all right so while we're doing this session because this is all about what really works in artificial intelligence and we've seen a lot of companies that are applying machine learning sometimes they're about to apply machine learning obviously there are a lot of talks around what could a I do in the future but as investors we care about what's really working as an incubator you care about what's really working so that's why we wanted to do this little session we prepared some questions but actually this is your chance if you're thinking about what you want to do in machine learning in your company in your startup maybe even as a job ask us anything and we hopefully can answer it just as I would like to ask you to introduce yourself but just very roughly because there are certain things that maybe you don't want to say yourself so been on the on the left side is from atomic oh I think you're managing 850 million I always get one point all right there so much
more money than we do huge fund one of the founders is the founder of Skype do a lot of very very big technology bets can invest a lot of money in those companies which is pretty impressive we like to partner with them a lot and they invest all across Europe but also outside and Ben has a background in computer science and has invested in quite some nice artificial intelligence companies after that please tell us a little bit more about your motivation and what you're doing and Rasmus also has all by the way and you have a I think computer science in from Cambridge yes your degree and then on this side there is a rowing I think opponent from Oxford I don't know if you guys did with the rowing challenge but Osmos from Oxford with mechanical engineering and then I think from Zurich with computer science PhD who is the head of neuron takes a company that is starting a eye on machine learning enabled business models so various ones by attracting very very
good talent and helping them to start companies so very glad to have you both here thank you very much Thanks then tell us why what you like about machine learning artificial while you're here and what's your background sure I think you gave a great overview of anatomica I think we're we're a venture firm based in London and we were a generalist fund so we look at a lot of kinds of businesses but I think we're one of the few funds in Europe alongside Holtzberg who are kind of committing a significant amount of capital to invest in in companies who are really building with new and emerging technologies and that's kind of what's kept me excited I got into venture from a kind of computer science and engineering background with a view to really wanting to help support companies who were really kind of innovating at the cutting edge and soda Tomiko we kind of look at two groups of companies one are kind of I would say like applied AI and analytics companies so companies
that are taking advances in technology sort of come about over the last two or three years and I'm really applying them to specific commercial problems and that's why I focus most of my time we also look at some more kind of frontier technology investments that are often using AI machine learning as part of things like computer vision or or in things like healthcare use cases like drug discovery and other areas like that so that team also kind of invests some of the kind of most capital intensive bets that we've made but on things like we invested in the business developing a new semiconductor for AI workloads called growth core and those kind of businesses kind of really they're the building blocks of a I are kind of it's great to see those coming from Europe but those need kind of real investment behind them as well so another kind of business and but but still very much an AI business so yeah that's kind of what we're about and kind of why I why I'm
kind of investing in this area great so Russ most tell us why incubation makes sense in the face of it yeah sure so I was actually an academic research before and there were like a ton of really nice spin-offs at ETH but many of them had like great technology and great teams and we're working on really big problems but like commercializing them properly like it was quite difficult for many of them just because the team setup wasn't right because they didn't have access to the right industry partners maybe didn't have the right investors on board and that was actually quite sad to see I mean many of them got acquired by like Google and Apple and large companies but like quite early on words and so that could have been much more potential if they would have been in the right environment and that's basically what the rantings is about so it's a it's a venture studio focused on incubating AI companies for one to three years we're very flexible on how long we incubate
the companies we can fund them with up to a few million in euros so there's this you know like a significant amount capital behind them really bring them to a strong series a and basically during this process I mean Veronica's not only provides capital but it's obviously also very great to try to attract talent in different areas and also builds kind of the network to industry partners and I think one reason right now we have a you know yonas is building the healthcare annex healthcare we have a company in the autonomous driving space working basically in the validation testing space of autonomous cars and handling basically the raw amounts of data coming from the cars we have a few smaller companies which are you know just a few people and I think one special thing about neurotic is that we run MX labs which is which is also here today which runs actually also projects with the industry partners so one way we actually get involved into new ventures potentially also with corporate partners so
for example healthcare with like you know like a dozen Hospital partners across Europe or an automotive is some large car companies is that we basically run projects with labs and collaborate with them and build a partnership to then later incubate a company in that field potentially with a corporate together and I think that's that's really important because like obviously like we can for a lot of the consumer stuff you can you can basically just develop that at home and like put in the app store and see what sticks and then iterate but like for some of this stuff like like healthcare or automotive like any other industry anything where where you go and these like you know like quite complex industries you need to get out you need to work with those companies and so in order to get a growth rate product market fit because yeah it's it's quite easy with a great team to raise some money to for a great team in the beginning but at some point you will have to
show traction and great for like market fit and I think that's that's very important so that's what we really focus on with more on takes and maybe one last things so actually all our companies are incubated in different industries which sometimes sounds a bit counterintuitive but actually has the huge advantage that there will never compete with each other because they all said two completely different clients and so internally actually we can share basically all the knowledge all the you know all both even code which is just developed jointly to basically make everybody more productive and so miranda has very little overhead it's just ready to service just for people it's micophone Adrienne to me and then two more in the in the core team and everything else happens and the companies and them basically helping each other so yeah that's that's what we are we're in Billund to another years old and obviously are right now also looking at new industries we want to get into and so that's why it's very it's a very
like ongoing thing to think about business models where I can be really used how is the technology ready is the market ready is it large enough for our existing companies but also new companies we're gonna work on great everyone one other thought I think it's it's one of the most exciting things when you look at the last ten years this is not my own point but I've stolen it from somebody much smarter than me in the US but if you look at the last ten years and the kind of companies that were built on problems that were sold they were a lot of the easier problems they were how do I get food to myself quicker or how do I get from A to B on in a taxi faster and I think a lot of the problems that you're tackling now and the problems that great companies are tackling now are the much harder ones they're in healthcare they're in manufacturing they're in these other much more complex areas but the opportunity to solve
those problems although the the improvements that can be made if we do solve them are kind of much more significant and so I think there are also problems that are kind of naturally problems really well-suited to being solved by by machine learning and AI and its various applications and so I think that it's and also to be solved in in Europe where a lot of we have a lot of historic strength in in the industries and I think that those things come together to be a kind of very exciting time at the next 10 years in these areas and maybe adding to what you just said and then to give you a little background about myself and what we do we do our investment we come from a consumer internet background so the londo were first investor flicks bus we were fastened first investor so for us actually there's research projects we don't really like them we want to see a product that you can give to a customer and get instant feedback like I
to XO ultimate AI where we invested in and with machine learning sometimes and we saw that with with a company called site gold where we deployed a lot of money to accounting automation it sometimes takes quite some time until you get the automation done so it's actually a learning process for us as an investor as well that it's great what happens in machine learning but you should actually take a bit more risk around let's say you bet on the team but that the application might fail at the beginning as long as there will be a product at some time and maybe looking to start off and then it's it's your stage but looking into the future and it's your feeling that we are still in this kind of experimental phase where people are applying newer networks maybe reinforcement learning maybe even genetic algorithms whatsoever and just testing them around hypothesis what could be a product or are we a step ahead of that I think we have like different kind of streams so I think
now they're definitely the first couple of companies who are actually really using this stuff and scaling that and actually showing that it's it's a real company as opposed to like if you years ago when like most of these companies we're like we're researching more like aqua hires or tech tech hires at the end but at the same time you also now have super research II like frontier companies who are working on some of the more advanced machine learning which is still not really working fully or so like application may be a bit unclear but which probably will go to this production stage in a few years so I think it's like from the time when it works in research to the time when it's really in a product it takes some time because you know needs to get robust people need to understand it and I think this time it'll shorten but there's always gonna be like parallel things happening I think yeah I think there are there are some kinds of i/o a machine
learning areas that are well developed now so if you need if you want to build something that's based on speech recognition there are mature libraries and approaches that you can use to do that and build great applied products using those technologies I think there are other areas some of the things around kind of massive scale simulation or something like this where the approaches are still kind of being worked on and kind of more research led and so I think companies that are trying to apply those are kind of naturally solving two problems at once both how do we correct the underlying technical approach and how do commercialization and so I think when when you look at those companies from an investor perspective the opportunities are often big because they are solving a new problem that they can only solve because of the technology approach they're taken but then you've got to have there's kind of two different buckets of risks there's the technical risk and there's the commercialization risk and kind of two types of
team you have to have a kind of bigger technical and research team and again the kind of commercialization side of the business so they're they're often they're kind of more intensive businesses to build but again kind of a big reward if you if you can get all the pieces together I think there's also another thing like even for like I think medical imaging is actually a great example like you know like medical imaging in deep learning has been around for many years and like now training an algorithm on some public medical data set takes you like you can probably build that in a day and let them working now getting a medical grade product out there where you have a lot of considerations okay how you know how do you handle hundreds of terabytes of medical data what do you annotate if you would annotate everything would burn 50 million I mean I'm would happy happy to do that but it's like it's not it's not very smart and I think our bestest wouldn't like
it so being smart about for example what you annotate how you annotate also annotation so like the basis the how you prepare the data works very differently for like a medical product as opposed to for like research because research these standard data sets are more about comparing each other unlike some standard benchmark medicine and now like a medical product you need to annotate much more because you need to actually also predict much more about all the inventions so there is a lot of these things also and for example also dealing with uncertainty so if our algorithm predicts something and there's obviously some uncertainty involves how can you quantify that these are like all very art machine learning problems but very applied and kind of getting busy these 80% like some algorithms work takes a day building a medical grade product takes like twenty men years of work so it's like and I think many people underestimate that also because like academia has less of a focus on often on on this part and sometimes not
even the same challenges because they might deal with smallest data sets and so the challenges are just different cool so now it's up to you I see a first question over there we have a microphone they are those throwing microphones maybe we should introduce them yeah it's the guy with the head or you just showed hi so my name is Tommy I'm a science journalist based here in Berlin and my question is a rather broad one so what could you name or single out a couple of business branches that you think will be where the impact of AI will be most disruptive and can you think of any well-established laws and practices that will have to be revised or even rewritten in in the course of establishing bringing AI into it I think it's gonna be big and just cuz it's like generally space where there is already a lot of data and you could actually collect much more data and quantify it and like I think if it would be health care would be
more data-driven like it will be better for everyone in here and also anywhere in the world so I think it's just a very obvious application at the same time there are a lot of challenges around product what Yanis talked about a lot about regulatory and there I mean yeah you have some some regulatory regulatory topics iana's has also been talking about so I think for that specific thing I think here's a few things you would like to change in the law also mobility autonomous driving I mean you have similar applications right like who's liable when time something goes wrong how do you need these how do you certify these systems so often I mean your networks are kind of black boxes in there they're very large I mean they're you know a lot of parameters obviously will work on on making them more explainable but in the end you still have a few if you solve a hard problem you can have a big neural network trying to solve it because the problem is so
hard and so then it's very it's very optimistic to then think you can basically explain it with a simple rule so it's gonna be hard to explain and so then it's always about like liability so if this black box does something wrong wrong or as bias like who is liable for it is this allowed how you test this properly and so I think about these topics there's a lot more to discuss generally large industries to wrap this up Healthcare automotive I think anything in manufacturing robotics like there is there some big applications anything around like back office automation huge I think like one space we're also interested in cyber security so that that is becoming a bigger bigger and bigger problem obviously a lot of data there so yeah I think basically any industry you can UNIX and somewhere even I guess like kind of cross already go yeah I mean I completely agree and I think that that touches a lot of our ears I think one other perspective on the question which I
think it's interesting here is dealing with personal data and obviously there's a set of rules around what you can and can't do with with personal data but I think there's also a societal change that maybe needs to come or or kind of not necessarily rules but people's willingness to share or kind of I guess in the in the end people giving in to company timing and kind of using these technologies on their own data and I think that whether that's something kind of trivial like retargeting the advertising based on kind of knowing more and more about you through to something less trivial like a credit decision made on on your financial history something even less trivial like a kind of healthcare provision decision based on your genetics like those things are kind of different steps with different regulatory burdens around them but I think that I think over time I'm the optimist view says that we all in the end that all the technology is built for the best and in the end we won't
give in and go well it's gonna make the best decisions in the end but and so we're gonna trust it I think there's a lot of things that have to happen for for most I mean maybe people in this room we're probably kind of further along and in kind of the trust we have and that comes I think from understanding what I think that there's probably a long way to go before the general population are kind of happy to to buy into those decisions based on that personal data being made in that way heading to your points what we see is wherever you have certain set of rules that you for example can can match the algorithms to so the algorithm knows okay whatever I detect it should match the set of rules of formalization that works nicely accounting is a space where there's a lot of manual labor still back office automation you mentioned what we also see an eagle Leviton was active there so data extraction from documents but then matching that data
to actually the legal system there will be things happening but the big question is do you have a set of rules that you can I mean it's at the end of the day that statistics behind it there is huge potential and I think especially Michael from i2x highlighted that also Rachel from ultimate I people sometimes don't want to do this kind of work they're happy if you take it away from them and then they can actually invest more time in other things so we think that the adoption in those spaces whether where there's a lot of repetitive boring work would be very very high because it's not just about the CFO saying oh this is amazing I can automate everything so I can save cost but it's also about the people using it because as you said annotating the data are getting the feedback loops that's pretty important you can't just put it plug it in like a software and then it works instantly that's one year and to save rasmus from the potential subjective
bias may be a small fort for Jasper and Ben but if you look at the US or Asia especially China what they're doing in AI I mean it's it's a big world you feel humbled sometimes in Europe looking at the companies here I mean yes I have been a lot of success stories I mean US investors have been very successful identifying those local stories but maybe apart from well barest barriers to entry language regulatory framework all that what motivates you being in the European mouths ensure space or AI sector what what kind of what is our competitive edge or how do you motivate young people to join that and not run off to the US or Asia I mean what keeps you in Europe so I did discussion yesterday was a member of parliament German Member of Parliament and and we really see at least in our companies and that's not thousands of companies but we have invested in five AI driven companies that first of all we have pretty good talent here that is
willing to work in a start-up at okayish salary levels and I'm not saying they're underpaid I think they get good salaries but if you look at the Silicon Valley we're just a project leader at Google gets maybe a million or two if they are on a certain topic and have a neuroscience degree from Cornell or I don't know that's an issue right so you basically need more money and then you don't know what will happen so more risk and the second one is people like to come to to Germany and work here so we get a lot of talent from Eastern Europe also Western Europe they migrate here and come to the startups it's also the other way around so they're moving around obviously but that's a positive thing about Europe and there is pretty good education around it and then German universities have very good talent and I think we have certain data versus the Chinese I mean we learned it from conics and and only from from my cropsy that the Chinese don't
have around production maybe even the railway system our accounting system is quite special health care a certain data I this is this is a rumor but I heard that a certain a huge US company trained their name by I think thing a scientist algorithm on training data from Boston so rich white people from Boston Massachusetts and then it was working in Germany anymore so so that that's definitely an advantage but language barriers also one yeah I mean I think that I mean I think the talent in Europe is it's very strong and there are there are lots of sistex that show that we have more across Europe we have more developers in the u.s. I think that as you say the kind of deeply technical talent particularly the the depth of that talent in Eastern Europe as well I think is it's kind of a great kind of potential source for building these companies I think that I mean there's other things like some of the most innovative global people owning companies team members as
an example were well European companies to start with so we're not kind of I don't think we're on the backfoot there I think that we we still are coming from a place of strength and then I think for me the other thing that's most exciting and I think I touch them as before is that when we look up I think basically same thing when we look at applying these technologies even if we're we're not gonna be building face recognition companies that are as valuable as the multi-billion dollar companies in China when you look at applying these technologies to manufacturing to healthcare to areas where Europe has had history and the financial services I think we're in a great position to combine the the kind of great technical talent from our universities with people who have had careers in industry building building corporates or working in corporates and now more and more ready to come out of those corporate jobs than to start disruptive companies kind of solving problems that they they spend their careers dealing
with I think that's another thing that really gives us a strength that's what I guess you could build your lab your studio anywhere in the world if you moved you also on purpose so we're also thinking to to go to other places but like we also moved it here on purpose for talent and also for kind of proximity to certain industries which are interesting for for machine learning applications so next questions there's one in the back hi Andreas from speed invest I'm wondering so they are certain areas where you don't have enough data right and there are certain areas where you have enough data but the data quality is is is lagging behind so how do you see the whole like the whole issue of like data preparation and data quality and and and data cleaning and how it endures like scalable business models and drives it more towards a project-based business I would be interested in that yeah I think I think you need to think about one like standardization of the use case
so like yeah especially in like some of the manufacturing use cases like each each Factory looks differently very simply put so that there that can be challenging for example like in healthcare mammography like it's it's pretty standardized globally like more or less so it's certainly easy in those industries where some of the way the image is taken is standardized because of some law or other data if it's not the case which is also fine I think you need to in order to not do too much project based business you need to build it in this kind of self-service way so in the end if you think about I don't know manufacturing quality control you don't want to send three engineers to to every factory and spend like three months there to get the system ready unless the contract is like ten million then you know it's probably fine but otherwise you want to do it self-service so that somebody who doesn't have the technical expertise can basically just with the user interface like basically configure
the system and so that that is kind of I think how you need to think about the product and then in general like yeah you need to you need to annotate data quality is key sometimes you can go with it not that great product in the market and kind of iterate and make it better so basically already provide some value and then use the user feedback to be smart about what you annotate but yeah it's it's a big part and then the other thing is also you can work with simulations and some surveys I mean I mean you're very good at simulating the real world you can basically generate data or train your system in simulation that way kind of circumvent the whole problem of dealing with real world data I think I have a good example from our portfolio company called site code they automate accounting for small businesses and it's a space this accounting space where you would not go to because you basically don't get the data instantly so what they did
is they started with an app and they automated everything to the merchants and then in a background they were doing everything manually and then step by step they were automating through machine learning and other applications but they through that they generated data and they annotated the data which was extremely expensive so they already raised 25 million euros but that gave them the data set in the understanding of the data set we have other businesses were for example transcription in verb it from from tel-aviv we're then freelancers do it and for ultimate AI even the customers doing that for free luckily so so owning that whole process that's what we like so when you as a start-up can see end-to-end what's happening there and you really understand it and then you can start with little data at least as long as you can control it and pay to make it even better or more data I think getting the customers to do the work if the business model is or the product is right can be
really powerful so we have two companies we work with we're investors in who do that where we're actually the the machine learning enabled features are not the kind of v1 or the kind of initial product so one company is missiles called Odin technologies which is a kind of manufacturing analytics platform and the first product they took out to market was really a kind of data aggregation and visualization platform for factories so they would go out they would connect all of the machines and the factories by retrofitting them with a small IOT device and put all that data into a visualization platform and the visualization itself was useful to the factory floor managers and to the factory management but what that allowed me was getting all that data to use kind of user labeling of some sort which was that users were kind of drooling in when there was a problem and kind of highlighting those areas and tagging them for their own use to say oh I saw there was a there was downtime in
this time or this is when the machine was running outside of the right parameters and then Odin can use that data that the kind of users labeled to to kind of train machine learning algorithms in the end will allow them to provide their more advanced features like predictive predictive maintenance and kind of variable optimization of production runs so that's an example I'm like that another one is a company called Street bees that at its core is a survey in products so they go out to two users with a kind of micro survey type proposition and again the v1 for their customers who are kind of large FMCG brands is that they get survey data back about users and how they feel about their products but what street bees do is they'll ask you to take a photograph and write about how you feel about that photograph so I am one of the questions be like what we're eating for lunch and in a takeaway restaurant and you'll take a photograph and you'll say this is
a burger and I thought it was juicy and tasty and suddenly they've got really deep labeling of that image they can then start to train algorithms on and and also to do things like user segmentation which is a kind of big ask of their customers so again it's like user labeling and user inputted data that in in v2 of the product or on the in the advanced features that they can then charge their customers more for then they can unlock those machine-learning these cases any more questions we still have a bit of time else I would have the way we talk about my own those conference it always appears to me like close chop like this it's nice tiny forward-thinking AI world and everything else is old business do you really make in terms of investment in terms of what you're considering a company this is clear clear-cut what is now I driven business which is just a software or a data analyst driven business or just a business which was just not producing goods
because in reality but the clients really want to implement something I find this distinction very hard to make I mean is that I I the moment I have a TensorFlow or the moment I you have Google Maps or doesn't state distinction even make any it doesn't even matter at all in real applications I think in in most in most applied use cases of solving big commercial problems AI is now a tool part of a tech stack it's like saying I only invested in businesses using the internet 10 years ago or I invested in businesses using the cloud five years ago now for a lot of businesses it's a component and I think that like in the two examples I just gave it isn't even the core of the the kind of first version of the product now there are some businesses like the health business um around sex which is built in it at its core and wouldn't exist without these technologies and there are some businesses that are kind of vertical AI sorry horizontal
AI all kind of enabling infrastructure of AI that are AI business is like a a chip or a hosted infrastructure or something else but I think the point is right I think as we look to some kind of Applied side we stopped needed we at some point we stopped talking about AI conferences and it just becomes a core part of most people's tech stack and as as these technologies get more and more accessible as you have to do kind of less and less of the R&D work yourself in-house to use them I think that that distinction pretty quickly kind of blurs I think the right term is we would invest in data science driven companies which apply machine learning but also other techniques around data science that enables something but we don't care if it's called an AI company and I actually recommend when we do fundraising for larger rounds for the companies not to put that on the slide if they're not doing it because people will challenge you I can I can assure
you been really understands what a is and what is not so don't put it on the slide if you don't do it and and that holds its true for most of the investors out there so at the end of the day it's it's unlocking crazy new things great new things for us but for us more the problem and the solution of the problem what you actually unlocking is more important than what you do then and I guess former antics you know just don't throw algorithms at something and see what comes out of it yeah no I mean reinforce is also important like is is like a business problem right like is is there actually a business case and that that should come first because otherwise it's it's more academic research yeah I think I mean one of the things we were talking about last week was there are lots of companies now I kind of used the term advanced analytics it's kind of the same as data science I get it like there are lots
of companies who are using non deeper learning approaches that depending on how far you should get to a different definition of a I are still AI all they're using accommodations so we're we're working closely with a company now who do large-scale simulation and that's kind of the core of their product but then there are components of machine learning that kind of gets followed into that improve efficiency or or kind of increase accuracy of results and so I think we're seeing again coming back to this previous point machine learning techniques go into this kind of broader tech stack with other advanced techniques as well and and again kind of supercharged or all kind of refined things but it's in it's still only one technology and I think if you're not if you're building a company without using it that's that's not a bad thing right there there are lots of other technical approaches that can be kind of equally as effective and sometimes the right solution is a is a simple kind of rule-based rule-based system
it may be a slightly kind of lower tech mode but I also think there's that there's a growing view that kind of data moats and and tech data moats particularly as a kind of defense ability when you're building machine learning business are or not a kind of long term sustainable thing anyway I know you know I don't think I don't I don't think like having large static data sets of huge competitive advantages I mean it's a competitive advantage in the beginning but like not in the long term it gives you basically a head start and the same you know being being strong machine learning you know helps you to cook about the the products but like algorithm itself I mean everything is published it's gonna get commoditized I think the harder part is like integrating that robustly into some workflow and I think that's different now with kind of all these AI companies that's many of them are in workflows which were completely human before and very tedious or we're not even anyone could busy
leverage that data and I think that that is basic creating a lot of value that you suddenly have like kind of quite high skilled labor which now could focus on something else all right last chance for the last question please in the back hi I'm fan firm and that capital um as VC world we always look accompany or evaluate the business according to financial KPIs and marketing KPIs that's according to like normal business models or the business does not in AI world and I'm wondering now with the AI coming in and with AI specific businesses is there some specific KPIs related to it that I can use to evaluate performance of AI company I think we use proxies I mean I don't understand how an f1 score works or how I measure precision and recall really as a VC I mean I understand it mathematically but I guess I need a lot of benchmarks what we're really looking for is what is the outcome what is the result so for example faster onboarding time more
precise questions or answers that are matched so happy customers and sometimes you can even see that with MVPs so first indications that it really makes sense to use machine learning or whatsoever but the scores themselves we do we do our our due diligence with our CEOs from our portfolio and they basically challenge then the data scientists if they do proper work yeah I think it's always just basically like you should always look at the actual business value and it's very different in different applications so depending on how they see you integrate machine learning into some workflow for some cases you need to be like super precise like if you do I don't know like quantitative trading like whenever you trade you want to be right because always you gonna lose a ton of money for other use cases if it's more kind of as supporting tool for people and people accept mistakes for example I know if you want to find compliance issues and large amounts of data of corporates like you can even if
you just hire even if you just provide like hundred cases of which then only ten are in the end like very very big problems that is already huge value at and so in the end I think you always should just look at it like any other business like what is the actual business value its driving and then more interesting will be then like how how can that improve but time that's maybe like a bit more kind of vai part like if you now get a thousand ten thousand whatever amount of customers and collect more data can you improve in that way make it harder for other competitors to come in or are you basically very much for towing somewhere then you got the final words how do we get your money well I mean I was gonna say to answer that question I'll kind of roll into that I think it also depends very much on the stage you're investing as uh anyway I think if you're saying the same thing in different way if
you're investing in a company when it's all that's using AI but when it's already got fifty million dollars of revenue then clearly you're just investing or the the the a I use case is kind of proven if you're investing when it's two people and an idea and you've got to be investing in those people and that kind of technical credibility and at different points in between that kind of those milestones look different so for us if we're investing in a very technical business pre-launch we're looking to see the kind of technical part de-risked with a view that that's a key milestone that kind of drives daily in the business and that we can support on the commercialization or if you're investing one step further you want to see the kind of early customer proof point so I think it so to try not to Jessica's question then I think at each point in in your business when you're if you're going out to raise money and you're using these technologies you have to kind of
be aware yourself of where you are on that journey and so what is the what is the most recent thing that you've achieved that is a kind of valuation or a kind of value kind of milestone filled business is it a big technical team risking is it an initial commercial de-risking is it a scale-up do risking and what's the next one you're heading towards and that really helps you to kind of pitch where you are and where you're going and that helps me to be very clear about kind of where the value is and and what needs to happen so kind of to see that next stage of growth and then value great so we already had the question about Asia next coming up in five minutes we do a little break is the future of AI in Asia thank you very much thank you thank you very much Ben Thank You audience as well and thank you