Transcript: DevReal: Leann Chen Interview

hello everybody I'm Alexi cabro founder and organizer of baera AI which is the biggest longest running AI metup in the Bay Area and probably the world and here we are now at Amazon we Services J Loft which used to be the start of Loft and now it's exclusively J Loft which we open went a bit ago and now we're back with a massive and awesome AI Meetup uh we have three talks today and here we have leanan who is developer Advocate dbot and uh we're really looking forward to her talk uh so but now we just do the meeting greet and uh to get to know the speaker so welcome Leon thank you thank you for having me so you're developer Advocate at dbot what does dbot do so in two sentences defb cross the entire web data and organized into the world's large uh the world's largest Knowledge Graph we organize the internet data into structure data which is called knowledge this is super exciting right like as I mean I'm now yeah architect at G we work with graphs knowledge graphs are fun and like putting the world into Knowledge Graph is super exciting so uh I mean how did this happen why you know is this interesting to you how did you get engaged with this what's so cool about it like tell us a bit more wellow um so long story short um I actually discovered knowledge gra um around two years ago and like back then I was doing some like network analysis and and I I got really really interested in like like the relationships between like different such as such as the relationship between organizations and and people like who work at which company and like the experience and all of this these different data points can be organized into a network and like that is very uh interesting to me cuz CU like right now you literally have a like like a map of of like things to get connected with and learn about right so uh that's a starting point and I actually learned a lot about knowledge graph from uh toas foran like he's he's going to be interviewed at this but uh I he really shared lots of great articles and uh also sharing some like codes on GitHub and I just start playing around and then also I like to get like kind of connected with more people in the community so I start making YouTube videos and like people find it interesting so I just like um get more involved in the community and like so right now I uh make educational content trying to contribute to this open source communy this is really awesome and you you told me before the previous meop right we had this like multiple companies right we had like writer and others all together and you was speaking and like it was super cool talk and you said like you're a NAD you're digital Nomad you travel the world and you take the show to the road previously yes um so yeah well I uh previously I I travel like from in Canada Japan Thailand and I'm from Taiwan I actually spent most of my time in Taiwan and uh because as like my main um tax is like I I want to come up with content like people would hopefully people find interesting and like valuable to them so like uh the ability to travel around and and gain different perspectives from like networks or uh networking or just like you know see how people think and getting New Perspectives really enrich that proc process of like see how can make content better content so this is interesting right because Taiwan is the birthplace of gpus as we know them right and Jens is and I a bunch of like people from Taiwan like there is a lot of strong presence in the community uh but you know but area is kind of the epicenter of Jia Revolution so do you notice this like when you travel do you believe that we have this critical messy which is different do you also see kind of communities of gen other places do you have to be here to do it or can you be somewhere else well um so uh originally I like to be fair like I benefit a lot from um social media such as like I um I know I knew about much stuff from like um social media like medium articles things like that and then we further connected and and like there they just like open up the doors like there's lot actually a lot of insight in these like hidden hallway talks so to your question yes I do feel like right now um and specifically in in San Francisco the area that there's a higher concentration of how people are building things how people are passionate about Ai and you know as like they're building things they like through those like trial and errors they will share some of the Insight so it's the the brainstorming effect is uh a lot stronger uh compared to the other places cuz like uh for example in in Taiwan it's um like we we're more for focus on the manufacturing part psmc is is is at so um of course it's you you can definitely feel a strong contast between like the hardware and software and how people are like moving so fast with like gen yes here so you know I think you know so what you're doing is very important because uh you know I'm also now in the organization at NE forj and so what I find that you know the developer advocacy we kind of Frontier we Vanguard of this revolution like you come to the mid up and developers are looking at you to explain this stuff to them right but nobody really know what's going to happen right like so so how do you deal with this uncertainty right people basically learn from you and you need to explain this to them so that they can go and build it right and they like need to do something fun uh but uh everybody is kind of in the same place like we don't know what's going to happen everybody's looking at like all these models coming out and every week they have new stuff how do you handle this uncertainty this stream of information so first of all I want to um also like make this claim is like uh I even though I make educational content on social media I I don't actually see myself as an expert because literally llms or became uh have become popular literally like two or three years ago right so actually the fact is we are all exploring things right yep and like this is a good question that you asked about like how to handle uncertainty because they're they're so much unknown but the way I look at uncertainty is actually okay uncertainty actually opens the door for us to collectively explore new things and learn things together M so that's how I see like um the the the network or the community on either social media or in person is like we are getting connected like when NE has been like emphasizing a lot like stay connected uh build relationships and and get like those insights or information uh getting getting getting across and that's how I see things and because like the UNC another part of uncertainty like what it gives us is actually we also have the space of making mistakes which I I think like why silicon Valley or or San Francisco has been so um so so known for like startups because if you already have a known Ro you just literally follow the procedures but if it's the space is unknown you get to make mistakes and learn from it and improve so I think right now if we can like build this like like open source community that learn from mistakes and improve then that's actually the like the the ideal part of like having this community absolutely I mean TR ER right like and and really that's the only way forward uh so uh knowledge graphs right we talked about it you know the gra gather and other places it's like really interesting and so it to me it's fascinating because I remember semantic web right like I was always interested in this and then like symbolic a I kind of went away and statistically I came forward and then you know deep learning and then suddenly llms like hallucinate and we look back and find the knowledge graphs are useful right so I wonder uh like do you think kind of uh that's you know so like we need to build the systems right we discuss it there's a lot of work in this so uh and obviously looking forward to your talk but like why do you think knowledge graphs are helpful in this new llm era yeah so um well like l llm i uh for example what have experience with Chad me is like um yeah it it gives up this like magical like conversational experience right and and like that's pretty cool that's so accessible to to the to the the general uh I mean the general audience to people that's great but but like if we're talking more seriously about how do we verify the information that we get from lolms are accurate right it's not like LM have this like feature of like hallucination right such as uh I mean quoted from Andrew capari is like Hallucination is not a bug but a feature of LMS so how do we and a knowledge graph is a a structure of deterministic ground truth it's the the ground trth that you have in your data so how do we like H how can we make llms more trustworthy or like more more like explainable of like okay the information that I get I know why I'm getting this and Knowledge Graph can give llms like sort of like like a map that LMS can re reference and said Okay I I see you know this uh there's a relationship between such as Nvidia and and AMD because they're competitors and LM can uh refer to that information and return the output for you uh instead I mean it's um it's a method to reduce hallucination it's not like a guarantee like uh like 100% definitely we're still working on it but uh it it proves to be helpful to reduce hallucination absolutely and uh you know kind of uh my last question is about so like you know we've been at this meetups with multiple other startups each of them does the piece of gen St right some people do data prep some people do Knowledge Graph some people the vector rag right and there are different kind of uh folks and a lot of them work in open so kind of you know as being kind of commit architect I'm always thinking how can I put together some kind of gen stack right because people are starting they have hard time like what should I do this or that so what would be your advice let's say you know developers are smart they can program various languages they know some devops but they're not AI people right so they come to you like and they ask you like Leon you know it's a great talk I want to build Knowledge Graph I want to do this kind of stuff where should I start and like what would be your advice to people right who like are good developers but not knowledgeable gen and they want to engage you know and can we have like do you think we can have this kind of stack or do we like does everybody have to pick their own pieces and like what's kind of a good approximation to an open source stack MH well this this is um this this question is easy to answer my simple advice and the advice that have been giving to people for long time is go to medium and type toas ban and watch every single articles of his and just Implement his calls it's very hon very straightforward and you will learn a lot and and then he he's also uh like he's actually here he he's very uh he's very kind and if you have any qu not not not saying that he's going to reply to everyone but uh he he's been very active on social medias and and I think hands-on experience very important and definitely go to his articles and you will learn a lot byting hands off from his CL fantastic well thank you Le we appreciate it and uh hold on we're gonna talk to toas in the minut thank you very much thank you