Ravin Thambapillai, Credal AI: AI security, enterprise data integration, and the future of AGI

another founder wisdom buds AI Edition with a really interesting founder today uh Ravin Tamba pilai he is co-founder at credle dot Ai and he's in a Lake Michigan or somewhere up there uh retreat with his friends so we got the opportunity to know each other in this short 15 minutes and I was like why not record because Robin is quite interesting um he's an ex-bounder one of

the companies I've not sure if love would be the best word but I'd love the business aspect of volunteering division is really cool and Peter Thiel uh yeah mostly loved the guy like 90 um great entrepreneur yeah great contrarian thinker um I've obviously followed the premise of his book zero to one since I started this entrepreneurial career but yeah obviously these powerful companies they look small by

the way it's like on the front end but I think they can become the most powerful companies in the world they need to be maintained in check you know including Andrew um by the way also shout out to Palmer she's amazing founder Evan uh welcome to the Pod weird introduction but uh tell us a bit more about yourself and what do you do yeah so I'm Robin I'm the CEO of Credo AI uh crudel's uh information security company for large language model so uh Enterprises typically 1 000 plus people uh will use us so as a Gateway between their employees and all the different AI applications Foundation models and like internal tools that they want to use so uh we kind of are a place for those companies to integrate all of their data and

then their employees can kind of chat with that data or they can build applications that LM applications that use that data in a way that um actually respects all of the permissions set up in those underlying Source systems like Google Drive Microsoft Office Salesforce Etc it automatically masks any pii or sensitive information that they might have and it also automatically enforces their acceptable use policies so you

know if you're a European company you have to keep up uh comply with the EU AI act so you might have an acceptable use policy that says don't use AI to make hiring decisions if someone's trying to do that with like a CV that would they were sent will automatically detect that and alert the user like hey this looks like it's out of compliance with this policy

the company has and that's what we do uh Jack and I actually started credle a little while ago at this point we started it in October of last year which in the terms of our industry is like 10 000 years right but um obviously it's not actually that long but when we initially started griddle we had a slightly different Vision which is that you know we would

be a place where companies would be able to integrate their data and use these large language models to sort of act as a kind of strategic decision-making assistant or like a chief of staff almost and what we quickly found like building that product out was that you know because Jack and I were obsessed with security having you know come from palantir we built all of these security

features and yeah a lot of the companies and customers we were speaking to were like yeah this product is cool but those security features are really cool tell me about that because I want that in every single product I use um and that's how we sort of realized like oh actually the security story that we've built is potentially even bigger than the specific application and so we

really doubled down on doing that and you know that's where we really found um you know the growth that we've seen and since so I think there's like two aspects here and I want to explore both with you but the first is if I'm this big company and I give all this data to GPT and open AI I might have a competitive problem they might cut they might like pull the plug on me and just take all the data that's one and the second one is if openai gets hacked or anyone like in their API men in the middle or whatever if there's like a security breach my whole data can go to so are these two valid problems right there yeah I mean those are definitely things that uh our customers are thinking about

I think you know one of the things that's changed actually is like since we started when we when we started cradle open AI were like openly training on any data that they were sent like there was no like policy around there since then companies like um you know us started really pushing them to be like hey you guys need to take security seriously because Enterprises will never

use you otherwise and since then they've gotten a little bit more mature about that I will say like um as have all the other Foundation models um but you know at the end of the day uh Enterprises need two things to be willing to use basically any software but LMS in particular one is visibility and the other is control right um and so that's what we provide

like the ability to get visibility into not just open AI but any of these like AI tools that you use because people don't just want to use chat right they want to use the sales assistant and the marketing assistant and the engineering co-pilot and so you want visibility and control not just across open AI but across all of the different AI tools that you're using to say

okay what are people using what data is going to whom how can I put controls and what data can be used by whom and can be sent to him and what would be the business model here I mean you guys went to YC good job you got funded by uh spark and uh this other VC that I forgot the name of so congrats on that uh how do you how will you make money with that is it like per API call uh yeah so we have a sort of three-part uh contractual system there's a license fee for uh you know the product and that works on a per user basis so depending on how many of your employees or developers are using the product then there is a per API call fee as well um

and in some of the bigger instances we also have like an implementation fee if there's like a particularly um complex uh deployment story or a particularly complex uh data integration that they want to use like a custom data source that they have then um you know we'll charge some kind of setup fee for that as well but those are the three models and most of the revenue

actually comes from the license fee and then there's a little bit on top for the API costs and the implementation fees tell me about your background at palantir you know I started this spot saying that you know it's an amazing company like from a business perspective I think their cause is somewhat Noble as well just to defend people and if you ask me like if I'm more

uh us and NATO or Russian well you obviously know my answer I'm all about Freedom democracy and yeah letting people live their lives you know and not intervened in in theirs and try to bully people um but I still think these companies need to be maintained and checked imagine if like there's an AI in the back end and these companies can truly scale you know it's not

going to be like the classic Lockheed Martin building missiles but like truly scaling a an AI that can hurt at scale or can go out of control so what do you think about these things yeah I mean I think are the two questions there right the one question is okay there's a set of companies in you know you name check Andrew Leah Palance is another one I

think there are others also where you know they're not necessarily companies that we're always talking about from the perspective of like what are the regulations that should apply what are the rules that should apply to these companies um but they're enormously significant to our way of life right um palantir works with you know so many national governments not just the US government it's you know very involved

with the UK government it works with the NHS it works with the you know UK defense uh folks uh you know is technological significance far outweighs honestly it's like commercial value and I think there's a very very important question about what processes do we have in place that govern like the influence that um that that those relationships create I think you know from my perspective I always

found the reputation of palantir a little bit baffling that I worked there for a really long time I was there from 2015 all the way through till um 2022 so seven and a half years you know during that time I had a lot of people talk about uh you know Panther is an evil company because you know it you know works with ice or whatever and uh I never fully understood that for context you know my whole family of refugees they fled from the war in Sri Lanka um I was born in Australia like during that um time like just after they left their house was burnt down in Sri Lanka they eventually made it out to the UK and so you know the question of how countries and Nations treat immigrants and refugees is

like very very personal for me um people talk a lot about the work that part did with ice they don't talk about some of the work that I personally was involved with apparently in Syria uh helping you know internally displaced people in Syria escape from Isis and from Assad and get resettled in you know safer Nations so I do you know I saw and frankly like did

a lot of work that I felt was like very very good and I'm like very proud to have done and yes it's true like parliamentary work with ice but yeah it's really complicated to say that just because I have a contract with pounds here that the entire company is evil as a result of that right is that true for every person who you know was involved in making Microsoft Word if Microsoft Word has used it I said is it all of those people therefore evil you know that's pretty strange a pretty strange video to me um but I do agree that the influence and importance of these Tech of these technology companies is extremely high and that we haven't really figured out as a society yet how we govern and control and honestly the most

important thing is provide visibility into that right like actually making people understand like hey this technology company goes bankrupt or if this technology company decides to pull its contract like we might lose the war in Ukraine right and what is the implication of that to the relationship between the people of the United States and the you know 10 or so people that own 50 of pounds here

yeah true I mean fog of war is one keyword that comes to mind every time and or on your LinkedIn profile there's the Covent vaccines right and there's a bunch of anti-vaxxers out there yeah it's always makes feeling every time you talk about life and death there's gonna be emotions and emotions usually they blur the conversation yeah me I've had you know like I'm not necessarily for

the vaccine I'm a biohacker but I had countless of folks that I talked to and almost died from covid I heard about many deaths of covet this is serious you know so um every time that you bring up that Convoy you'll get pros and cons now Peter tells amazing uh founder it's just that yeah eqis not sure it's like I guess that's one doubt that uh people

have uh the The Ice thing yet like it's immigrations and other like tough topic right like I do want to bring a solution to that one because I'm in Mexico right now and I think if we just educate people that if you have a laptop and you stay in Mexico you can make crap tons of money you know I don't necessarily need to cross the border and guess what being in a house in Mexico is safe you know it's not uh it's not like how people think um I want to ask you though like you're in security you're in AI um a lot of people talk about the like uh stop it all button button or like the plug on AI what are your predictions with AI in the future because we only have a

like one or two minutes left here um what do you think will happen five to ten years from now with AI and as a security guy what would be measures of security or the the best course of action we could take to not have an AGI blow us up yeah I mean congrately I think we're very far away from AGI right like it's not a question of like uh the software Engineers just need to do a little bit more work and they'll and they'll figure it out it's like we need like multiple scientific breakthroughs before AGI could happen and like it's impossible to predict when those will happen just as it's impossible for us to predict like when when Newton first discovered like you know classical mechanics it was impossible to predict how long it

would be until relativity was discovered like that's the exact same situation we're in right now with like current AI versus when we'll get AGI like it could be 200 years it could be like 10 years like nobody knows um in terms of like what we actually need to do about today's technology like the real stuff that we have in place honestly there's nothing more valuable than just

putting it in people's hands in contexts where it's in control like people are in control and being able to understand and learn about how the technology she works I think there are specific examples or see like people suddenly realize like stuff about the like how the actual technology works that makes them much more safe and secure and how they use it I think one of the really

good examples of this is like stop sequences when you like see a developer learn how like stop sequences control like what the model's output is like they get much better intuitions for what the model is doing under the hood and then they're able to design much more like safe and secure systems as a result of that so just like putting the tech in people's hands in a

way that's safe is like or in a way that's like in a box and controlled is like easily the like most important step for educating people I also so I would agree with your um AGI predictions with the multiple breakthroughs um I think these can have been pretty fast if you put like a bunch of brains in one room they're like Miracles creators very smart folks especially

if they work together under a great leadership like we've seen at open AI um that being said I don't think we even need AGI to disbalance this world I think like the next big thing will be agents that you deploy you know I have a VA right now and I think that nay I could do it like a better job to be frank um even with the current technology I just needed to give it my accounts and my permissions right I just need to log in and tell and sign the legal stuff saying that if they up then it's on me um how far are we away from this you know like me deploying a bunch of a agents one for sales one for marketing I was about to say one for HR but no

not for HR because I wouldn't yeah more HR uh one for uh legal and run for product let's say how far are we away from that yeah I mean in terms like fully autonomous agents agents that like need no supervision I I honestly think we're still quite far away from that in terms of agents that like assist you and like make you much more effective we're not

like that's today you know that's like literally happening already you see in that you know if you look at the YC startups in my badge like all of those companies are like loaded up on these AI tools right there's the finance AI tool that does their like bookkeeping for them and then there's the like you know Jasper AI that writes blog post whatever and obviously like those

tools are not fully autonomous right they are taking the first graph maybe even like doing the second draft doing some editing but they're still fundamentally under the supervision of a human they're doing maybe like reducing somewhere between let's say 30 of the work in in co-pilot's case to like 60 of the work in Jasper's case um but you know 100 thing even for like relatively narrow domains

maybe maybe the EA example is like an example where we might be closer but for anything where like it really requires like someone with like a very switched on brain to get it right um you know I think we're still a little bit away from that and honestly even in the EA case really good EAS are like really really smart right and uh yeah you're not going

to replace that with with chat TPT uh for years I don't know interesting well Robin thank you for coming in today where can people find out more about you yeah you can look us up on the web on our website credle c-r-e-d-a-l dot a I um and cooperate uh follow us on LinkedIn or Twitter where we talk about security and and what our customers the amazing things

our customers are doing with AI today and I'm Charles Cormier host at founderwisdompodcast.com that was 11