Can an AI scientist solve humanity's problems?

0:02Hello and welcome to the Times Tech Podcast with me Danny Fortson and only me today as Katie is away for a couple weeks. Sadly, it means that unlike last week's episode where we talked about cricket and my weekend in the mountains, there will be none of that. But even if there's no cricket, there is never a dull day out here in Silicon Valley. On today's episode, I've been speaking to Sam Rodriguez, a physicist, bioengineer, and co-founder and CEO of Future House. This is a nonprofit based in San Francisco that aims to build an AI scientist. But for what exactly? That is the$64 ,000 question and one that I put to him.

0:50That's coming up in a little bit. But before we get there, I want to pick up on a few of the big stories that are kind of happening out here. And there's a lot going on. So right at the top, firstly, last week, Katie and I spoke about the poaching of the best AI talent in Silicon Valley, this huge talent war that has broken out here. Sam Altman talked about this in a leaked memo where he said, quote, quote, what Meta is doing will, in my opinion, lead to very deep cultural problems. And of course, Metta, he's talking about Mark Zuckerberg throwing around nine figure salaries, you know, or pay packages for the top engineers, treating these guys like LeBron James or Ronaldo or Messi or, you know, pick your top sporting hero.

1:41That's what he's paying these guys to build what he says will be the future. From Meta's perspective, they're like, look, we're not paying every AI engineer$200 million. What we are trying to do is basically assemble an all-star team. These are pretty rare. We're paying the all-stars like all-stars. That in the context of the opportunity, this makes sense. And he doubled down this week. Zuckerberg did. In a Threads post, he said he's going to spend hundreds of billions of dollars building huge data centers all over the US. Now, the question is, what is his game plan here? And I think it's scorched earth.

2:21It's basically to scare the bejesus out of the rest of the industry while also trying to attract the best people, being like, you know, I'm going to pay you top dollar, and I'm going to spend more than anybody else. So, you know, that is to individuals paying up to$200 million for one guy, apparently, this Apple engineer named Ruming Peng. And then he's talking about data centers the size of Manhattan. You know, it's insane. And just two thoughts on what is happening in this kind of crazy AI war that has broken out. One is when you're thinking about Zuckerberg. Think about the metaverse. Remember the metaverse?

3:06zuckerberg spent 50 billion dollars on the metaverse trying to get us all to you know don glasses and goggles and if you recall it wasn't that long ago he was doing these demos where he'd show up in some virtual meeting room with him and a bunch of other avatars with no legs and be like oh look this is the future etc it was all very cringy it didn't work and it cost 50 billion In other words, money is not the salve. That is just because you have more money and willingness to spend it does not mean you will win. To wit, a senior engineer at Meta wrote a scathing letter on his way out of the company last week, and it was leaked to the information.

3:51And in this letter, he wrote about this culture of fear inside Meta, all the bad vibes. He said we're, quote, spreading like a cancer. Nobody knows what Meta's mission is. if you have zuckerberg he says i've created this new super intelligence lab we're going to build super intelligence so i think this is more like a signpost for if you want to build really ambitious stuff talent come to me in response to the leaking of that meta said we're excited about our recent changes new hires in leadership and research and continued work to create an ideal environment for revolutionary research but the whole idea does not seem to be landing at least certainly with this one guy and it just went viral as people were talking about you know what meta is trying to do what zuckerberg is trying to do so money money isn't everything it's a lot but it's not everything it'll be interesting to see if zuckerberg can kind of pull it off and get meta right back at the top of the race in terms of ai that is one story which leads to this other story which is so crazy and very, very Silicon Valley and really kind of captures what's happening out here.

5:02And that's about this company called WindSurf. You may not have heard of them. You may have a quick recap. WindSurf is this very buzzy AI coding startup that a few months ago in April, Sam Altman at OpenEye said they're going to buy that for$3 billion. You're like, oh, great, cool. However, that deal swiftly got stuck because an issue between them and Microsoft, which is of course OpenAI's biggest investor and number one frenemy. And the whole thing kind of got stuck in the mud because there is an issue between OpenAI and Microsoft over who would or would not have access to the service once the deal was done.

5:40The deal stalled. In the middle of that, Anthropic, our friends at Anthropic, Dario, Amadei, and crew, which Windsurf uses to build on top of, they use clod anthropics to build on top of they were like you know what we don't want to help create a product that is going to end up getting sold by our bitter rival open ai so they just cut off access to their llms which meant all of a sudden windsurf had to scramble had to retool it made everything much more complicated and so they're kind of panicking trying to uh retool on the fly. This deal is stuck. And then all of a sudden, last week, 40 top windsurf engineers get this interesting email say show up at this hotel in Mountain View.

6:30So they do. They all just kind of troop out. They walk in and who's their their CEO. And next to him is Sergey Brin, Google's founder and the 10th richest man on the planet. And he's saying, we'll hire you, we'll double your pay, you know, your shares, your salary, etc. Come with us. They basically all agree. The total price is$2.4 billion for these 40 folks and some licensing fees. Meanwhile, back at Windsurf, there's 200 people who are like, why is a fifth of our company just not here? That's strange. And then they hear about this, this deal where the CEO and the top people have all left taking Google's money.

7:12So then the people at Windsurf are basically like, oh, no, we're manning a ghost ship because Anthropica's cut us off. Our CEO and our top team have left. And now we don't know what to do. The new CEO who's been in the job about two minutes, he then scrambles, has all week negotiations and ends up getting bought by another coding startup called Cognition in a Shotgun marriage. Crazy. So just a few kind of lessons I feel like worth drawing out there is that Silicon Valley is really losing its mind right now. This is proper bubble stuff. It's also getting very, very messy. These top AI developers trying to grab the best talent.

7:58And most interestingly, for the kind of the bubble aspect of things, it shows the fragility of a lot of these startups, which are basically building what they call wrappers around other people's technology. This is a lesson that media companies have had to learn time and again around building businesses on top of other people's tech stacks. In the case of social media, it was news feeds and algorithms that if they get changed, if they are altered, if Mark Zuckerberg wakes up and decides he wants to do something different, all of a sudden you have no business. and there are so many AI startups in that boat.

8:39And Windsurf has managed to find a landing spot, but it was just one small example of just be careful. You have to really do something special. You can't just build on top of other people's stuff because they can change their mind and all of a sudden you're up a famous creek. And then lastly, last story before we get to today's interview, Nvidia is starting to ship chips to China again. Big reversal from the Trump administration after months of saying, no, you cannot send these powerful chips to China. They now can. Jensen Wang, the CEO, he had been doing a lot of lobbying and it's a really big deal for the Trump administration to be like, look, this is a war for supremacy.

9:24We're talking about warfare. We're talking about new weapons. We don't want to like give them our best stuff so that they can build their own weapons, their own AI systems that are as good as ours. But now NVIDIA is starting to ship chips to China again, selling H20 chips, which are not their most, most powerful, but they're very powerful. And the reasoning, I think, is that they're saying, look, we want the world's AI systems to be built with American technology. We don't want to hand the market to our Chinese rivals. like Huawei. So we should be in the game. We should be in the arena. Yes, there are dangers here, but there is kind of a risk worth taking.

10:08We'll see. We'll see. But that was the other big deal for NVIDIA, which of course, recently crossed the$4 trillion market value threshold. So it's wild out here, y 'all. There's a lot of things happening. It feels like every time you turn around you hear like another just crazy story that I think I mentioned it before it does feel to me having started my career because I'm old in 2000 effectively it feels a lot like that just the kind of every day you wake up and hear something crazy a new story a new valuation a new takeover whatever it may be and it's it'll be interesting to see how it all turns out so that's the news that means we should get to our interview which I think you're really going to enjoy so as the global race for AI heats up, I thought one of the interesting aspects that I heard about was this company, Future House, and they have created an AI scientist, an autonomous AI scientist.

11:08And, you know, the question is, what can it do? The possibilities are endless. You know, we're talking about if you talk to folks like Dario at Anthropic or Sam Altman, they're like, well, these AI scientists, once we get them sorted, they're going to cure cancer, They're going to double lifespan. We're going to be 3D printing organs, et cetera. So I spoke to someone who has created his own nonprofit in San Francisco to do just this, to create autonomous AI scientists. It's called Future House. It's backed by Eric Schmidt, the former CEO of Google, and by Open Philanthropy, which is a major funder within the EA movement of effective altruism, which is a philosophy just as a reminder to, quote, focus on the questions of how can we best use our resources to help others?

11:57In other words, get as rich as you can and then basically take very big swings with a goal being how many lives can I save? And of course, the biggest backer of this whole idea, most famously, was Sam Bankman-Fried, who is now in prison, the former CEO of FTX. So EA kind of has a bad name these days, but future houses backed by Open Philanthropy, as I said, as well as Eric Schmidt. So I started with Sam with an easy question, I hoped, which is what exactly is this FutureHouse thing he has created? So FutureHouse is a new nonprofit research lab in San Francisco focused on building an AI scientist.

12:45The inspiration behind FutureHouse is observation that I made in biology. I was a biologist originally, a theoretical physicist originally, and then biologist after that. And I had the impression when I was studying biology that even if we had all the information that we needed to understand how the brain works or to understand how the cell works or whatever, we wouldn't necessarily know it because no one would have enough time to go and read all the literature. And even if they could read all the literature, they wouldn't necessarily be able to go and assemble it into some comprehensive kind of model in their head.

13:16And so this is what basically convinced me that the most important thing that we could be doing in science today is figuring out how to go and build an AI scientist to build a system that is smarter than we are, that is better than we are at understanding complex science. So that's what we're focused on at Futurize. We've been around for about a year and a half. and our goal is essentially to build systems that can do literature search so it can go and retrieve information from the scientific literature go and retrieve data sets and so on go and analyze data and go and generate hypotheses at the same level as human scientists or maybe even better than human scientists why non-profit we are here in startup land i would guess there are some many a whole bunch of startups trying to do the same thing in some form or fashion so there are fewer than you might think, actually.

14:05And part of the reason we're nonprofit is because we're interested in automating basic discovery research. And it is very difficult usually for, for-profits don't usually work on kind of basic research, right? They usually work on more applied research, drug development, and so on. As we get further in our journey, I mean, like many nonprofits, I have a tendency to spin out for-profits. I do anticipate that as we get further in our journey, we will likewise see opportunities to spin out for-profits and we will take those opportunities when we have them but you know fundamentally the core of like automating basic discovery research is like a is a non-profit activity and where is the funding from who's who's backing you guys we're backed primarily by eric schmidt at the moment but we also get funding from a bunch of other people we have funding from open philanthropy for example yeah which is a another kind of san francisco-based philanthropic organization they're ea are they not they're effective altruists indeed yep yeah um and they're really interested in the question they funded us to investigate the question of how you measure the performance of like you know ais on scientific tasks so there are several other organizations that have funded us what were you doing before this so i i did my undergrad in theoretical physics i then did my phd at mit on in neuroscience and neuroengineering basically building new tools to get more information out about how the brain works And then I spent three years at the Francis Crick Institute in London as a group leader there, basically running a bioengineering laboratory.

15:34And then I moved from there to start FutureHouse in late 2023. Bioengineering. Yeah. Is that what it sounds like? Bioengineering encompasses both how do we engineer biology, so get biology to do new things that it wouldn't normally do. and also how do we build tools to interact with how do we build like you know new kinds of microscopes like molecular microscopes that can read out the inner workings of cells for example that would be an example of bioengineering also so how hard is this and i know that sounds like a facile question for example open ai has you know there's a story that came out a couple months ago now then like maybe we'll charge 20 grand a month for a quote-unquote phd level agent you know Here's your AI physicist, go off and do physicist things.

16:23But again, as you say, there's not a ton of companies doing this. And I imagine it's a bit more complicated than it might sound. But maybe it's not. I don't know. No, I do think it is complicated. There are several reasons why it is complicated. The first reason why it's complicated is that you have to figure out how to teach these models. A lot of the things that humans know about how to do science, they don't necessarily write down. And it does turn out that there are many of those things. There are many things that are like, you know, about how to do data analysis, for example, that don't really exist in literature.

16:58And so as a result, like the language models may not have learned them in the process of training that nonetheless are like well understood that it's an important thing to do. Right? Right. So you're talking about like kind of like the unspoken art of science. Yeah, there's some amount of unspoken art. And I think that people tend to overestimate how much unspoken art there is. There's like there's certainly some. It's not like I think most things that humans do are written down somewhere and language models will probably have picked them up. But there is certainly there's certainly some. And so you have to figure out how to identify the pieces of the unspoken art and then train the language models in them.

17:32And so that's one thing. Another big thing is actually infrastructure, right? So I think that when OpenAI talks about having an agent that can go and do physics for you or whatever, they imagine an agent that is able to interact with its computer much the same way that you can. That is one approach. And OpenAI kind of previewed this approach with Operator, which they released back in January. One of the problems with that approach is that it's relatively slow. And it does turn out that if you have, you can build systems that are much more performant if they have purpose-built infrastructure. So, for example, we have played around a lot with building purpose-built infrastructure to allow agents to search the scientific literature.

18:13That allows our agents to search like 50 to 100 times more papers than OpenAI's deep research can, for example. And so there will be a lot of work, I think, around building a dedicated infrastructure for agents in specific areas. Right. There are a variety of other reasons. Right. And I mean, I'm sure it wouldn't work out quite like this, but there could be a data center X. It was like, that's the science data center. That's a scientist data center. Yeah. Maybe not on a data center specific, but like, you know, I think humans just, we may take for granted the extent to which the internet is like built for humans to interact with it and is not yet built for agents to interact with it.

18:53I want to emphasize yet because we will certainly get there. It's just, it will, it will take a little bit of time.

19:08Are there any examples of that? Talk about that unspoken art of science. Cause I just can't conceive of what something like that would be. Is there a simple example that comes to mind? There are questions, for example, about a better example. There are many more of these examples like in the lab. So I like to tell people there's a protocol that I was doing once where I had to wash some cells. So the cells were like in a plate and I had to wash them. So you put some liquid on top of them and then you want to get the liquid off quickly. And so what you do is you like turn the plate upside down and you flick it.

19:40And like the process of flicking it gets the liquid off, but the cells stay on. And it's very difficult to describe. Like, I don't know how many papers there are that actually like describe the art of flicking a plate. You know, so, I mean, but the thing is that there are a lot of papers out there and the odds are that it's probably written down somewhere. Yeah. You know, if it's not in a published paper, it's maybe on Stack Exchange or whatever. It's probably written down somewhere. But that's an example of the kind of thing. And these examples become more important as you go to more niche fields.

20:11Yeah. So there will be a field somewhere where it's like there are only three people who do that thing. And they've all talked to each other and they all know the unspoken art. but they've never written it down because you know there are only three of them and it's not going to be written down anymore so can you help me try to think about this because i feel like in a way especially with ai our imagination is a little bit limited as to what this could all do but yeah for example if you guys are successful in building a truly kind of autonomous super capable ai scientist or scientists in specific fields or whatever do you think it's realistic that, you know, we just send them off to do their work and be like, they'll come up with some magic material that sucks carbon out of the air or cracks nuclear fusion or whatever comes up with something completely novel that we just can't even conceive of now.

21:02Like, you know, when I talked about that Dario essay, he's talking about basically longevity, solving all diseases, all of this. We just don't have enough brain power, these really complex problems. And if you create that synthetic brain power, we're going to solve all of this stuff. I mean, I do think that in the fullness of time, it's reasonable to expect that we will solve most or all of those problems. Now, look, it is unclear to what extent it is possible to solve some of these problems, right? Like, maybe the laws of physics don't prohibit it, but there may be other kinds of things that make it effectively like impossible or whatever, that we just like don't quite understand yet.

21:37But I think barring any hard constraints like that, I do think it's possible that we will discover solutions to many of the problems that you mentioned, many of the diseases and so on. I'll put some caveats on that. The first caveat is that the diseases that we are aware of are the diseases that humans encounter up to the age of like 100, right? Right. And if humans lived to be 150 or 200 or something, we would encounter new diseases. Other things in your body would break that just like are more robust and they don't break until you get to be 150. And we've never seen it because no one has gotten it.

22:15Right. And so I do expect that as we cure more diseases, we may discover more diseases. Now, if you fundamentally figure out how to cure aging or something like that, then that may be a different situation. But I think aging is one of those things where it's like, it's not 100 % clear right now whether like curing aging writ large is actually possible. Yeah. Right. So the jury is out on that and I could dive in there. But then the other thing I think that's important to emphasize, we may have very, very good ideas very soon. Right. If we have a superhuman AI scientist, we will get a ton of really high quality new hypotheses very fast.

22:50That does not necessarily mean that we will like get cures immediately. Right. Yeah. Even if we had a cure today for like Alzheimer's disease, it may take us five or 10 years to know that it was working because five or 10 years is the timescale on which that disease progresses. Or longevity is a better example. Right. Like if we had a cure for longevity, it would take 10 years for you to like notice that you weren't aging or something like that. Right. Right. And so we just need to be a little bit. Biology happens at a given timescale and we need to be a bit realistic. But that sets a bound on how quickly we can like make new discoveries.

23:29But I think it will happen. I mean, for a lot of people, it will happen in our lifetimes. Right. That we will cure many, many diseases, maybe solve aging and so on in our lifetimes. If you were a betting man, how long do you think you're going to live? That's a great question. I would say my uncertainty in that is very, very high. and so you know assuming like death from natural causes or whatever right like uh yeah not like hit by a bus tomorrow yeah exactly i don't know i mean i would say like you know a hundred or something like that's not that's not too crazy right i would just say my uncertainty there is extremely high if you take the expected value and there's a possibility that it's like that there's like some non-zero weight on like you know a thousand then my expected value will be like 500 or something, right?

24:12But like, I basically, I don't know. I don't have a really informed opinion about that, actually. So what is the goal of Future House in terms of who's using this? And like, how close are you? Or maybe it's already out kind of quote, in the wild, being used with effect? What's the kind of the game plan? Yeah. So we released a platform last week that allows people for the first time to go and use our agents. And so that's been very popular. We have tens of thousands of people using it. And so that is, it's already out in the wild in that sense. We've been getting a lot of feedback, which is very nice.

24:45I think what we're really interested in, there's some extent to which people will use it as an assistant, which is fine. What we're really interested in is people using it to make new discoveries. And that will just take longer. We'll really see the effects of people using it to accelerate their own process of discovery on a six to 12 month timescale, I think. and even before then we will be using it i mean like we really built this for ourselves we are scientists right and we will be using it to make discoveries in house and you know we'll have more on that soon i can't say anything about it right now but right so you guys are using it to run your develop your own hypotheses which then you would test so there is a world that you could kind of be like hey we've got this new drug candidate or therapeutic or whatever it may be we're going to spin that out into a company yep etc yep exactly so you're kind of like scientist as a service something like that i mean scientist as a service is actually very interesting i you know that's a great that's a great tagline i don't know that that's necessarily the model as opposed to using it to make new discoveries or uh i don't know but got you is there a thing you're most as a scientist and coming at this from you know scientist perspective is there a thing that you're most excited about that this might allow you or the world to do?

26:02And also the corollary to that is, is there a thing that freaks you out the most about the potential for how this might be used? Yeah. I mean, look, so the thing that excites me the most is the opportunity to go and cure all human diseases. I think it's pretty cool. That is pretty cool. Yeah. Lots of other interesting things. I mean, fundamentally what I would really like is I would like to be able to go and like explore other planets or something like that, right? Which sounds completely insane. But actually, you know, if we figure out how to do cryonics, we figure out how to like freeze humans in a way that allows us to warm them back up, then maybe it's possible.

26:38So whatever, we can talk about all that. The thing that concerns me the most, I think, usually when people ask me this, they expect me to say like things about biosecurity, about viruses, and, you know, can people make killer viruses? So these are like very real and serious concerns. There are a lot of people thinking about it. I tend to be a little bit less concerned about this because probably it will also help us to make really strong vaccines. Fundamentally, viruses have been trying to kill humans for multicellular organisms for 500 million years. Actually, probably way longer than that. They've been around for bacteria also.

27:11I don't know. I don't want to dismiss it. It's very important. I think the thing that's going to be more interesting to talk about that I worry about is the effect on scientists. I don't think that scientists will ever be fully out of the loop because i think like fundamentally if you imagine like the research that we do will be limited by like resources right it's like how much money do you have like how do you want to allocate that money you have an ai scientist that's going to tell you hey probably the most impactful things we could do are like x y or z you as the human probably have to make the decision right like i do think that grants will for the foreseeable future like scientific funding will be given to humans to go and do stuff using AI, we're not going to fund AIs directly to go and do research.

27:51But I do worry about like, I worry a lot about the effects that we'll have on the roles that scientists play in the future. You know, what does a scientist in 10 years do when like a lot of the stuff that we think about today, like an essential part of being a scientist is automated. And, you know, I worry about that a little bit. I just worry, you know, for humanity, making sure there's still like a lot of really interesting things to do job um and i it's a job it's jobs basically yeah jobs basically um i i think that there will be but even in the absence of like you know science is fun there's like jobs from an economic perspective and i think lots of people have thought very hard about that but they also just worry you know like i want to make sure there are like so fun things for for people to do and so maybe i worry about that a little bit i don't worry about it that much because i think we have gone through many, maybe if you were around in the 1800s, and I told you, hey, you know, pretty soon, you're not gonna need to like make shoes by hand anymore.

28:48The cobblers would have been like, but making shoes is fun. What are we going to do in the future? Maybe there's not gonna be anything to do. And in reality, people adapt. And so I do think that there will be like new things to do. But I just I don't know what they are yet. Yeah, I guess the question that a lot of people are wrestling with is, if the machines you're working with are smarter and more capable than you what is your role as the human are you a manager of those machines like an orchestra conductor or are you a servant you know just playing the violin yeah i suspect that there will be some like many problems actually where even a super intelligent ai is not necessarily better at solving them than humans simply because it's like not limited by intelligence right not everything in the world is limited by intelligence i think i can say this with high confidence being a biologist because like a lot of biology obviously i think that there's a lot of biology that is limited by intelligence that's why what we're doing but a lot of biology is also not limited by intelligence it's just limited by like data or whatever and so i think there'll be a lot of problems where it's just like you know yeah you could have an ai work on it but the ai is going to do exactly the same exactly as well as the human will so the human will do it or something right right right right deciding where to go to dinner is like a great example of this right like there's no right answer ultimately an ai will be able to make you recommendations that's probably not like going to be able to tell you what you want or like yeah to some degree like you deciding what you want is the whole point so yeah lastly just in terms of the funding how does one how does a scientist at the francis quick institute get in touch with eric schmidt and convince him to give you some money to start this crazy non-profit idea in my case i got connected through some other work that I did in my PhD and was lucky enough that I got to form a relationship with him and get to know him pretty well.

30:35I do think it takes some amount of luck and some amount of persistence, but it also just takes putting your ideas out there. It turns out that ultimately the thing that got me connected to Eric in the first place was an essay that I wrote, which had some ideas in it about these things called focused research organizations. And it turned out that Eric has been very interested in those ideas. And so I would just say, put your ideas out there into the world. And if you have good ideas, people listen. All right. So what do y 'all think? One logistic point when he said, Oh, we just released this last week, I did this interview last month.

31:12So it wasn't just last week. But listening again to that conversation. I don't know about you all, but things I was thinking about, he talked about living forever, or not forever but living to say 200 years old and it will have new sets of problems which is kind of interesting the thing that always kind of comes up for me is your skin what do you do with your skin if you're 200 years old i mean maybe we'll have some mega moisturizer that is kind of revitalizes your skin but it's like this this suit that you wear every day it's like you know if you're if you are a house the exterior of your house needs to be repainted and replaced and redone it but if you have a 200 year old you know skin suit what's that look like i don't know probably we're never gonna have to find out that's the other thing i whenever i listen back to these things kind of talking about these fantastical futures is part of me is like is this all just going to look ridiculous in five years from now like we're talking about curing all disease tripling lifespan all of this stuff in 2030 if i listened to this i'd be like oh my god we were just so far up our own behinds we were just right in the middle of this crazy ai boom and everybody was so excited about the possibilities and the reality is just a lot more mundane and a lot less exciting or interesting or maybe it's not imagine it'll be some some probably somewhere in the middle but it's just uh it's interesting that these are the type of conversations people are having all the time based on the kind of the excitement about this technology and the potential power of this technology at least by the people who are building it so yeah we'll see but i hope you guys enjoyed that my head is certainly spinning from all the ai news and this weekend i'll be writing about shocker ai stuff i'm not exactly sure what yet so you actually have to log on to thetimes.com or even grab a physical pile of dead trees, a physical newspaper this Sunday to find out.

33:20But I have some fun stuff coming, including one interesting, very interesting UK tech story that I'm working on that I can't really talk about right now. So you'll have to wait till Sunday for that. Anyhow, that is it for me this week. KDR is off again next week. So we're going to have another old school version of the pod, an old school Danny in the Valley version. so please come back so I'm not flying totally solo next week thank you as always for listening and we'll talk to you very soon bye bye