Kevin Systrom on the Future of Artificial Intelligence, Art and Artifact | TechCrunch Disrupt 2023

[Music] [Applause] foreign [Music] may I have your attention please welcome TechCrunch general manager and editor-in-chief Connie loisis that voice of God thing always freaks me out a little bit thank you so much for coming day three we're so excited uh you know thank you all for coming these last few days it's been such an exciting wonderful big event I've loved every minute of it I hope you have too we have heard from Kyle vote Dario amode Lisa Jackson of Apple um it's been amazing and we've got a lot of great speakers coming up right after I get out of here we're going to see Kevin Systrom the co-founder of Instagram who's gonna be telling us about his new company artifact uh we have Gary Marcus who is a sort of pioneering voice in AI we have Meredith Whitaker who is the president of Sigma she is really forceful I'm really excited to see her speak she also nicely spoke on our AI stage just a little while ago and I'm getting to talk to Katie Hahn of Han Ventures where I haven't had a chance to sit down with in a couple of years so I'm really excited about that um anyway we also have of course the last day of the battlefield we are down to six contestants they are awesome companies and I'm really excited to say that half of them happen to have uh include female Founders which is great um and we have an All-Star cast of Judges uh joining us which we're very appreciative of Marissa Mayer Dana Suttle Mar herschensen Mahmoud Hamid and Charles Hudson so uh that's it uh let's see no that's not it we also have um our deal uh our deal making room next door if you have the right Badge go over there that's been on fire I guess deals are actually being struck there so uh don't miss that um now I think that is actually it so I'm going to leave thank you again for coming and let's welcome Daryl atherington our managing editor may I have your attention please welcome TechCrunch managing editor Daryl Etherington [Music] yeah all right final day of disrupt the best day the most exciting day for all of you here as Connie mentioned the deal flow Cafe has been popping I went in there and someone approached me for a deal and I was like I don't have any money so hopefully if you have money go over there uh by the way don't worry about taking any notes during the event we've got otter it's right there as the words come out of my mouth that go on that thing they also go on the app and they go on the web so you can get access to those for free you can see all of our past uh panels and all the ones today too and you can also check the app and the signage around the building for our other industry stages we've got great programming programming going on on those two for the first time ever so many industry stages today we have hardware and security and those are going on to my right over here we have pitch showcase breakouts and round tables on the second floor so check out all of that and let's get ready for an action-packed day which starts right here uh with me I'm gonna be sitting down here and joining me will be Kevin Systrom who is co-founder of Instagram of course and also co-founder and CEO of artifact a new type of news app that I've been using personally quite a bit uh and that has been updating at a really aggressive pace so clearly they're not so searching over there but we'll hear more about his plans for that and his views on AI right now so let's welcome out Kevin systrow [Applause] [Music] all right Kevin thanks for joining us thanks for having me you you've been to disrupt before but it's been I have many many years many many things have happened since then I think right uh so I want to talk about AI to start just because it's it's been a theme generally with the tech industry I'm curious where do you fall on the AI dumerism Spectrum are you uh it's coming for us all are you a no problem are you somewhere in between um maybe the answer is Nuance so uh even if it was coming for us all what are you going to do about it we make the most of it um my sense is when new technologies arise people always look at these new technologies and say to themselves jobs are going to go away it's going to kill this these are the problems with it and if you just like look back in history at every major Revolution whether it was the internet the PC the mobile phone like imagine how many things have been unlocked in our lives because of those yeah sure you know jobs have gone away sure uh you know classroom teaching and education has changed where you know you used to trust that people wrote their own papers now they don't potentially we just adapt we just we find ways to adapt and then all of a sudden um you find that new jobs exist people are super powered because of these Technologies and I think that's much more likely to happen you know like I was talking to people and they're saying oh like illustrators are out of a job and it's like I don't think so I think what will happen is the really talented illustrators will learn to use these new tools and then Super produce right that's my sense uh but you know it comes back to my first comment which is I don't know even if everything's going to be terrible you might as well put a positive spin on it that's true yeah so I'd like does that extend to like do you think that they will become sentient and sort of block out the Sun or anything like that or is that a possibility no I so I spend most of my day programming these days yeah yeah staring at neural networks and staring at why either neural networks are working or not working trying to train them um and whenever someone says you know there's a network that's sentient you're kind of like just like a bunch of numbers but and then that person will respond well you know we're effectively just a bunch of chemical connections so and then that gets my mind racing but um but generally speaking I think uh we are we are likely very far away from these doom and pending events um and we are very much in control of how we build these Technologies how we regulate them um and very smart people are in charge so I think it's good to be asking these questions but I think the reality is the next few years are are going to be completely safe yeah great okay I feel reassured uh so you mentioned that you know as I'm coding so like let's talk about that let's talk about um what is what is your day-to-day what do you as I mentioned I think you've been really really aggressive with the pace at which you're releasing updates for artifacts because just released your links feature last week I think so yeah tell me more about that what do you do all day what are you excited about and what are you building right now for artifact I think the best way to answer this question is to actually say like why start another company at all absolutely a lot of people if you start a company it gets really big that sets a high water mark and I mean let's be honest how do you surpass an Instagram that's you know yeah given that it's growing constantly it's very hard to see where anyone could ever do that um you start something because you believe there's a problem to be solved in the world you start something because you love building you start something because you're pained that a problem exists in the world and you need to solve it and for me there there's this framework that I I feel like someone in college taught me it's these three circles and the one circle is what are you passionate about and the second circle is what's a big problem in the world and the third circle is will someone pay you to do it and uh and I looked at these three circles and I said to myself well one you know at Instagram I feel like I saw you know if if the wave of AI is crashing right now we saw the swell and you could look out five to ten years and say okay this is going to be a big deal I don't know exactly how it's going to work but specifically in social networking how AI dictates what you see and it's not just about who you follow it's about what your interests are that was really exciting to me and I think we saw that in Tick Tock absolutely uh so that was the passion what a big Pro what what's a big problem in the world I mean information consumption news consumption specifically at at the peak is an enormous problem today what's true why is someone sharing this uh should I believe it like those are the types of questions we ask every single day when consuming news and by the way is this just being produced to basically game the algorithm on something like Facebook so that it gets you know a lot of clicks or does it produce interesting useful information for me and then on the publisher side I mean you know and obviously you know much more about this than I do but maybe not on the publisher side it's an existential crisis of you know is it ad revenue is it subscription model hey like you see all this fracturing of Publishers where independent Publishers or writers will break off and start their own newsletter thanks to something like sub stack and they're actually being successful doing that it's not everyone but some people are yeah um so where do those people get discovered how do those people get distribution that's a big problem and then the third circle I'm going to avoid for a while which is how do people you know will someone pay you to solve this um but it worked out with Instagram so here's here's crossing my fingers yeah yeah I mean those are huge problems to contend with and you're right like we think about them a lot uh every day too and and I think an interesting thing you bring up there is you know that that a lot of time and effort and money was spent on figuring out how to game some these algorithms like Facebooks and to some extent Instagrams a little less so for news I think but um and yeah that maybe LED people down a path that was not in service of dissemination of truthful information let's say right uh did you see this recent post by Mr Beast no about like whether or not like open mouth versus closed mouth oh yes thumbnails performed better yeah like this is like this is like the best example of optimizing for the algorithm I mean Mr Beast obviously very very successful but we're now at the level where people are questioning whether their mouths should be open or closed in a thumbnail to get distribution so if you're doing that for for for a video imagine you're sitting there as an independent writer at a publication trying to make the decision do I write about this thing don't I write about this thing What's the title of this thing of course people are thinking about how to get distribution yeah I just bring this up because it's it's so clearly in every corner of our consumption including YouTube yeah absolutely so but one thing that you've done um at artifact is you've sort of like said well let's take that decision away from perhaps you right the writer not take it away but let's make a new decision on top of that like let's change the headlines uh like perhaps avoid the more clickbaity headlines for the worst actors right and I it's if it's been controversial right it's been met with some like well why are you the ones right are able to do that right and what's the thinking there and what's the service to readers so what do you say to the kind of that criticism of you shouldn't repurpose this if your purpose is to sit at this distribution later right well for context for the people that don't fully understand the future um go download artifact um and uh and you'll see that sometimes by the way the large majority large large majority of Articles have no editing whatsoever yeah um but we kind of see ourselves as a list of links and I think it is arbitrary what those links say if you go to Tech meme Tech meme rewrites headlines for clarity to make sure that people understand what's happening and if you click through often that title is different than what you see on the page so it's a practice that's happened forever which is headline rewriting yep for clarity um what we decided to do was two things one is we farmed out the decision which is if enough people um mark an article as a click bait title if enough people then it triggers some internal logic to go off to one of these large language models read the description of the article and the title of the article and rewrite it in as factual a way as possible it was kind of a neat idea we had late at night when we were working on something and it worked really well we've done that and and users are enormously happy when they come across an article where the headline has been Rewritten just for clarity for conciseness the worst is to come across an article that says you know this amazing restaurant just closed down and it's got kind of like a shady photo of the entrance you don't know where it is and it's like I wish someone would just tell me what restaurants so I knew if I cared to click on this thing or not um so we're trying to do is just level the playing field a bit uh the other reason we do this is because we use clicks as a really important signal of what people are actually interested in if you purely use clicks as as the main signal what you get is what are people interested in interested in knowing about rather than what was valuable right so it's important to kind of like level the playing field of titles to make sure that things aren't um click-through rate is in isn't somehow exaggerated by a clickbait headline which it often is right but so like what is going on there with incentives and how would you advise like I think of a really interesting thing that you can probably share is what you've learned about readership about users to media and Publications right so we have done this I'm including us I'm sure we've done this before and probably still on occasion right like what should we be thinking about our audiences because that to us is signal like we want them to come and we need them to come and click on the thing and that's a reason behind doing that right but what are we getting wrong about that signal that we're rewarding it in this way that is not necessarily productive or best for audience I think maybe a few anecdotes so um the first is when we started artifact we went to all these writers and we said will give you stats about every one of your articles because you don't know click-through rate generally because your stuff's being shared on Twitter and Facebook but we'll tell you of the people that came across your article in a feed how many of them actually clicked in and to my surprise every single writer was like no no I don't want to know because that's gonna that's gonna change what I I will naturally want to publish on yes and it's not surprising but like you know most of the most the articles on on artifact that get the most click-through rate like there's a word that rhymes with Elon in the title and um and that like you know and it's like okay if you want to get you know clicks you can you can publish on all of elon's antics and and you'll get a lot of clicks people are obsessed um it's it's the it's like the first modern you know True Tech celebrity and um and that's cool but uh but I think writers want to be a bit more measured in what they cover and independent I think that's a good thing sure um absolutely so I'm not sure there should be that feedback right which is you know and then the second thing is I I've been so surprised on artifact um the writers you think would do well do okay you know the famous ones the ones who do really well you probably don't even know their names right and that's one of the most exciting things about artifact right now is like I actually think there's an opportunity for more and more of these writers to be discovered as independent personalities um because they are thoughtful because they do get the Scoops because they there are like their take on a new device or a new technology is so much more interesting than maybe one editor at a very large publication thinks um and they're able to get the distribution via artifact maybe because they weren't placed on the home page or they didn't get the front cover of the tech section or whatever yeah um so that's been really fascinating it's just like what what happens if there's a meritocracy among all these writers and what happens if you expand Beyond writers to anyone who picks up I guess not a pen but a keyboard and decides to write on a sub stack or a medium and throw their Link in links sorry I'm talking a lot but the last thing I asked about was links yeah um we decided that okay we have this list of Publishers we try to crawl and you know and we distribute their links and hopefully they get traffic and clicks all the nice stuff that they like um but what was failing a little bit in our feed was just this texture of the unknown you know this texture of like this really interesting link that you wouldn't have seen otherwise that you might get on something like Hacker News but maybe it's not technical yeah and so we launched this links tab which is effectively a giant bucket of content that anyone can throw stuff into and say hey figure out if this is interesting so the algorithm that I worked on basically distributes it tests it and if it's interesting and seems to resonate with people at distribution more and more in bubbles and it's been working really well over the last week so this is a weak old feature but I'm really encouraged I mean I'm seeing new apps to download I'm seeing interesting recipes yeah I'm seeing interesting GitHub repos I didn't know existed it's very Tech Centric of course because it's personalized to me my feed um but that has been an interesting unlock for artifact and and I'm really bullish on where that's going well that I mean it's a little bit retro right like it kind of sounds like delicious or dig or something like that like from back in the early internet days I guess dating ourselves here but uh well my sense is every great idea is is an evolution of an idea that has happened before yeah numerous times you just have to modernize and update it so you know delicious was great dig was great um many of them were either I guess sold to other companies Yahoo included um and uh and their Founders who had the vision went away and it just kind of sat back there nothing I mean the number of people that told me I was crazy to start a photo sharing site when Instagram started they're like ah photo sharing sites haven't those been done a thousand times and it's true but what was different was okay we were going to add creative tools we were going to be mobile only we're going to be focused on iPhones so if you change enough of those bits yeah you can take an idea from the past I mean come on Tesla like it's a car yes like change some interesting things about it that changed the game and then be good at it and it works really well and is the the change for you then is this ML and the algorithm that you've built is that basically what I think I think the difference is is um ml first so uh definitely using large language models and all the tech that exists now to both understand text and personalize that's one the second is being really focused on mobile I think links to things like YouTubes and tick tocks and Instagrams and such um work really well on the platform um and and third is just like I think the craft like if if I open up other news aggregators today and again I I think we now consider ourselves kind of like a a it's like a post news aggregator right like you're you're trying to go beyond just news and and do other content news aggregators you know there was an article about uh a news aggregator launching um like a like a calm tab the other day and it's like if you tap on it it's like just positive stories and the story talked about this feature in relation to the fact that their Home tab was just like murders traffic accidents you know bad politics stories and it was like oh thank God they now have this feature and I was thinking to myself like or you could just produce a product that doesn't you know push people to terrible awful things all the time um so my point here is I think there's a level of quality or at least a level of Integrity that we're trying to shoot for editorially speaking to make the experience much more than you know just plain accidents and stuff that obviously draw a lot of eyeballs and click but again back to that signal of clicks if you focus too much on it you get you're equating clicking with value and that's just not true yeah we all know that yeah you get yourself into a lot of trouble there but that I mean the thing I was going to share earlier when you were talking about that is like we we have uh very uh diversities room and a lot of different perspectives in terms of how you like achieve success right and and we've had people like myself I've never ever looked at the numbers like hey look at the numbers I look at the numbers now because that's my job at not writing as much but um uh but then we have writers who are obsessed with their analytics and always looking for analytics right and you wouldn't necessarily that wouldn't result necessarily in the content that you expect it might result in right but it is uh it's very personal to writers but it's interesting to see use like have that and provide that but then also wrestle with like how do we serve that to that audience in a way that is productive rather than just getting everybody to write about Elon all the time which is kind of what the thing would tell you right no it's the biggest downside to this recommendation driven social network phase that we're in yeah is that we tell people the rules of the game which is if you can have a high click-through rate if you can have a Lara long watch through rate or repeat rate on something like Tick Tock you win you become rich and famous you become Bella porch right like and what happens is someone decides ah I know what I'll do I'll take an old plane I'll fly it up real high over the desert where hopefully no one will get hurt I'll pretend like the engine failed and then I'll parachute out meanwhile I've recorded the entire thing with GoPros on me and on the wings and it was it was an engine failure right and then I posted to YouTube because that's how I win yep and like in what world are we better off because that is the outcome it's entertaining yeah you know it's it's almost like a Hollywood production but is it unsafe yeah do you want to encourage people to do it no um but again back to the like mouth open mouth closed things I mean we're we as technologists are responsible because we are telling people the rules of the game we're saying do these things and you win yeah and sometimes you win big it's not just like oh great I got a few views on artifact it's like you know people are here on Tick Tock yeah it's their career yeah so I you know I feel a little bit like dated at this point because I've been through at least one cycle of social media with Instagram but like sometimes that experience that wisdom allows you to look at a system and say like are we um are we incentivizing the right Behavior absolutely and I'm not saying that we've fixed this on artifact we definitely get clicks on clickbaity stuff up at the top and we have to manage it um but it's interesting because I don't think enough leaders at these companies are focused on setting up the rules of the game to provide valuable content in the long run rather than just people jumping out of airplanes so well let's talk about that a bit then because like we've talked about Elon I would love to hear your thoughts about Elon and what he's incentivizing over at X right now and how that has come together do you have thoughts about that um no it's just like what was the uh I'm gonna get this wrong at Amazon they had something called like the chaos monkey that would like go around and fail systems randomly uh erase meetings from you something like that right yeah um I kind of wonder if elon's doing that to Twitter if it's he's just like we're gonna make this thing really what is it like anti-fragile yes just like if you break enough things at this thing it'll become a super being hardened the remaining core will be hardened yeah um but I don't know it's like never count Elon out it's like you just batting average is pretty great so um I guess we'll see you know would my instinct be to charge people would my instinct be to bring back a bunch of these accounts that have violated rules would my instinct be to do any of these things no but maybe that's why maybe that's why it'll save Twitter because all of everyone's Natural Instincts someone I really respect this guy Ray dalio is an investor you know thinker um he makes this point which is not his point it's just a well-known point in investing like to make lots of money you need to be contrarian and right you need to take the opposite bet and be right starting Instagram was the opposite bet because everyone thought there were no there was no room for social networks but we were right and you know I think we can all agree that many of elon's tactics currently are contrarian yeah and the question is will he be right right that's the fascinating thing to watch and and I can't tell you whether he'll be right or not I think we just have to let the game play out okay but I mean like on the other side it just seems like he's also anti-Semitism like whatever like what about that side of things yeah sorry to be clear I'm not talking about his personal views I'm talking about his management style at the company laying off a bunch of people um taking bets on charging for membership um I was thinking about this on the right here actually how you know because I assume we wanted to talk about threads at some point given that it was we definitely do um so I'll just go there um like I wonder you know because I assume you you'd be curious like do you think threads will win do you think Twitter or sorry X will win um and I wonder if there's like a third outcome which is like a very Silicon Valley way to think about this right but um a third outcome where it's almost like you know how when uh uh uh you know you have you have a presidential hopeful and people tried to discourage the second you know uh Place presidential hopeful not to run in that party because they're worried they're going to take away votes from their first hannibalize the vote right exactly like I wonder if that happens here where what you end up with is you know the original Network for distribution of text and ideas so weakened that it allows for something like threads to actually do pretty well but actually never be as big as Twitter could have been if there was just one right and what you end up with is this like weird fracturing of the network where it's everyone's worse off because now there are two or three or four players all trying to do the same thing by the way I think this is kind of true with Tick Tock and reels and and shorts and um where where the lack of network effects lack of density leads to they're not being a great Network for discovering ideas and real-time information and whether the world is actually worse off I know this is like a very strange way to think about this but um I think what will likely end up happening is all the distraction of that competition will open up some white space for a brand new Social Network that just takes a very different View and I'm not claiming artifact is this I'm just saying someone a couple kids in a garage will start something that just completely crushes both of those things because everyone's so focused on competing for previous ideas in the past well you're saying both but I mean it includes you know Mastodon or blue sky or those things too which are all doing that dilution effect that you're talking yeah but and I guess I'm bearish on all of these because they're rehashing of of the same idea and I'm just maybe I'm a little bit of an optimist here but like I don't believe we're out of ideas for networks I don't believe they're done iterating and evolving and I believe there are a lot of really smart people in Silicon Valley willing to try new ideas and something is going to stick it's just I kind of you know I've got my popcorn and I'm looking at you know all of these folks competing for what feels like almost like a you know a very two thousand yeah six last decade yeah that's fine but um meanwhile Tick Tock is kind of running away with the Baton on short form video um and and they're iterating very quickly so I I guess we'll see but um yeah I think there's a third third world where all of this competition actually leads to Consumers being like weirdly worse off because people are confused where to post things and find things yeah um those were like that idea requires like a certain critical math yeah right which it just seems like they're fraying well this is why um at least in in consumer winners win big yeah because there is that critical mass and then you know someone once told me I like hurt a little bit but someone once told me we were like three years into Instagram they were like you guys could just leave now and this thing will reach a billion users and no that wasn't Mark um come on guys no no this was an investor um and it's not entirely true um but things do reach a a certain critical mass and then they just start going and it doesn't matter what features you add or you know they just kind of take in consumers specifically true yeah yeah uh because you were on a market what do you think about uh meta and their stewardship of Instagram what do you think about where the product's gone where it's going I mean you know the worst thing in the world would be to wake up as a as a founder of something and the company has crushed it it's gone to zero they've pulled the plug how many entrepreneurs do you know where that's been the case tons tons yeah tons um the fact that Instagram is still crushing it still winning uh they're iterating you know all the decisions that they make would I have made exactly the same ones of course not but does that matter um I'm proud of where they are it's super cool and uh it's so funny like I'll go to a party and people are like man I love threads congratulations I'm like I've been gone since 2018. they're like what this is the great thing about being a Founder everyone thinks you're still work there yeah yeah but then the problem is you're like I swear artifact is a thing go down uh so then back to artifact uh thank you yes yeah I would love to appreciate that give you more opportunity to talk about that but we've got links that has shown kind of like some hint of where you would hope to expand right because you've said I think repeatedly that news was just the beginning um but is that kind of your focus for the next little while or where do you think you go from here what is this I'll ask the next later but where do you go from here are you are you a big Lord of the Rings fan sure yes yeah you know like I'm gonna get this wrong oh my God I just mentioned Lord of the Rings and everyone loves Lord of the Rings here so um there's like that Montage okay let's go to the movie I know now I'm really screwing myself you're losing uh I'm losing everyone um but there's that Montage where like they draw out the map and they show you all the worlds and like where their head I kind of think entrepreneurship is kind of like that where you know where you're headed you know there's a land out there but the map's not drawn and you're way over here and your job is to connect the two yeah not to know each and every step along the way and where you're navigating but um that's what life is right now I mean we start off and we say hmm that's it that's resonating with people let's let's double click on it and and push on it and iterate on it and see where it goes we're definitely in product Market fit finding world I mean we have a lot of people who are very passionate about artifact but it's not hundreds of millions of people yeah so we need to figure out how to get to that scale and the thing that has very clearly uh resonated with people is not being locked into a set of specific Publishers but rather allowing people to go far beyond that and discover interesting links anywhere via artificial intelligence and that's the bet we're making and actually if you think about it there's like I think the reason that that resonates is there's not really any version of that out there today um so we're going to see if that works but um I'm really proud of of the team and I'm proud of what we've built so far and um it's it's cool it's it's nice to be back back in the back in the game yeah yeah back building so just the other question I had was then and you uh teed me up nicely you got that map you've got the end state do you know what the end State thing looks like then and since you're building out the the middle and that is the kind of the mystery do you know what you want to end up with uh should everything go exactly as you want I think the pattern if you look at every successful company whether they knew it or not at the beginning every successful company started oddly specific whether it was with just a desktop computer whether it was just selling books whether it was a social network for Harvard students oddly specific and whether they had these grandiose Visions or whether they basically were able to realize them over time actually isn't relevant but what is consistent among every great company is that they know that in the long run they can be much more and what they need to do is build along the path that allows them to be much more you know if Amazon had gone deep on you know writer tools or buying different Publishers and basically focused only on books they'd be a tiny company today but instead they focused on Logistics and infrastructure and they thought about like okay we're selling books now but what if just what if we sold things other than books and now they are where they are so I it's a pattern in business to learn from and apply to your own startup and that's what we're trying to do at artifact is to know that basically distribution of content via artificial intelligence feels like the future to me and I don't think it should be limited to short form video I don't think it should be limited to news I don't think it should be limited to Links so what can we do what what can we do other than selling books that's like the daily question and it took a really long time for Amazon to get there and we'll see how long it takes artifact great yeah we will see I love the product now thank you excited to see what happens appreciate it thanks very much for joining us [Applause]

Channel: TechCrunch

Views: 1,592

Keywords:

Id: t6EZlKa1L60

Length: 34min 37sec (2077 seconds)

Published: Thu Sep 21 2023