hey everyone welcome to sck podcast called AI business and comedy uh if you like that type of stuff hit that like button I am joined today by my friend um Tim novakov uh he is actually I have gp4 wrote an awesome intro so we read it um today we are excited to introduce Tim novakov the Innovative mind behind super teacher Inc where he currently serves as founder C and CEO offering unlimited tutoring for just 10 bucks a month since June 2021 before this Venture Tim held pivotal roles at Google including lead product manager at Google research where he was spearheading new product Innovations and manag Google collab from 2018 to 2021 he also served as product manager for Google photos leading teams on assistance editing and sharing and integrating Advanced computer vision Technologies Tim's entrepreneural Journey began with fly Labs Inc a startup he led from the Inception through its acquisition by Google and academic at heart he has also contributed to Corell University as a lecturer and researcher creating courses and advanced algor algorithmic education Theory let's give a warm welcome to Tim novakov so I always say that chat GPT its voice is a 24-year-old uh Yale graduate who majored in political science who's now working in HR is also a doula likes avocado toast and reads Harpers so that's how it and as GPT 40 is a little flirty also exactly exactly it's like I I you know it's like it's like hey you know he wants to put some game at me I'm not like her or anything you know no what did you do did you just ask chat GPT you know can you write an intro for Tim novakov that sounded a lot like my LinkedIn so I'm assuming that's where it exactly I did a rip off your LinkedIn I said this person is Tim novakov please write an introduction for him that I can read on my podcast one paragraph and I could I be be more specific I can say like write like you're someone who's um uh I want to write in my tone which is like depressed um a little bit narcissistic self-hating self-loathing hates corporate just hates life in general go sure right like Jordan Tibido and then that way so so Tim thank you for joining us uh really appreciate it um so uh me and Tim go back because uh when he had fly Labs he sold to Google and I was working on that that deal with uh Megan Summers and so that's when I first met Tim and uh it was been it's been fantastic ever since and we we've stayed connected which is good um it's it's usually uh HR and entrepreneurs are like friends like cats and dogs but actually I was I shouldn't have been in HR I was kind of like Island to misfit toys and Tim was like there's probably a fit here I think I can consider person friend I thought the entire time you were part of Corp Dev oh really I didn't even have hair oh come on I didn't have slick backed hair I wasn't wearing those those women for me you for me you were like just part of the m&a team and you know when we visited Google before the acquisition I think you were the first person I met you sort of greeted us you shephered us through and then when we joined you were also the first person you know who uh who shephered us through or orientation and so you know cuz the other interactions besides with product were with Corp Dev I just assumed you were you were part there it's it's okay you were part of HR I won't hold that against you than thank anything you stay sharp no if I go to St Peter's Gates it was like oh sorry you you're at HR for Google first of all you read at Google during Google Plus like Sorry God does not not like that product he's a Facebook person uh no no I I I that's why I loved about being uh in m&a was I had to meet Founders like yourself because you had outside experience you went out tried to build something and it was just a really really cool thing so I want to start off with like you were working at Cornell you were on the academic track there and then you decided to um like you were lecturing there and then make a pivot into fly Labs maybe you can talk about that yeah sort of that that's uh not quite right uh so I I finished my uh PhD from Cornell and as I was wrapping up I started fly labs and I was working on labs for the few years after grad school but you know initially there was no investors there was no investment there was no money and so there was no pay and also I had no money in my bank account and so in order to yeah in order to make ends meets and this isn't this is a slight exaggeration but basically I started at the same time as I was doing the startup teaching a course in the computer science department of Cornell in iPhone app development and that sort of ser served a double purpose once for one they paid me you know a lot of money $7,000 I'm rich I'm rich and uh you know like actually when you have no income whatsoever you're just doing a startup you're trying to find investors you know you're um at that phase the pre-funding phase that's actually great to have and also it was well first of all I was you know it was related to what I was doing day-to-day which was iPhone app development at the startup and also it allowed me to recruit potentially uh people from the class after they graduate in fact I we did end up having an intern who was really great uh from from my my class at at Cornell and you know I've met some pretty amazing computer science students who are now doing great in the tech industry through that course and so it was also just like meeting great engineers and you know when you're when you're an entrepreneur when you're starting out even if you are an engineer a big thing is knowing lots of great engineers and so that's that's something that any great founder is always doing is just trying to keep in touch with all the best people they know in terms of um you know software Engineers they're the ones that you know build the things so those are some really good points Tim now uh I mean people like have this they believe some some people believe like my technical skills Will Set Me Free I just need to code something great and then all of humanity will give me money this whole networking and people stuff and whatnot maybe you can give some tips to Engineers on just like how can you network without selling your soul or how can you stay in contact with people like any any thoughts on that yeah it's a good one it's hard I don't actually enjoy well there's different types of networking actually networking at like technical events or like you know where you're meeting other Engineers um that's fine I love that actually because I think the conversations are really interesting you know as a Founder you often have to or maybe should go to networking events where it's more like other Founders lots of investors and that can be a lot less fun partly because I mean it it can be fun but you know it's it depends on if it's your crowd and if you really enjoy that I I've gotten used to the fact that I have to do a certain amount of networking in order to be an entrepreneur and so I go to a certain amount of you know these events but it's not easy I remember once going to an event and you know I just wasn't into you know the the hobn knobbing and the you know here's what I'm working on and everything's going so great uh that that everybody was doing and I was just kind of hanging out by the side and there there was another person who was just kind of hanging out by the side and he looked kind of nerdy and like you know he was in the same place that I was and so I you know started chatting up with him and he was great he turned out to be the the founder of stack Overflow forgot his name Joel um Joel something no way let me stack Overflow founder let's see here uh Jeff Atwood and Joel spolski Joel spolski yeah and it was a super fun conversation turned out we knew people UNC common and we ended up meeting up again in New York City he actually had lots of great advice for me um even advice about entrepreneurship and fundraising but he was clearly in the same sort of mental space that I am where it's like you know certain certain types of these events are a bit of a burden but I do think it's important you know I'll go to networking events and maybe nine times out of 10 nothing really comes out of it you know um you know you meet some people you know the old days you be exchanged business cards now you like take a picture of the of your the other person's badge or you just like scan their LinkedIn and you connect or something like that but it's people that don't end up playing any type of role in your company in your company's Journey you know that they they're not business partners they're not investors they don't make any useful intros you know you just kind of got to know them and who knows what what'll happen but some really key things have come out of networking and when they do matter they tend to matter enormously so as an example my last company one of our investors was SV Angel you know a fund out in Silicon Valley and I went to their CEO conference and um you know just kind of meeting meeting random people and at one point uh this great investor uh to for Conway he says oh here's you got to meet this person from from Apple you know Sean Pruden and she's actually very senior executive at Apple she I think she might be in charge of devel relations she's the person that presents the apple design award all the time and she um she just made introductions within Apple to the App Store people and then Apple started featuring our app as like you know number one featured app in the photo video category which is a category that we were in but then also number one across all categories and then it was like in all countries and then that led to being featured in the Apple retail store demo devices so you go to an Apple Store you you know pick up an iPad uh our app running on it our video editing app because it helped Apple sell more iPads they made an introduction to Starbucks and we ended up being like Starbucks's app of the week which was is surprisingly like a big deal you get like hundreds of thousands of uh of of downloads to that and so all this came you know through that that you know introduction and it's hard I I never lose track of the fact that a lot of great things came out of networking they all almost all came out of that with one introduction that came out of that networking event and then like this wasn't networking event but it was also just sort of an introduction more like a email networking yeah 10 guy toer Conway just like fantastic investor and he introduced me to Dave leeb at Google photos he was the lead product manager at Google photos Dave yes now he's at y compor isn't he he now he just became partner at YC yeah and that introduction led to Google buying my company and US joining fly laabs and so these things both came out of networking and so I probably went to like 30 or 40 networking events you know maybe there's a tiny tiny bit of impact from a couple of uh you know people I met um besides those two but then huge amount of impact from the you know two most important connections that got made networking so I feel like it can matter a lot but it's like a um it's like a I don't know what the right analogy is but it's like like a spray and prey type of approach you meet lots of people and then like you know you pray that one or two turn out to be you know really big we just started a Discord we have 71 people in there you can get access to it by uh becoming a patreon or YouTube contributor for $299 a month have a lot of fantastic people in there people introduce themselves we talk about robotics in here we have a channel for AI we also discuss recent episodes and you can help us create new episodes and see what Joe and I are going to be talking about before the episode happens so people can come in chime in suggest topics we should we should talk about then we have places people request uh interviewers and folks that they think we should be in the podcast place to request what kind of sounds we should have the podcast and the soundboard place for research people share their research papers in there themes and comedy gaming entertainment investing documentaries name it we got it so if you can hit that join button right now or support us on patreon so you can get access to this community it's a great investment in your future also we will be um live streaming now every Monday Tuesday and Wednesday from 11:00 a.m. to 12:00 p.m. Pacific Standard time you can catch us on YouTube and twitch that is Monday Tuesday and Wednesday 11:00 a.m. to 12:00 p.m. Pacific Standard time thanks talk to you later bye 101 would listen to this again we have a media board so I'm going to give you one of these sweep the leg no mercy that's of course you know that's what that's from right cry kid KOB Kai he was like Sensei he you want me to sweep the leg his ankles hurt he's like sweep the leg no mercy and that that means when someone does something really cool new AI launch has really good tip you going to sweep the leg so thank you um also such good tips I think one thing it came across was you're opening yourself up to luck you're just going to events and like something happens great if it doesn't uh that's fine I'm just going to keep on talking and see what happens here and keep on meeting people and that's what you got to do it's sort of like prospecting in some way kind of like sales say great way to put it yeah exactly so I thought that was really good and then you met Joel the founder of um stack Overflow what were his some of his like if you remember some of like his highle tips on fundraising because I get Founders to come up to me and ask me this question I'm like I actually I do remember because it was it was great you know we were putting together a round for for fly labs and I was thinking about some different ways that the round you know could be structured and um I thought he might be helpful his office was right near my office and so you know I you said hey could could you give me some advice and said sure come on over and we chatted and he was like you know he barely remembered me you he's sort of a big shot I was you know nobody he was like you can you remind me what is your what is your startup do I mean I've told him like trying to think about you know the round structure and you know this or that and and actually tell us real quick what what fly does and then we're going to go super teacher after sure flyabs the company of the time was a mobile video editing app so we were the number one video editing app when video editing on on mobile devices was brand new so we were the first company to sort of like crack that uh be able to make that happen and so he's like yeah remind me so I you know I told him that um he's like okay cool listen Tim that's your like expertise that's where you innovate don't try to innovate on like how the round is structured and I know like as you know and again I was sort of like an engineer at the time like as an engineer like you have that mindset where like if I you know you can adjust the mechanics and this all that this works that way and such and such but he's like you know there people have already figured out how to do rounds you know at the time it was like convertible notes and there's like price rounds and you know now this you know for the early stages convertible notes for the noobs what's that mean convertible notes is sort of like the pre predecessor to Safe notes um I think for the most part people don't have to know about convertible notes if they're like um you know preed seed stage companies anymore because everybody uses safe notes but uh they were essentially loans that will eventually convert into the terms of the uh the subsequent funding round so just like a safe note safe notes are better they're easier they're cheaper uh there's more protections all around so safe notes are better um that's why everybody uses them now but um yeah he's like listen there are a few structures that everybody knows works again this just Prett safe so convert note was the other one price rounds the other one there's a couple of you know mechanics there's the you know the you know the value liquidation preference some other stuff that you hopefully never have to deal with as a as a Founder like participation rights and you know sort of like uh investor friendly terms but like that's it just deal with the things that are out there they can solve all your fundraising problems you you're not you shouldn't be trying to innovate on fundraising you should be trying to innovate on video editing on on mobile devices that's your thing just do the normal stuff when it comes to when it comes to fundraising and I've taken that not only to be true about fundraising but kind of about a lot of stuff you know there's the stuff you're trying to innovate on and there you really are thinking from first principles and doing things that nobody's ever done before because that is you know your actual core competencies your area of expertise and then for a lot of the other stuff just do normal stuff that's kind of people have done it you know you you know how do you structure like a a you know what you know you shouldn't be running your own payroll you should just be using like you know just works or austo or something like that you know just do normal stuff when it comes to like healthcare 401K stuff for employees just have a normal vacation policy like you just don't need to overthink any of this stuff just kind of do normal stuff and that saves your energy for like okay how are you going to actually make video editing possible on you know iPhones and this is like I don't I don't remember what era this is but like iPhone probably like five or six or something was a long time ago um how you going to make that possible or how you going to make an AI tutor that can you know tutor children whatever it is that you're you're you're building that's where you innovate and everything else just keep it simple that's perfect well said I mean because you have so much cognitive load going on with just building a product that's good and all the other things going on focus on that enough exactly then oh I'm going to innovate on payroll I'm also going to innovate on corporate structure innovate no just it's just way too much and then you're also saying is like don't over engineer things you know and I I you probably have dealt with Engineers too who like hey I want to go build out this new feature here for some Edge use case that might matter and you're like no focus on this don't get get caught in these random paths that can lead to disaster now you mentioned also I think you Sundar you edited one of sundar's videos or presentations with the app can you talk about that right so I think Google IO is happening right now like as we're recording this right or did it already happen I don't know it already happened um ke happen it's like a full week yeah um so that was on my mind as we were uh exchanging emails this morning um yeah I guess the I mean it's not much of an edit but you know the story was that uh I was a a product manager at Google photos at the time and I was in charge of a lot of the like the AI features that were using breakthroughs in computer vision that was coming out of Google research so that that's sort of how Google works actually Dave Lee explained this to me and this certainly how Google photos interacted with the rest of Google that you know first of all Google's a bit like a microcosm of the entire tech industry you know there's everything there's Hardware there's you know OS there's programming language people there's you know you know application layer there's you know storage everything happens at Google and and outside tech industry you might have lots of little different companies you know providing various bits of this you know in the startup world there's this notion of an API startup you know a company where you know their main product is they make you know some online service and they sell access to their API right and Google photos or Google had had a similar thing in the sense that there were these research teams that would do research in you know the equivalent of foundation models that's not what they were called back then but basically you know imag net and like you know um breakthroughs and computer vision and then they would basically try to find products that could actually use their Tech and they were almost they like making sales calls to to product managers like me and saying like hey could you use this so there was this fantastic researcher there uh her name was y prit and she was a leader of the computer vision team and she had figured out a way to do image segmentation where it could take a photograph and then it could identify you know which pixels corresponded to a person including their you know their clothing and their hair and anything they might be you know carrying uh from the rest of the the photograph so if you have a person in the foreground um you can isolate just the pixels for for that person compared to the rest of the photograph and she's like is there anything you could do with this in Google photos and so I thought about I said you know this would be kind of cool if we uh if we could do uh we called it colourpop where basically you find a person who's wearing you know you automatically detect photos where there's a person wearing bright colored clothing uh ideally in in in a background that is also somewhat bright colored and you know you have other her istics like they're smiling you know we have a good sharpness score they're kind of their gaze angle is somewhat just the camera says like a good photo of somebody and then we'll leave the person and and their clothing in color turn the rest into black and white and make it sort of an automatic Creations like hey you know we made you a new a new like a a new Twist on one of your recent photos you call it colourpop and so we ended up building that feature and launching it um live with the keynote at google.io and sindar I can't remember what he what he said about it but he announced it as you know part of you know like a a trio of features launching in Google photos as part of his big keynote address and so there was somebody who wrote the you know his script maybe it was Sundar maybe it was you know a script writer or something like that and then the people who are you know relevant to the things he was talking about actually got to see the script in advance and make sure it's you know technically accurate or or what have you and you know it's interesting the secrecy in this is is was was fun to witness they didn't just give us access to the whole script which is you know for an hour long keynote uh it was actually chopped up into little bits so we could only see the bit that was actually relevant to us yeah so it's like really really you know um kind of security on it so that was kind of cool to just be part of that process and you know I I edited whatever it was about Google photos and um and colourpop and I just remember Sundar somebody had put into the script some announcement about something else that was coming and it was something that would have been within my domain I think it might have been taking a black it was actually like the opposite in a way taking a black black and white photo and like colorizing it a a feature that we had looked at technology that we had looked at many times and and I knew from testing it that it was ways away from being ready for prime time and that I I had my doubts that that that whatever the feature was the technology was would actually make it out in the next you know six to 12 months it would maybe even be years and just can't remember whether it was a whether it was colorization or something else um but the previous year Sundar had already announced like something in Google photos that never ended up launching and I was just like what you know you know what is this that's not something that that um that I was aware of and so I I made a comment that like you know hey this little thing I'm not sure this is actually going to launch anytime soon and you know I'm afraid that if uh you know SAR makes announcement like this you know especially coming on the heels of something else that never end up you SAR would end up with a reputation of you know being all hat and no cattle like an old cowboy expression well I love that expression continue and uh I was you know really pleasantly surprised somebody I don't know if SAR or the marketing head or somebody actually took that bit out and so I actually had a tiny little bit of an effect uh on sar's uh sar's talk outside of the context of what we had U announced that's great so many questions first of all uh good for good on you for calling it out so sudar didn't have to commit to something that wasn't even I just wanted to make sure that they were kind of aware of it you know it wasn't my call to put it in or out but like just so you know this thing you know exactly Joe used to tell me um they were he knew that when Eric Schmidt was in charge of the company and before they would do presentations on new products to the Press he would go to all product managers and be like do you know your crap on this thing because if when you're presenting if they stump you on the question and I have to come bail you out I'm not going to be happy and Joe Joe told me one time it was some who who's Joe uh Joe taski uh he used to actually back in the day be my manager when I was an admin and now he's he's Jo he's co-hosting the show but right now he's he's he's he's on a pathway to uh he's getting ready for a vacation and so that's why he's not here right now but he wanted to be here um but he was telling me that I guess Google was launching Google pay or wallet or something and they did did a tour to the press and the product manager is talking about this the Google pay or wallet whatever the hell was called and then it was some journalist was like so how is this different from like apple wallet or Apple pay and the product manager is kind like self-destructed on the question and Eric Eric swi just came in like like an Android or gpt2 and answered perfectly and assuaged the fears of the of the journalist oh and then it ended and he was not happy with that product manager oh wow my job to bail you out that you're high six figures plus Equity you should know your stuff so uh that that's that's that it's also it's an impressive story about Eric Schmidt although I found that to be true about a lot of Executives at Google that they are the real deal I've heard many people say Eric Schmidt was just insanely smart um you know our RVP at Google photos this guy Anil saberwal he was that way too he could seemingly speak extemporaneously about any aspect of Google photos but in like perfectly crafted crisp like marketing sentences you know like ready for the public but totally accurate and and you know everybody else was kind of like rehearsing their bits for like TGIF and like know was just like you know I could probably Wing this whole thing um you see people like that and you're just like w that's really really cool yeah and that's what you want too because then it it makes a more natural conversation with the audience with customers and users it doesn't feel pre pre-script pre-rehearsed and fake and so I thought also you worked at a golden age of Google photos you and David Lee because yeah that there was a lot of Google's going into the wilderness period of time all the different products we were launching but photos I was proud to use that product it worked out so well like the fact that I could s say brown dog and it would give me 50 photos of when my Daisy rest in peace uh when she was around was like fantastic and it felt very magical and then today and IO they launched now they're going to embed all your photos so you can ask like direct questions or something like whenever I been into a photo with an ice cream truck and it will be it will show you all these different things Co so that was really cool so um Fly Labs gets quiet Google you're now now in the machine um and you were telling me about how this researcher came up to you like hey we have this new capability here would you like to integrate it into what you're doing does that normally happen like just random researchers come up and say please look at my research try to incorporate it into your product yes that happened all the time and and sometimes I it was you know I'd have to say no because there it wasn't quite ready for prime time you know actually the same team it's probably y making the pitch I don't I don't remember but the the the same computer vision team had developed uh something on top of image segmentation I think I mean it's very related where you know they could take a video of um if if the camera is still and like somebody's kind of like running across this uh the the the you know the screen soort to speak then they could they would make one of these like strobe effect photographs that you know used to be popular in like sports magazines where you know you see somebody running and then they're at the next stage of their their run and so they would took a photo photograph that was like a sort of they would take a video and make a a strobe effect photo showing kind of the the action and they could similarly make you somebody's like swinging a tennis racket or or whatever so they had their their examples and as with as always with with AI when people yeah exactly that type of thing and you know as always with with AI like people tend to when they're presenting the AI and you know open Ai and Google iio are not going to be exceptions at all they cherry-pick examples where it works perfectly and so you know part of my job was you know conceiving like a way that this would work well in Google photos and I said like hey this would be pretty cool yeah strobe effect photograph you know there bunch of people do take action videos you know had some doubts about you know a lot of times people tend to like move the camera when somebody's actually going across the screen so wasn't clear whether it would actually work but like we tried it and we tried it on like my photos and you know I a baby at the time and it would make like strobe effect where it would make like a two-headed baby please tell me you SE have you seen Total Recall with the original arold was it like qu start the reactor save Mars yeah it's it's it was weird and it's like I can't even imagine what my wife like if she's just hanging out and she gets a notification from Google photos is like hey we made a photo for you oh your baby has two heads now it would be like you know traumatic to Mom's everywhere and there was another example that came up where like somebody had a you know like a dog that you know ran across sort of like ran across the screen you know like looked at the camera and like went on and again it was like a made a sort of like a a worm dog that had like a s a face sticking out the side of itself you know it's like listen this is cool technology and perhaps the day will come when we can make amazing strobe effect photographs and we could try to work out like sort of a set of heris sixs to avoid the failure cases which is what we sort of worked on with with with all all the all the um computer vision but that one just wasn't quite ready so I said like listen this is where I have to push back on you Tim and this is why I should have been the senior VP product I would have said this is not a bug it's a feature we now have a wiener dog converter for all your dogs Google can now turn all of your dogs into wiener dogs it can just extend the body but with a second yeah and a second pH yeah yeah no continue this is great go please please no so yeah it was common so I would I would get field uh field uh you know basically it it was this the equivalent of a sales call of an API startup it was a computer vision uh you know research team that would say hey we got this thing do you think this could find its way into Google photos that was actually my favorite part of of well I loved a lot of aspects of working at Google photos one of my favorite Parts was working with the researchers that were not within our team but with within sort of uh Google research and eventually I'd actually go to their you know their big Summit you know where they they gather all the like Global you know computer vision researchers and I would just ask them can you develop a certain technology because I would like for this technology to exist and that was super fun too I remember asking a team that was working on um like Optical flow whether they could develop an action recognition system that could recognize certain actions within videos like you know shooting a basketball or jumping into a pool because then you could basically take you know these long videos that's you know you recording your kid and they're hanging out at the edge of the pool and they're are you watching are you definitely watching you're watching right are you recording okay and then they jump into the pool you know there's a bunch of sort of like um footage there but there's one key moment where the child's jumping into the pool and if you could you know just extract those moments you can make highlight videos you can make little gifts and all sorts of things and I ended up working with that team very happily for you know a year or so uh developing action recognition eventually you know was you know published research and we we ended up launching the feature on Google photos so yeah this was this was common this was one of my roles within Google photos basically be the person productizing the AI the AI of the time that's that's that's amazing and that's that feature sounds like automatic ESPN highlight reels exactly and that that that's that's really cool um and also I always feel like a product manager at Google you had to have one foot rooted into like what is currently available we can do and another foot rooted into the future of like what is POS was possible and try to like join the two but then also make sure it's reliable when it hits production and I feel that uh I don't know if you heard about the rabbit R1 that orange device the large action model me and Joe I was hyped up about it I was like hey if I can get my cell phone just start doing stuff for me great but we're thinking like if it doesn't deliver on this reliably then Apple and Android are going to be like we'll Goble them up and that's exactly what happened so they didn't have the right product leadership to actually say hey you know what this isn't right for prime time we can't ship this right yeah I mean sometimes there's an element of risk where like it may work out it may not work out and some of us have a like an unhealthy addiction to that exact space where it's like you know there's technology that definitely works and you can just make stuff that's useful and you know probably make a lot of money if you just you know build things that people want with technology that totally definitely exist you know Works some of us just like can't help it's like we want to be in that area where like it may or may not work uh it is most exciting to be on the frontier but yeah you do ultimately pay the price if you make the wrong call you know you know it's to it's totally true I tried the uh you know this is not the rabbit but I tried the AI pin I was at a a networking event actually and somebody had the the Humane AI pin go for it let's hear it's it's exactly as good as everybody says yeah it's terrible yeah like $700 you mean you're going to pay me $700 to use right yeah know even then even then like you know for $700 I would wear it because it's actually a very spiffy looking bit of hardware but I wouldn't use it you'd have to pay me way more than that I mean it was crazy you know the guy is an engineer had it and he was showing it to me and you know what we were just trying to do I mean he was actually just trying to get me to like type in the password and it was like you know it was it was displaying numbers on my hand like and if I moved my hand back it would go up and it it would go and then I had to squeeze my hand to be like you know like 3 1 you know s two and it was just impossible cuz like as squeezing my hand my my hand would go a little closer and he's like oh yeah I think you changed the number right before you know you meant to squeeze on three you ended up squeezing on two and it was just like this is crazy it was such a such a bad experience and even after trying just to master that gesture for like five minutes it was still very much you know not good you know I don't know I I think the the what's the guy like MKBHD yeah yep he I think his review was was very accurate I I don't understand the people who are in a huff about that review like he didn't cause that problem that was perfectly reasonable frankly levelheaded review I mean you know right and and that guy's a really good dude he wants to see entrepreneur successful he's not like grinding an ax and also you just explained about just how the bar for launching consumer Hardware is so incredibly High yeah and if you can't stick that Landing you're good as good as toast and Google tried multiple times with Hardware they finally kind of got it with pixel to a degree but it's still iPhone still controls the market but they struggle with different Hardware products it was until I think Rick or Rick took over yeah and we started getting some discipline because we're so used to oh you know Google we're running 1 million experiments per day if the experiment doesn't work we just it out of production where Hardware it's like no you have the 18 24mon lead cycle and there's no like oh sorry we screwed this up let's do it again it's like no yes um so yeah uh and regarding the that uh MK I keep on screwing up his acronym CU it's so confusing mkb got for doing that it's like you're blaming the wrong person completely it's just ridiculous I I don't understand why he's getting even the slightest shred of hate it was quite frankly polite I mean he was you know it's weird to say it was polite because he said this might be the worst product I've ever reviewed I thought he was very polite about the fact that it probably was one of the worst products ever he could have said he could have came out and said this is uh dog crap and just like but he was trying to be charitable to it also and the same thing same with rabbit R1 um and I think the issue is that there's some people who just get so caught up in the potential of a product that they lose all sight of it and they just become True Believers and they just criticize anyone else because they've invested so much into the potentialist product if someone comes in and says no the emperor's working no clothes and they're going to blame the person who says that not the emperor for walking around naked yeah yeah yeah so no that that is interesting uh very good points on that also uh Vision Pro um I think for us us being a Google we played with cardboard um I had friends who had Oculus from from Facebook and I also had friends who worked for magic leap all really cool products and I think the video game use case for Oculus is great uh if I want to be in an X-Wing uh dropping off a a missile missile uh for the the missile into the shaft of the uh Death Star everyone probably everyone stopped watching right now because you think I'm a Star Wars poser now it's terrible that's a really cool use case but I'm not putting on these goggles all day long and wearing these things at work I'm just not going to do it and when you're telling me about the Humane pin and and you're trying to do this with your hand Vision Pro they're like oh yeah you just use your hands and you pinch and everything it's like no I I can speak clearly the chat GPT app and understands me I just want to talk I don't want to have to use my hands for anything um what are your thoughts on Vision Pro I haven't tried it so I I try not to judge things unless I've actually seen it and thought thought about it from first first principle so I'm not going to say anything good about about Apple Vision Pro um you know I agree people don't want to walk be walking around with with goggles um maybe it has some great use cases um I guess I'm I'm a little skeptical but you know it could be turn out to be like great for something that I'm not super plugged into like gaming I'm not a a gamer it's possible that it's just the best thing ever for for for games or maybe you know Zuck is right about you know the Facebook version of it um I have no opin hey we're looking for advertisers everyone loves advertising right yes from you I really want to do a Home Shopping Network like type of deal where I'm either selling samurai swords or your Enterprise product or your invention or something that I can Market without actually going to jail or anything so I've never been to jail and I don't want to go so anyways if you would like to Market your product to our community we have about almost we 4600 Subs 30,000 social media followers 2,000 folks on our newsletter well-connected group of stem employees if You' like to reach them then consider being Advertiser on our show and reach out to me at nfo SV investors club.com as INF nfo letter SV investors club.com thanks okay before uh we got cut off um we uh Tim was going to go for super teacher and what they do so Tim uh take it from there yeah sure um so I'll start with a bit about myself um you know I think we covered some of this already but uh you know my name is Tim and I'm a former teacher I'm also a second time founder sold my previous company to Google I was a product manager at Google photos and my role there was sort of specializing in in applied AI for consumers um you know researchers would come to me they have some cool new technology in AI computer vision and then I'd figure out how to productize that into Google photos in a way that was great that's actually my favorite part in a lot of ways of of working at Google photos was working with the researchers yeah I actually loved a lot of Google photos but uh but I loved working with the researchers and ended up switching over to uh to Google research where I was closer to researchers ended up managing a product called Google collab which is how uh developers often uh you know develop AI it's a developer tool that uh people who do AI sort of use all the time nice yeah uh I myself have a PhD in applied math from Cornell my dissertation a lot of people don't know this but it's actually it's well it's called algorithmic education Theory and it's basically proving models uh proving theorems about mathematical models of educational software I actually made educational software as a hobby that's how I first learned to code back when the iPhone SDK first came out in about 2008 and I actually created the course in iPhone app development for the um for the computer science department of Cornell I think we chatted about that earlier yeah all all of which to say is like I'm you know somewhat highly technical myself but I still kind of identify in a weird way uh as a teacher more than anything it was my first job out of college I Taun for four years in the New York City public school system uh I initially was uh teaching doing my student teaching and living uh Up In Harlem uh where actually generations of my family have gone to school or lived or uh taught uh I ended up teaching at ston high school which is New York City's best high school it's sort of like the the equivalent of of LOL Academy okay yeah and um had a great time teaching uh one of my colleagues there uh when from when I was uh teaching you know we started dating she's now my wife the mother my children um also my parents were teachers uh one of my grandparents was a teacher my my wife is still a teacher my daughter wants to be a teacher my brother's a teacher his wife's a teacher my wife's mom is a teacher so it's sort of like teachers all around and according to family lore my my great aunt was the longest serving teacher in the New York City public school um system history she taught for like 40 years but that was also a long time ago so maybe that's awesome maybe someone beat her record right now kudos to her that's like I just like the life of your family having a commitment to public service and I think that's really admirable so that's really awesome yeah I I mean I think we just really like teaching it's something like you know we would be talking about obviously the the lives of my you know my parents talking about their work and teaching but we actually like talk about pedagogy and like the right way to explain math and the right right way to deal with you know this type of student or that type of parent it was sort of just like always the the conversation and I'm not alone in this way there there are lots of families that are kind of like teaching families it's a thing all right so uh you know maybe with that said let me tell you a little bit of uh sort of foundational research in education this is something that people know intuitively when they think about it but they they they don't normally think about it uh here's the deal tutoring is known to be by far the most effective way to learn uh there's specifically this uh this thing that's known in the education research Community called blooms to Sigma problem basic idea is you take a a huge set of students who have a classroom instruction and then you give some of them tutoring on the side as well and the kids who have tutoring end up doing on average two standard deviations better than the kids who just had classroom instruction that's way way way better gets like the 10th percentile student to be like a 90th percentile student and that's the best intervention that's ever been found in education the second best one gets you like one standard deviation and and that's like having a great teacher with the you know deep knowledge and sort of all the best practices combined most interventions actually do close to nothing and a lot of interventions actually once you test them they actually do a little bit of harm tutor is the one thing that's just awesome like and and you know that like when a wealthy person um by the way are you still here cuz I can't hear you sorry I know I'm I'm just like uh saying if you can't if you can't improve silence don't say anything cuz I had questions were popping up one is why do you think tutoring at least such better results and then second direct intervention is that more of the teacher saying oh I know Johnny's failing so let me see if I can help him on the side a little bit but I don't have dedicated time to do it so intervention is a general term in education research basically anything you're testing to try to improve student outcomes so that could be anything from using one curriculum versus another like oh we're going to switch to using phonics or you going to switch to using whole word Lear or you know we're going to stop teaching the standard algorithm in math um or you know doing extra homework or using a bit of educational technology anything is just called an intervention and uh it's something that you're testing cool and and for tutoring do you think why it's so successful is because it's meeting the student where they are and finding your deficiencies and trying to like see where they're missing like where they where uh Missing the trying to fill the gap of where the lessons are not truly like congealing with the student mind okay it's a good question I actually think that the the main thing is simpler than simpler than that which is is that it's sort of intense instruction directly to you so when you're in a classroom you might be sort of like thinking slower than the rest of the class you're a little bit behind you might be thinking faster than the rest of the class the teacher likely has to sort of ask a bunch of questions to make sure you know few people have a chance to to you know engage with the Mater material but when it's one-on-one it's all directly to you all the questions to make sure that you're engaged and following along are directly to you it's not to one of your classmates and so it's just much more focused it's just a much better use of time then also it can be actually adaptive to you so if you're you know R running up ahead or you have a particular sort of like cognitive you know misunderstanding of the material then the tutor can adapt but I actually think that the adaptivity well super important and sort of that's how everybody sort of thinks about it's a kind of charming when you when you see it happening you feel really great when you're a tutor and you know you end up adapting to to to their kid I I actually think that the main reason is better is just because it's like it's private it's it's it's uh intensely focused on just that one student you can ask stupid questions and also there's a discipline of you have to show up and do the work in front of someone and it's kind of like a a a gym buddy to a degree too that's right you can't just like you know phase out get distracted you know it's your learning during that time and it's it's very intense nice continue this is great presentation yeah um so you know this is great uh tutoring is great and when a wealthy person's uh you know child is struggling in school like they don't even think twice about it they get a tutor like that they don't even you know it's just immediately they're going to get a tutor or even if their kid is like so great they want to get them ahead they want you know they're type a parent or whatever they will also get their tutor child to tutor they just know that the tutoring is where it's at if you want your child to to to thrive as much as possible and um that's great for those people I guess but uh the problem is most people can't afford it that's that's why it's the problem that's why it's Bloom's two Sigma problem like tutoring is not the problem the problem is tutoring is expensive because the tutors are paid by the hour and most people can't afford to pay for somebody else's time by the hour for enough hours that it takes to sort of uh you know have the effect and that's kind of like the the macroeconomic you know sort of uh you know view of it is like not everybody can have a a private tutor cuz like you know the uh the the like labor economy can't be organized in that way um but from a from a consumer standpoint it you know it's felt you know without all the statistics it's just it's just like you're in a painful situation it's just like okay my child's struggling I know what they need so they need somebody to be doing this one-on-one with them it's very obvious when your child can't do it it's like I wish somebody was just explaining them oneon-one maybe you try to do a little of it yourself but you know it's it's takes a huge amount of time and so you know the right thing to do is get a tutor maybe you also don't know how to teach a kid how to read or you know what the new math is all about so you want to get a tutor you look into it and the very cheapest ones are like $20 for a half hour session and um you know once a week that's $1,000 dollar that most families can't afford you know the families typically have stretched budgets in the first place but also any tutor who's actually good ends up spreading by uh By Word of Mouth and quickly uh having clients for like you know being able to charge 40 60 80 even $100 an hour or more that's just completely unaffordable for the vast majority of people and so you know that's why I started super teer super teer the whole idea is that you get unlimited tutoring for $10 a month and it's unlimited tutoring in reading math science art poetry social emotional skills geography Greek mythology you know unlimited tutoring in any subject for all your children combined for $10 a month so you know pretty much anybody can afford that there's also an annual plan you can get it even cheaper nice and uh you know when I tell people you know if I'm at a sort of one of those networking events and people say oh you know what do you do or they you know they see my my swag they say what's super teacher I say well we offer unlimited tutoring for $10 a month and you know I pretty much have to follow up pretty quickly with you know just to be clear this is not something we're thinking about building this is like an actual product that works you know you can get it in the App Store it's available on iPhone and Android um it's free to download it's free to try the first lesson in every course is free and already it's done over 100,000 hours of tutoring with real children over 10,000 families have now upgraded to the paid version uh it's done actually I think closer to two million questions answered and um you know this is this is what it looks like I you you want to uh see what see what the actual product looks like so here goes Michelle can you read this sentence the chair is blue so close let's try it together the chairs are blue blue a chairs your turn the chairs are blue Michelle that was great reading perfect that's actually one of my daughters and that's one of our ads One Stop Dancing and get back here Lily where is the two supposed to go under the five and the Seven what was so funny the two is dancing on the side of the one's colum like this oh my God that's too adorable yeah those are my two kids and those are two of the first ads that we made and they they sort of uh they show how the product works pretty well uh and you know you can kind of see it there so it's it's conversational the kids are speaking out loud back and forth to the teacher uh it's multimodal in the sense it's highly visual there's also sort of like audio things you know you know good good job little little sound effects you if you see the full app it's got you know Awards and there's you know flashcard sessions and there's worksheets you can print out and it's you know it's a very robust robust product that uh that can actually teach kids how to read how to do math I mean it's really crazy to see um my kids have learn how to read learn how my younger daughter learned how to add and subtract from super teacher my older one had already learned sort of the basics but she learned you know carry the one in subtraction with borrowing uh she learned how to read an analog clock you know entirely from super teacher now she's learning multiplication and division and fractions all just from super teacher so it's it's pretty it's pretty cool um maybe I'll pause here maybe you have you have questions or this is this is fantastic um I it reminds me of one laptop per child where the dream was in the early 2000s just give a laptop to a child and it's going to start learning and then here's your daughter is actually realizing that now with super teacher and that's really neat so yeah congratulations on getting this all built out I think it looks really good um thanks uh I know you can't go I mean it's um assuming here my question putting words in your mouth can't go into the secret sauce but like kind of like how does the tech stack work for this thing is it like what what's powering it is it uh like did you hardcode all these examples like how how does all this stuff work yeah good question so you know maybe I can give you a a high level uh you know overview of it we started out we wanted to make an automated tutoring system and everything we've done has been simply to make you know make an AI tutor I'd say that uh the at the base of it is a sort of a protocol a lesson protocol that sort of defines an Adaptive delightful lesson so you can kind of think of the lesson protocol as like a a lesson plan but on steroids where it includes everything that the teacher might say or you know instructions for you know how to how to construction construct it in real time you know including like the child's first name and you know that type of thing uh every award that might be granted every little sound effect every video every gift every dancing too you know every question the teacher will ask along the way the the lesson protocol that we developed is encodes an Adaptive delightful lesson that will be effective in teaching children and also delightful and you sort of keep keep them coming back that's kind of like the bottom of the stack on top of that is our what would I call sort of a um lesson execution engine it's basically a virtual teacher in the cloud and it's basically a server where you have you give it two inputs one is a lesson that conforms to the lesson protocol and two is a um a uh a connection a Communications connection with a real child somewhere you know uh with the microphone and the lesson execution service it's like teach this lesson to that child and then it goes it's just a teacher like a teacher service that will take a lesson any lesson you give it conforming to the protocol and in a child on the end of a communication Channel and like teach that lesson to that child and obviously there's a lot to be said in terms of like how we built that you could say like a lot of the sauce is is within that and there's a lot of AI involved there's you know speech recognition that's based on nurl Nets there's natural language processing there's generative AI for the you know the voice of the teacher uh and as as new models come out we basically incorporate new models it's sort of the Google photos Playbook where you know we created a product that is almost designed to benefit from all the advances in AI that we saw coming you know I was at Google research I could see there's a huge huge amount of advances coming in speech recognition in general and for children and with different languages and different accents and all that natural language processing was uh growing like Gang Busters you know there's you know thousands of models on on hugging face generative AI was starting to take form synthetic voices that are sounding more realistic more emotive and similarly that can speech speak most multiple languages and have speak with accents and dialects and stuff like that so there's a lot of AI in our lesson execution engine the sort of virtual teacher in the in the cloud that that's uh importantly that's scales right on top of that it we've built a lesson authoring environment which is where human lesson creators we call them lesson Engineers that we have on staff create the lessons and you can think of the lesson authoring environment as sort of like a uh a no code platform specifically for building conversational multimodal lessons and so we have a staff of creators of these lesson Engineers that basically craft these lessons until they're you know complete and you know uh you know conforming with the protocol and we do use AI there too you know we've built a sort of a co-pilot for the lesson authoring environment where a Creator can just kind of click a button and then a chunk of lesson gets uh gets written including including like the question that the teacher will ask and the the answers that they expect the kids to give and the responses that the teacher will will respond with based on the answer that the kid gave to the question that's fantastic I I like how this idea of like it's like a you're using AI but you have hum in the loop and you have the humans as the rails to be like okay AI right now can't do it can't long-term blinding it struggles with but humans are fantastic with that yeah one of the reasons you have to do that and you know despite the uh the buzz around you know like fully uh you know right fully modelbased AI tutors is that first of all parents don't want that parents moms don't want their children their five six-year-olds talking to whatever came out of GPT 40 or whatever the next the next thing is yeah it's just just doesn't feel safe enough and also it actually will go off the rails we we every single hackathon that we have you know about once a quarter will try the latest stuff from open aai from from Google research from meta uh you know you know Facebook AI research and uh typically they start going off the rails right away typically because speech recognition doesn't work well enough uh we have a very a very mature speech recognition system where we have using data from real users from you know the 10,000 children and hundreds of thousands of hours of tutoring we have fine-tuned our speech recognition system on a question per question basis and so it sort of can far exceed the you know the capabilities even of the you know the latest version of whisper that's fantastic so the more more students that are using this the better it gets and then not like you create two products there's a product that your your awesome beta tester student jogers are using and then there's a product the back end for the Educators no the lesson authoring environment which is where our our human creators you know craft these lessons with the ative AI and the lesson execution engine that uses the I as well uh but then out of that come our actual lessons and then we have an app uh maybe have a slide with uh sort of showing next slide yeah uh here we go yeah and this app is where it's basically a thin kind of Thin Client as we saying you know the tech industry it's just it's just a sort of a lightweight client that just allows parents and kids to browse the lessons pick one and then start it and once they start it it's like they're having like a you know a zoom session with our virtual teacher in the class this look super fun so one thing I was talking about uh so we call We Now call GPT 40 GPT 40 oce and we know I'm very you know we're going to go uh we're going to go uh what Edward 40 hands ex thank you we're going to pour pour one out for gbd4 now we have 40 o here uh when the 4 came out everyone was like what does this mean for dingo everyone's going to like use this use this now it's like no actually I look at the the translation aspect of 40 ounce as a way of like I'm G to pinch I need someone to translate Japanese so I can get something done but then for I want to learn the Japanese language I'm going to do lingo because I don't want to be dependent upon that cell now things can happen in the future but what I was going to say also is what I love about dualingo is it's character it's UI it's very fun it's engaging it's distinct and I look at this design we have right here and this is very fun and I it looks really cool and engaging who who designed your UI was it this your view or was it like you got a great designer we have a ux designer he's fantastic yeah and he came from a actually came from an animation background a computer animation background he also codes he's highly technical believe it or not he had worked as an animator and as a sort of a creative producer um for his whole career and he originally um we contracted him to illustrate the teacher and he had a bunch of ideas about making lessons all right well why don't you make a lesson he made a lesson it was great and then he was like oh you know there's this thing in the lesson authoring environment I feel like it'd be a little bit more clear if you didn't have to click this thing twice as like hey you know what you have like a natural eye for ux and so we started talking and basically doing a uh you know like a trial period so like you know what if you changed your career became a ux designer and it was a huge career leap for him uh but it's worked out great that's fantastic it's another side point of when tell people like I have a friend right now who works at Google he was talking to me about taking a job open AI I talked him through my first conversation was like just shut up drop your stuff and go work there yeah like well you know I'm might I'm get down leveled and the scope of my role changes I'm like you understand startups they run at the speed of light what you're doing today will change in 90 days to 180 days of 6 months and you were just a great example of that Designer who never would get the same opportunity at Google he would had an interview you know that's sort of like wrapping up I guess you you had a lot lot of great questions about um about you know GPT 40 40 as you're calling it 40 ounce you know one thing that's funny first of all I completely agree that dual lingo is a is a crafted product and this is not going to uh you know replace du a lingo I see these people on Twitter saying like it's so over for teachers like this is going to replace a teacher Insanity teachers are irreplaceable nothing will ever replace a teacher um and it's certainly not that you need a full crafted product for what what parents want parent wants want the whole thing they want first of all the teacher to actually have an opinion about what lessons should you do as you're learning to read you know GPT 40 isn't going to um you know just like teach a child to read over the course of you know months and lesson after lesson it's it's just not there yet um you also need you need the rest you need the worksheets you need the flashcards you need the extrinsic motivation I mean you could probably torture it to to do some of that but the thing that the thing that the thing that was true about Google photos and it's also true about super teacher is that we're need to incorporate all the advances in AI so as new speech recognition as new voice you know generative AI for for voices as Dolly and llms and GPT you know 3.5 and four came out we've Incorporated all of those in various ways you know for the llms it's just in our lesson authoring environment because that's actually the right thing for that GPT 40 is all similarly going to be slotted into Super teer and it's going to probably enable new types of modalities within the AI tutor we have to Fig figure out a way to do it safely first of all we also have to H have it make sure it's uh reliable and doesn't go off the rails and doesn't get things wrong and doesn't tell a kid that they're wrong when they're right you know if a kid you ask a kid you know what's five Time 5 and they say 25 and you say the teacher says actually no it's 25 the kid gets Furious the mom gets furious at the product like it's over you can't do that even once with a child so if we figure out a way and I think we will to to use GPT 40 in a way that actually serves a serves a full polished product for for parents and for children uh we'll incorporate that too and we'll sort of uh you know it it might enable more open-ended interactions with our teacher we might have like a TA that's powered by G GPT 40 opens up a lot of exciting possibilities so it's just like one after another when there's an announcement in AI people are like oh no what's that going to mean for blah BL blah but if for products that have been founded and crafted with the strategy of Simply incorporating all the latest and Advance advances in AI as they come out it's exciting like the more the merrier so we watch the the uh the announcement from open Ai and from Google IO with you know great excitement each time because we know super teacher is going to be getting even more Super even more powerful fantastic I I love that um you you hit home with a lot of great points about education and the holistic approach to it it's not just the technology there's so much else going on there um now I was thinking we have a lot of parents watching so is there anything like try something that we could give them as an incentive to try using your product sure so first of all you know the website is getup teer.com and or you can just search the app store or the Play Store for super teer it's free to download free to try the first lesson in every course is free if you choose to upgrade to get the full version use the promo code success and that will get you 20% off sweet well hey uh Tim thanks for being awesome awesome person a good friend I'm just so happy to see how great you're doing a super teacher looks fantastic and this was my this was my first podcast ever and so thank you for thank you for uh for hosting me now now I feel like I should go do more podcasts dude you're a natural I want you back thank you all for watching don't forget to like subscribe and share and we'll talk to you later peace
