, that felt like a big deal for the whole digital therapeutic space, right? That we have health systems that are willing to partner with innovators and then collaboratively go to payers and make a case for value.

Vic: Yeah, I think that is. A great model that I honestly had not seen before. Drew educated us about it. I think it makes a lot of sense to go sort of in partnership, a large, [00:46:00] large ish health system with a tech startup as a partnership saying, here’s the outcomes that we have done kind of on our own account with our own money and bringing results to the payer.

I think that model has a lot of promise. How it will be received from the payers is. I know, but, but hopefully they will look at the data and if it, if it delivers better outcomes at lower costs, they should fund it.

Marcus: Yeah. I felt while I was listening to him, like. Hey, I was, I was watching a star rising.

Like I kind of feel, I kind of feel like this guy is going to be big deal in the healthcare innovation space. Like, like a lot of people may not know him yet, but I think more and more people are going to get to know him.

Vic: Two weeks ago. And then I had never talked to him before today. And he’s great. Um, and it’s pretty, it’s pretty hard to sit in that transformation seat [00:47:00] and Work with all the physicians and he, he, he characterized it very well.

They’re, they’re trying to deliver care. There’s all this new research coming out daily. That’s very hard to keep up with. And how can he be kind of an advocate for them as well as an advocate for new technology innovations that, that need an audience need to be tested. Now, whether they work or not is like I say, three out of 10 will work.

Maybe, maybe. Three more will need to be iterated and improved on, but just having someone who can bridge that gap, it’s, uh, there’s not a lot of communication between the existing infrastructure and healthcare systems and the innovators that, that understand the technology, but, but really maybe don’t understand the day to day life of a, of a physician.

Yeah. So I think it was phenomenal. And I think he’ll get, um, I’m sure there’ll be people criticizing it, but, but I, I was really impressed with his approach and kind of how he’s thinking about it.

Marcus: What did [00:48:00] you think about the case that he made to, he made, I thought two very compelling cases, one case, which is that we ultimately really need to get to automation in the clinical setting, um, because of workforce, because of expansion of medical knowledge, because of the Out of whack ratios from a labor perspective versus aging population.

So I think he made a compelling multi layered case for the fact that while it’s, we’re not ready today, we are going to need to get ready because there’s no other way to care for our communities if we don’t do that. And then at the very same time, positioning sort of the psychological challenge of humans having an acceptance of the fell ability of.

Of other humans and being okay, the number of deaths that we have, I mean, you know, we’re not okay with it, but we understand humans are fallible, but even if we were to improve the metrics, are we going to be okay with [00:49:00] machines being fallible? So, like, I thought that there was an interesting crossroads we’re coming to where just logistically from a viability perspective, we’re going to have to automate more things in the clinical setting.

And at the very same time, how are we going to get comfortable with, with the reality that we are going to likely improve things overall and still have mistakes that lead to death?

Vic: Yeah. I mean, listen, in a, in a healthcare system, you’re going to have patients that, that pass away. I mean, that’s just part of treating really sick, high acuity patients in a healthcare setting.

And I think that is, there’s a social dynamic that we need to restructure in our society. If, if, if a human error causes, You know, natural, um, there’s gonna be 1 percent of this procedure that ends up in a bad outcome, and we’ve lived with that [00:50:00] forever, and that’s just part of the delivery of care. If you then shift to half that, Half a percent will have a bad outcome, but it’s a technology based thing.

The liability, uh, the legal liability structure and kind of whose decision was made, how do we, uh, place blame? How do we decide when to implement that or not? We just don’t have the. Language or the legal structure or the the payer and provider infrastructure to do it. So I think part of what we have been saying has become more crystallized to me that the existing structure of.

Health systems deliver care and then payers adjudicate whether that was right or not and pay for the procedure when it’s right, that is outdated and needs to change. And so whether it is a sort of a holistic, um, value based system where you have all risk under one platform, [00:51:00] um, or there’s some other combination.

Basically, I, I took away that we need new, new structures, new legal frameworks, new payers frameworks, new ways to deliver care in order to take care of our population.

Marcus: Yeah, very much. Um, anything else you want to say just as in the debrief before we move on to the news of the week?

Vic: No, no. I mean, I, I, we’ve already said it.

I’m, I’m excited to see how it goes. I think he said he would give us an update in six months or so. So it’ll be good to follow it. And I think part of what I want to do on this podcast is find innovators like, like him that are doing great things around the country. And then get to know them and follow them and try to learn from this.

We can take best practices to other, other health systems.

Marcus: Yeah, I’m, I’m definitely smarter for, uh, having had that conversation. So thank you, Dr. Kapoor. And thank you, Bruce. We got to have Bruce on the show as well. He’s doing some really cool stuff as well. So we got to catch up with him and get him on the show.

Vic: Yeah, I think we should, we’re going to [00:52:00] have more guests bringing in outside perspectives is good.

Marcus: Yeah. Don’t spoil it though. That’s, that’s, that’s part of our surprise for the next year. All right. So, uh, yeah. Quickly, the White House continues to be very, very active, and, uh, Vic, you pointed out that we’re basically a year away, um, from Election Day, so it’s time for them to start packaging things up, and so, uh, they released the United States playbook to address social determinants of health, um, this is the second big announcement that they’ve come out with in, uh, Two or three days, uh, on the healthcare front.

Um, the last one we covered last week was the women’s health, uh, research announcement. And, um, basically I think we looked through the actual PDF and it felt to us like it was largely packaging a philosophy, the existing. Body of work that they’ve done, um, you know, over the course of their administration, that, that kind of falls into the buckets of addressing housing security, food security, uh, education, access, and a healthy environment.

And then they provided a playbook of actions, [00:53:00] three pillars. The first expand data gathering and sharing. The second pillar is support flexible funding for social needs. And the third is support backbone organization. So it’s kind of like. You know, a thought leadership, white paper. It’s not, it’s not really an executive action.

It’s kind of like a framework that feels like it’s part of a policy framework that they will use in the campaign trail.

Vic: They’re laying out their healthcare platform. Right. And I think that’s probably smart politically. There wasn’t a lot of new ness in this. And SDUH was all the rage five years ago, maybe eight years ago.

And I think it’s fairly consensus now that these factors are determinants of health and we have to address all of them. Didn’t feel that new to me, but it is sort of laying out the Biden administration’s work to date and where they want to go, which is, which is You know, net positive.

Marcus: I mean, I think, I think frameworks are sometimes as helpful as [00:54:00] actions, right?

Help us understand how you’re thinking about this. What are the structured steps or actions that you are taking and that you want to encourage others to take and what’s the guidance you’re providing for your, you know, canon of agencies, uh, to go after this. So I, I think, I agree. It’s not. It’s not action oriented per se, but the framing of it such that we can all have, you know, scaffolding to have this conversation and to understand what, how they’re thinking about it and maybe how this might filter into a set of innovative actions that companies will take the CMS might package up into codes, you know, that’s, that’s all pretty helpful.

Vic: Yeah, I mean, I, I think it’s like, like research into women’s health last week. These are initiatives that, I mean, it probably has 95, 100 percent approval ratings, right? So it’s a smart political thing to do. Yeah.

Marcus: Yeah.

Vic: It’s not that, uh, there’s not a lot of new there, but, but positive.

Marcus: All right. So. Now on to the [00:55:00] story of the week and, you know, potentially the story of the year.

And who knows in the future, this could be a historic moment. And I feel like every time we’re talking about AI and specifically we’re talking about open AI, I keep saying like historic, historic. Uh, but I just feel like I’ve got pattern recognition around. Massive shifts in, in society as it pertains to technology.

I’d certainly believe LLMs and especially the open AI version of LLMs represent that. And, uh, my gosh, we have had a week at open AI. So. Okay. Do you want to just start by trying to recap it? Because I, here’s the thing. I don’t want to assume that all the listeners know exactly what’s happening. I was talking to my wife and I’ve been basically keeping her up to speed, you know, and she’s very aware of what’s happening in technology.

So let’s maybe start with a rundown to get to this. Yeah. So

Vic: let’s, let’s do kind of a fact based, uh, over the last week or so [00:56:00] without a lot of, um, conjecture. Yeah. So yes. The. OpenAI developer day was 10 days ago, 14 days ago. We covered it here. Um, I believe that, um, unveil, which we talked about lots of new products, really powerful products,

Marcus: GPT, four turbo, custom GPTs, lower prices, giving

Vic: out like API and the GPT to make robust power to lots of, um, you know, People that don’t work at OpenAI to use and work with.

Yep. We were excited about because it was clearly powerful. Um, it seems like that was the last straw. Whatever the reason is, the board of OpenAI, which is an independent board, There’s, um, four members. They didn’t, they’re not [00:57:00] investors. Um, they are in place to guard the nonprofit mission of open AI. The, the structure is, is a nonprofit foundation and its mission is to bring AI to humanity and protect us from AGI, uh, which it would be intelligent, an automated intelligence that is smarter than humans.

Because

Marcus: artificial general intelligence, which is the same kind of intelligence humans have, not a specific intelligence, but general intelligence.

Vic: Yes. And so I don’t think, I don’t think anyone believes that. Chat GPT or the LLMs are there yet? No,

Marcus: they do not represent AGI as we have experienced them.

Vic: That’s right. As has been shown to the world so far. Yes. But the, the trajectory, the capacity and capabilities of the large magnus models [00:58:00] is, as we’ve talked about, has increasing very quickly. Yes. This board is in place. To try to protect from really bad outcomes like it turning on humans and killing us all.

Marcus: Okay. So so

Vic: so there’s a there’s a foundation then there’s a non profit Subsidiary, I mean, there’s a for profit subsidiary llc That has been charging people twenty dollars a month and making all this revenue And has done the deal with microsoft and lots of other vcs and they You Um, they fired Sam Altman Friday afternoon because, in their opinion, he was not being forthright, not communicating openly and honestly with the board.

And like all CEOs, he reports to the board and they have the power to fire him.

Marcus: And so they

Vic: did, so they did Friday and he, uh, was [00:59:00] shocked, I think, and was on Twitter X, the whole

Marcus: world was, let’s just, I mean,

Vic: yes. And there was a effort for, um, them to take him back. And Microsoft was pushing that, VCs were pushing that, that all invested in the subsidiary.

Um, and then they refused to do that because Sam, I think, demanded the board be replaced. As any CEO would after they fired him. And then they brought in a new CEO Sunday, I think. And simultaneously Microsoft extended an offer of employment to, to Sam to join Microsoft. And then Sunday or Monday, A letter with about 700 of the 770 engineers, developers signed by 700 of them.

So like 90 percent of 80 percent of them, uh, saying that they would leave the [01:00:00] business. If Sam wasn’t reinstated, I think that’s the summary of the facts. Yeah. I think, I think, I

Marcus: think the only thing you may have left out was that, uh, the president of the organization, is it Greg Brockman? Is that his name?

Yeah. Um, he quit, um, he quit the board and he quit the company with the board. No, no, no. You’re talking about Ilya. Ilya. No, no, no, no. Yeah,

Vic: so let’s distinguish that.

Marcus: Greg was alongside Sam. He was not fired, but he quit. He quit. He summarily quit. Ilya, who is the chief scientist, was on the board and was part of the vote.

Vic: Yes, and he has Gun back and said, he’s sorry about that vote and

Marcus: signed the letter

Vic: and signed the letter, right?

Marcus: And there are various threats from the investors who invested in the for profit open AI business to sue the board and the [01:01:00] foundation. Is that correct?

Vic: There’s lots of threats, lots of threats

Marcus: flying around, right?

Because, because there there’s, Assumed serious value destruction that has happened at open AI. And that was at one point, maybe Friday reflected in Microsoft stock price. And then over the weekend, Satya Nadella pulled, pulled another all time great move, brought Sam Altman into Microsoft. And then Microsoft had their all time best day

Vic: on Monday.

Yeah.

Marcus: Um, and so we are recording this Tuesday.

Vic: Yeah. With the holiday early

Marcus: evening.

Vic: Right. Yeah. So it’s changing quickly. Um,

Marcus: by the time this goes out, I mean, everything we said could be reasonably outdated.

Vic: Yeah. So those are, those are the facts as we know them now. I think we should talk through conjecture or what we, that aren’t facts, but what we, what we think, I mean, this article in the Atlantic that we’ll post, [01:02:00] it may be behind a payroll, but it’s,

Marcus: it’s behind a paywall.

So if you want to read it, you’ll have to pay.

Vic: There are a hundred news stories. But I think by far the best reporting on this is by the Atlantic, and they go back, uh, to before the release of chat GPT, and then sort of talk about all the, the sort of circumstances leading up to Friday. Um, and it’s pretty good reporting.

They have, they have 10 sources. In the company, they’re not named because they wouldn’t give their name, but it seemed like a very well researched, good news article. Um, and in the news article, they make the case that the board took their job of protecting humanity from AGI seriously and began to get concerned when they released the first chatbot, which was amazingly [01:03:00] not quite.

A year ago, so it was on november 30th last year. So less than a year ago Um, and it was released. Uh, we should probably get the quote but as a low profile research project Um, and a lot of the reason it was kind of rushed to market is their competitor anthropic Was about to come out with their chat bot and they wanted to get in front of that.

Um, and so the low key research preview there. Um, and so this was the biggest and highest profile product release in the history of product releases. Now, whether that was intentional or. Or unintentional, it was not, there was nothing low key about it. Um, and so [01:04:00] I think the article is suggesting that the board was nervous about the lack of control that opening.

I had over this. So they have a tool that is supposed to be monitoring the traffic and seeing, for instance, are people using this for malicious purposes. Are they trying to figure out how to create a bomb out of kitchen supplies or what, you know, other stuff like that. They built a tool that could supposedly follow the traffic and try to find bad actors and cut them off.

And the tool was designed for like a hundred thousand users at a time, because no one thought they’d have that many users and it, and it, it was, it’s a hundred million a week now. And so their safety controls were completely swamped. And there’s been a group inside open AI that, that is concerned. And I’ve heard, you’ve probably heard, there are people, there was a letter signed.[01:05:00]

There are people that in the. in the world that are worried about AGI getting out of control. I don’t have those concerns, but it, but it, there are serious researchers that have concerns about that. So it is a thing and OpenAI was designed to protect against that. And so then they released GPT 4, which again, was not, the board didn’t want to release it.

And so I, I think my guess is that they saw sort of the power that The for profit sub was getting and felt like if they didn’t stop it now, they were not going to be able to fulfill their mission. Now, whether that was right or not, uh, the board is pretty powerful and I think it’s going to be a challenging lawsuit.

So.

Marcus: Yeah. So, I mean, I think we have a [01:06:00] very complicated story because we have a nonprofit and a for profit and a groundbreaking revolutionary technology. In our hands. That’s that that’s kind of the recipe that creates what we’ve experienced over the last four days. Um, a nonprofit that has an altruistic charter of protection in mind.

Um, with a technology that look, we’ve made movies about this and not just one, right? We’ve made a lot of movies about this potentially being what ends the world as we know it. Um, so. I for one, appreciate the idea that someone is. Trying to make sure that this thing doesn’t get totally out of hand. I think safety is appropriate given the, the potential power of AGI.

Um, but commercial actors are at play. It [01:07:00] takes money to run these kinds of things. You’re going to get commercial interests if you need money to, to advance things. And I think that the. You can’t serve two masters. And so I think the idea that this, you know, most powerful version of AI in at least the Western world, uh, was being built in a house divided nonprofit on one side for profit on the other side with Microsoft hanging sort of in the wings, um, staking a major part of its future.

On this, it was, it was kind of inevitable that some type of fissure was going to happen. Uh, I think the way that it happened was totally surprising, but now let’s just talk about the implications of Microsoft potentially Aqua hiring Sam Altman and. A ton of these engineers without having to buy the company.

I mean, Satya Nadella has made incredible deal after [01:08:00] incredible deal and really resurrected Microsoft, not just in stock performance, but I think in brand sort of eminence. I mean, the brand of Microsoft is a great technology brand again, and it wasn’t for a long time. Steve Jobs almost buried that brand.

Vic: No, no, uh,

Marcus: No. Steve Jobs. Oh, Steve Jobs. Steve Jobs, Steve Apple sort of, yeah. Steve Jobs. Yeah. Not, not Steve Ballmer, Steve Jobs.

Vic: Steve Balmer sort of just like kept it going. Yeah. He just kept it going

Marcus: while Steve Jobs was burying it. Right. Right. And this guy’s brought this brand back to life and he hasn’t done it all through internal effort.

He is done it much through m and a. Yeah. Right. And GitHub was, was brilliant. GitHub, LinkedIn. Yeah, LinkedIn. Activision. Now he’s got Call of Duty and Open ai. I mean, just look at that cohort of brands of technology brands. Yeah. Um. And now the idea that he may be able to get the brain trust without having to buy the asset.

It’s just, it’s just [01:09:00] mind blowing. And I think the other thing is what other situation would create an incentive for Sam Altman, who clearly has the heart and mind of this team to go with the incumbent rather than to go take a bunch of VC money and do this on his own at unbelievable terms. And the answer is, Hardware and network, right?

I mean, we’re, we’re now in this era that clearly shows the power of Microsoft. Google, Facebook, Oracle, right? These companies that have cloud infrastructure, global cloud infrastructure, they have a different kind of currency than they had in the, in the pre AI world of, of tech. And I think that that sort of lends into this next story, talking about how VCs have been aping into these AI deals, not really recognizing all the platform risks that are there when you’re [01:10:00] trying to base everything on, on AI.

But. But before we get into that story, I mean, just,

Vic: yeah, I mean, I think just going back to the structure, I mean, I’ve done this for a long time, been a VC 23 years, I’ve never seen this kind of structure before, where you have a nonprofit with a subsidiary that is worth any significant amount of money.

Right. It was worth 90 billion or a hundred billion or something. Um, and is the world leader in a technology that is changing all these markets. Yeah. It probably wasn’t the place to, like, try new things and try to reinvent a corporate governance structure. But that’s, that’s what they did. And it is, I think, uh, it’s symbolic that Elon Musk was an early backer and then pulled away because he didn’t like the direction it was going.

The Anthropic founders [01:11:00] were founders in open AI, and then they left to start Anthropic. Um, it w the, the platform was started with a mission that I don’t think Sam Altman as CEO Really like fully believed in he was much more aligned with the for profit and, you know, I would probably be the same, but that that probably means he wasn’t the best person to be CEO and.

It’s going to be chaos now. I mean, I think Microsoft will likely gather say half of the, of the developers, but, uh, Benny off from Salesforce is doing an open call. Um, Musk is doing an open call. Like any, it’s like a free for all trying to grab talent, but the source code is not owned by those people.

They’re going to have to leave the source code there. And I don’t know exactly what that means for Microsoft. [01:12:00] It’s going to be. A little while for them to get going if, if chat GPT falls apart, or it’s just. So I agree that, that Microsoft has played its hand very well, but I don’t know that, um, I would count Google and the others out that, that have, don’t have all this chaos and have their team and can sort of go.

Marcus: No, I, I definitely see this as an opening, um, that’s going to slow some of the momentum that is a. Crack in the sort of veneer of how people view open AI in terms of, is it going to be the stable winner? Should I invest more time in Claude or Bard or Grok or right. You know, just kind of really,

Vic: you have to have alternatives now

Marcus: because who

Vic: knows what’s going to happen.

Marcus: Yeah. Yeah. And, uh, I just got an email from, from Google giving me like a whole, you know, slew of updates from Bard. Yeah. So look, I mean, it’s going to, it’s going to be a race. Yeah.

Vic: Yeah,

Marcus: it’s going to be a [01:13:00] race. And then I

Vic: agree with you that right now it looks like the compute infrastructure and cloud and the chips is a consolidator of power.

There’s only less than 10 organizations in the world that have enough compute to to work with this. Um, I’m hopeful that there will be open source models. That are, that exist that allow for innovation in other areas, because that’s better for the overall market. And selfishly it’s better for VC. Um, but I think that is what it looks like right now that the.

The early movers are the people that have huge amounts of cloud compute.

Marcus: Yeah. I mean, when you think about the power of, uh, the compute combined with the network combined with the distribution, this feels like a winner’s take all kind of, kind of moment, um, unfortunately. So let’s shift to this, uh, wall street journal pro story about, um, [01:14:00] VCs and how they’ve been aping into all the gen AI deals.

Um, I, I do find it interesting. I mean, you. With your new fund have certainly been focused on AI, but we’ve been very measured, I think, in our thinking around AI as it pertains to healthcare. Um, it still has to have applicability. You know, I think we know that, that the appetite to, uh, bring in AI is not evenly distributed and it’s not a hundred percent here in, in healthcare right now.

Um, you know, there’s a lot of stories popping up right now about how payers are likely using AI for, um, For, you know, to accelerate denials of claims. Um, but you know, in terms of like health systems, I mean, look, we just, we just talked to what I think is probably one of the most forward thinking, you know, chief digital transformation officers who is maybe going slowly

Vic: in the, in the Patient care.

Marcus: Yes, he was very clear. We’re not ready for LLMs in the clinical, you [01:15:00] know, environment yet. So anyway, having, having said all that, um, we know outside of healthcare VC is almost entirely enchanted with gen AI right now. And, and that has actually been pulling up valuations and deal count, um, where outside of that, almost no deals are getting done, but you know, you and I said pretty early that this feels like a winner.

Take all, I don’t think this is a new take that we have, that, that Microsoft and Google and Facebook, we’re going to be, you know, the powerhouses in this space and your ability. I mean, look, let’s, let’s put AI to the side for a second. We’re, we’re, we’re about to talk about a company that I think is well positioned to fight.

You know, Google, Google calendar came out with appointment scheduling last week. And it’s like, holy crap. That’s a, that’s a total shot at Calumley, which was a unicorn,

Vic: you

Marcus: know? But it’s like. It’s a unicorn that was built on Google calendar. And if Google calendar gives you [01:16:00] appointment setting for free, I think that’s a lot of value destruction.

If that was a publicly traded company, that stock would be down big time on that announcement, right? It’s privately. Held company. So, you know, we’re not able to see how that might impact evaluation, but there’s no question. Lesson is the

Vic: same. It’s still a significant threat. That’s right.

Marcus: And we’re talking about a unicorn, by the way, we need to be clear.

Cal only is a unicorn. So when we’re talking about these gen AI companies that are doing features, you know, there were a lot of companies that were out there creating custom GPTs and then opening, I comes out with the custom GPT thing and you’re, you’re dead.

Vic: Yeah, that’s it. Yeah. So this article. I think talks exactly to that, which is, uh, for ever, there’s been kind of basic rules of the road in, in venture, right?

Like you should get some governance and control. Now you may not be able to force things, but you should easily be at the table and be talking with the leadership. [01:17:00] And there are a lot of times when VCs didn’t get, didn’t get board seats. No one got board seats in, in open AI. Microsoft spent 13 billion and didn’t get a board seat.

Um, so that that’s the first thing, but the second thing is exactly what you’re saying, which is, um, it has been a significant risk to rely on another company’s technology that like you build on top of someone else’s foundation. And I mean, calendarly and, and Google calendar is a good example of that.

There are hundreds of companies built on top of open AI, chat, GPT, and And that brings risk. You’re relying on another party’s Goodwill, and they can just decide at a moment’s notice. Well, we, we want that market. Now, cowardly, you showed us this is a billion dollar market opportunity. We should take that.

And it’s actually better for the users. [01:18:00] Totally. I want to jump out of the interface I’m in right now. And pay an extra 15 a month, right? And so that’s a thing in venture that you don’t, you want to have freedom to operate. You want to be able to build your own tech stack and not be subject to the whims of some other corporate partner because their own strategy changes.

And in AI, because of the capital needs, not many VCs, I don’t, I can’t think of any that have, except for the open AI ones that have invested in a pure LLM, they’re all kind of building on top of these things and that’s subject to significant risk. And then the last one is, uh, typically I don’t want to fund.

Large scale CapEx or like, that doesn’t pay that well and the, the hardware. Like buying CPUs. Oh yeah. Buying chips. Buying hardware. Gotta support it. Depreciates it depreciates like in a week. It’s terrible. Yeah, it’s terrible. It’s [01:19:00] not a good use of equity capital. No. No. So VCs forever have been. Using lease structures or yes, right.

That’s, that’s exactly. So, um, there were a lot of EC money that went into the buy in chips, the video chips. And I think that maybe isn’t the wisest thing to do. So anyway, that it was just interesting to see the, the, uh, You know, the fear of missing out has caused all of these VCs to pile into this stuff and kind of giving up their own kind of best practices.

Um, I’ve been looking into it. I mean, I really am interested in the power of it. We have not invested in an AI company yet. I’ve looked at a ton. I did invest in a company that’s sort of trying to, trying to work on the interface around communications in the post acute space, uh, which I think could be a really interesting data set.

But right now they’re not doing any AI. They’re just sort of managing that [01:20:00] communication flow. One of the things that I think is an opportunity is getting unique data sets. In healthcare where you have a patient population or you have a physician nurse patient payer population and you have your own data that that no one else really has no one else has organized in a good way so you can sort of train the model for that particular narrow use case better than anyone else, right?

I think that’s pretty interesting and is defensible and sustainable. I’m trying to line up some assets like in those positions that maybe could evolve in that way. But, but we have to see how it kind of pans out.

We’ll see. I mean, it’s going to be exciting next, next few weeks.

Marcus: So listen, uh, you’ve hung out with us for, uh, well over an hour now. So thanks so much for listening. Um, you know, please share this with, with your friends, uh, go on your Apple podcast app and, and [01:21:00] give us five stars. We are. Lighting up some really fun stuff for y’all for 2024.

I’m really excited about where that’s all going and hope that you all have a happy and very safe, uh, Thanksgiving.

Vic: Yeah. One of my favorite holidays.

Marcus: Yeah.

Vic: Not much responsibilities. I’ll eat, eat a bunch and lay around, watch football. Yeah.

Marcus: Uh, happy Thanksgiving, Vic. Yes, you too. All right. Bye y’all.