Josh Tobin Interview: Founder and CEO of Gantry
Recording: Josh Tobin Interview: Founder and CEO of Gantry
foreign [Music] the founder and organizer of Bay Area AI a Meetup which was founded as an NLP Meetup 10 years ago and then went through data science machine learning deep learning waves and now of course Ai and now it's of all llms and we're back after the pandemic Hiatus we started a small last fall and we're coming back and forth we had our first full Meetup with Josh Tobin our founder and CEO of Gantry and startup which is doing testerone developments of fellow Lamson he gave a talk about uh tested and developed fillers which will be released separately so we just wanted to meet Josh and ask him questions about you know how he come to do this and what's really making it fun so thank you for sitting down with us thanks to be here uh so we met at LM boot camp which was huge Global success can you tell us a little bit how do you guys start this and you know how you gather such an amazing audience yeah so it kind of you know go back in time to 2017 early 2018 uh world was a different place um you know lens were just uh uh where's the paper there's a twinkle in the eye of some researchers who had this idea of scaling things up um but AI was not you know and um a lot of people were really excited about um AI back then in particular they're really excited about deep learning um so I at time was working on my PhD and I was kind of simultaneously working in openai um and had realized that like a frustrating thing about that stage was that you know at some point the hard part for us and opening I stopped being research and it started being like everything else around that like how do we organize people to work on these projects together how do we build infrastructure how do we scale it up how do we um how do we test our models and um so that Insight led to co-creating this class with um with my advisor assignment Berkeley and um one of one of our other friends who's running an AI startup and really the the Inception of it was just you know we had um we wanted to do some teaching over the summer um we got an approached by a company to teach a class for them we talked about the the curriculum that we want to teach and we realized like we want to teach stuff that was really different than what most other people were teaching is focused not on How to Train models but actually out on how to build products and models um and so long story short rather than teaching that class for just for that one company we decided you know it would be more impactful if we made it into a boot camp um we ran that and it was like way more successful than we anticipated and the community has kind of grown over time ever since then how common became so International like people flew from Netherlands from Brazil like those like really Global showing all right how did this happen I don't know I think um you know I think we we historically um we always do a really good job of um you know putting our content online for free after we teach the classes and so it's become over time just kind of this repository of uh stuff that you know we put a lot of effort and a lot of thinking into um and so a lot of people you know find it valuable and it's just online for free so I think you know um there's you know maybe not a lot of university courses or in-person classes in um most of the world and so folks gravitate to these these uh these free classes um but then when they see that they're having an event they you know want to come and follow Community yeah that's you know that's a Hallmark right like when people fly from around the world and all of them I talked to the person said like I am doing software engine jump on a starter top I want to get a machine learning so they see your course as a pivot to get into the space right and I think it's like it's a it's a ground floor of a lab so you give a lot of people a new chance in this new era so it's like I think this is awesome can you tell me a little bit about like there are three guys who do the element like like three musketeers you all have different character right like can you tell me how how did you guys click right like you work together as a team yeah um you know uh it was all it it was all sort of circumstantial and it just kind of ended up working out really well um uh Sergey and I met um to start working together on this first class um Peter by advisor um who previously was you know helped out a lot with the class and uh you know these days is more of an advisor um what you know sort of brought us together and we just you know we found that we worked really well together and then you know fast forward a couple of years we were both running other companies um but we wanted to continue this community and uh we met Charles maybe that time was working at Wheats and biases T2 classes there we watched some of his lectures and we were like this guy's really good he really great educator um he is it really is a little spot what's up well you see the spot yeah um and so we uh you know we um uh approached him to chat and you know turned out that he was interested in working together and you know we started teaching classes together from that well this is great yeah I think like each of you who has like very specific contribution to this right and like you know like I think yeah that was this is a great I think you you it's a great team can you talk a little bit about you know your GPS right all you guys worked there and so you know I collaborate with a lot of Open Source folks from you know amp lab and Rise lab and now Skylab so Burger is this culture how did that affect this yeah you know I think um from the beginning we were running this as a side business um so you know uh we um and like part of the the original sort of motivation for doing this was to kind of have this you know the side business that we could work on and um make some money from but because of the connection to Berkeley you know is always really really important um especially to Peter but then also to the rest of us to uh you know not like not pay all this content forever but to you know um yeah like make sure that we're breaking even on it and then eventually just open it up so that it can impact more people and another thing should really impress me is that you guys now have you know startups like you're a sort of founder right like server Founders generally like don't have time to answer emails uh right and so you are running education events and talks and right right like you're really kind of focusing so how does how do you manage to do that and you know why do you consider this important right this educational component yeah um well you know the the uh the how do I do that well I I only write content about stuff that I'm already thinking about and already really care about um and the way that I develop content often leads to ideas that get incorporated into our product or reflect things that we're building into our product anyway so there's a ton of overlap um and the reason why I think it's important is because you know we're kind of that um we're we're at it um an inflection point of adoption of a new technology and this technology has the potential to have massive amount of impact right it's like um as impactful as you know whatever metaphor you want to shoot impactful as the internet as impactful as software itself or maybe the most important technology that humans have ever invented um depending on how far this scaling curve goes um and you know but in order for this technology to have that impact we it's not just going to be models themselves that are going to drive that um it's also going to be how we build how we turn those models which are this raw technology that's really capable um how we craft those into the actual products that having impact on the world um I think in order for that to happen there are new technologies that need to exist and so that's the that's what I'm working on that's what we're working on at Gantry um but it's I don't think Building Technology is enough because um we also need people to understand how to use this technology to solve problems um and you know if you think back at the history of software um you know it wasn't just tools that made software successful it was also the development of you know agile and uh you know practices around how we can work as organizations to develop software and so I really think you know for me the mission is um accelerating the impact of artificial intelligence on the world and you know seeing AI become part of all the products that we interact with on a day-to-day basis much like software is today um and Gantry and uh the full stack are sort of are the two thrusts that I'm that I'm working on to help achieve that this is great you know I really appreciate that because a bunch of Engineers so that's kind of probably my last question is so you know I run a bunch of Engineers who have tens of thousands of Engineers right and who do backend Java Enterprise spark Big Data data pipelines streaming you know react for the ux and so forth right so we have a whole class of Engineers who are now kind of wondering right what's going to happen to them and uh one of my friends was at Google she said that very interestingly like if you think that AI is going to get rid of software Engineers it's like saying about magic is going to get rid of transitions right so so which which kind of right which which you know is critic but you're in the midst of this you build software engineering product with Engineers right what do you think about this and what's your advice social Engineers who I came today who want to learn about this and pivot into this like what's your advice how to be a software engineer in the age of Alliance yeah absolutely um I think you know uh maybe in the future um AI is going to be so good at doing everything that humans do that there won't be much of a role for us um maybe but in the short term which is all we need to really control um this is a I think people should think about this as a really capable tool and it might um and so you know right now like I think most software Engineers will be better software Engineers if they learn how to use hell UPS like when I write code especially when I write code and you know uh something like JavaScript I don't really you know use very often or not understand that well I probably a lot on co-piloted gpt4 and stuff like that helping um in the medium term you know the the capabilities of these like AI driven programming systems might increase very rapidly there's a very high chance of that and so um the tool might become so capable that it might feel like it's it's you know replacing a lot of the parts that software Engineers do in their day-to-day jobs um but what it's not going to do is it's not going um at least not you know until we get to this point of being close to AGI it's not going to make decisions about what we should build it's not going to make um judgments about what is um you know what do we care what do people care about when they interact with the software um and it's not going to be able to balance like all the constraints and trade-offs um that only come from like conversations with other people and a deep understanding of the problem that you're trying to solve um and so you know you can think of it as like software engineer as software engineer ultimately your job is to like build things that people use um and that they get value from and the you know um you know just because you uh like just because a better tool is invented like if a better tool is invented to build those things then it's your job to figure out how to use it um and it's and if you do that then it's not going to replace you um you know at least until it replaces everything sounds great and of course like traditional questions like what's next to a country I saw your product which is really really cool like it looks like a lab which people would actually use to to do this tdd which is a great engineering concept so what's next like do you want to diversify in different areas like whatever physicist or doctor do I need like special tools or are you going to say General TD to any ideas kind of where it's going yeah I mean folks are just gonna have to wait and see where uh you know we're uh not quite ready to talk about a little bit about that yet sound good thank you very much Josh we'll be following you and looking forward to news from thanks so much thank you [Music]