- 2 years ago
Kevin Guo is the Cofounder and CEO of Hive. Guo has over seven years of experience in VC and startup companies. Kevin completed BA, BS, and Master's degrees from Stanford University. In 2017, Kevin pivoted their career to focus on AI, previously they were with Mithril Capital Management, where they worked on investments in enterprise software, consumer internet, and healthcare.
Kevin Guo, chats with Alex York at Founder's Forum about the possible changes to come with AI, how he founded Hive, and how Hive builds tools to that help the greater good through technology.
Kevin Guo, chats with Alex York at Founder's Forum about the possible changes to come with AI, how he founded Hive, and how Hive builds tools to that help the greater good through technology.
Category
🤖
TechTranscript
00:00 [MUSIC PLAYING]
00:02 Hi, everyone.
00:03 I'm Alex York.
00:04 I'm at Founders Forum in New York City
00:05 with Kevin Guo of Hive.
00:07 Thank you so much for joining me today.
00:09 Happy to be here.
00:10 So I want to talk about Hive and the inception of the business.
00:13 You guys create computer programs that analyze images.
00:17 Why did you get into that space in the first place?
00:20 How did you kind of see this as avoiding
00:21 the market you wanted to fill?
00:23 So when we started this company, actually--
00:26 this was back in 2014.
00:27 I had just did my undergrad and my graduate study at Stanford,
00:30 so I did computer science.
00:31 We actually were a very different business.
00:33 We had our own consumer platform.
00:35 It was a startup.
00:37 We had millions of users on this product.
00:40 And we had a problem around things like constant moderation.
00:43 There were some misbehaving users.
00:44 We tried using third-party models out there to solve this.
00:47 And we realized none of that was very good.
00:49 And so we had to build our own models internally.
00:51 And so you can say that we were our own initial customer
00:53 for many years, using it for our own internal purposes.
00:57 Over time, many of my peer and friends,
00:59 these other startup founders, had heard about what we built,
01:02 asked if they could buy this model from us.
01:05 And so then we started almost a side hustle
01:07 of selling access to these internal things we had built.
01:09 By 2017, 2018, I had realized that the consumer business
01:13 maybe wasn't going to become a long-lasting, enduring
01:15 business.
01:16 But there was something really special
01:17 about the models we built for ourselves.
01:19 And so we ended up basically shutting down
01:20 the consumer business and turning
01:22 the rest of the company into what is now Hive in early 2018.
01:25 So in the last five and a half years,
01:27 we really just continued that original vision
01:30 that we had of, can we just build the best models,
01:32 initially, again, for us as a customer,
01:34 but now for the rest of the world to use.
01:36 And so today, we work with hundreds
01:37 of companies that are sending us billions of pieces
01:39 of content every month.
01:40 And it's really what we call is to understand content.
01:43 It's image, video, text, and audio data.
01:46 Content moderation is still probably
01:48 one of our largest well-known areas.
01:50 But we do a lot more of that now.
01:52 So in terms of content moderation
01:54 and analyzing all those forms of data and media,
01:57 what were you really looking for when you first
01:59 created the platform and the technology for yourself?
02:02 And how is that different from what customers are really
02:05 looking to analyze today?
02:06 Yeah.
02:07 So initially, when we started, it
02:08 was more of the obvious attributes, right?
02:11 So things like graphic content, nudity and violence,
02:14 things that definitely, as a kind of first principle,
02:16 all platforms need to be able to identify and prevent.
02:20 But how it's been has gone far beyond what
02:22 we could have imagined there, right?
02:23 So some easy examples.
02:25 Our age model is a very popular one,
02:27 making sure that there's no minors that are
02:28 present on certain platforms.
02:30 That's a big one.
02:31 Even things like counterfeit logos and products.
02:34 We have a logo model that's very good at identifying
02:36 10,000 different brands and making sure that's not
02:38 being present where it shouldn't be.
02:39 I think a model that's very recent,
02:41 it's very top of mind for a lot of people now,
02:43 is around is something AI generated or not?
02:45 We have the only production grade model
02:48 there that can tell you for a given image or a video,
02:51 one, is it AI generated?
02:52 And then two, if it is, which engine made it?
02:55 Was it a mid-journey or say with a Fusion or a DALI?
02:58 And this is really useful for a lot of platforms that, one,
03:01 let's say you're a creative platform,
03:02 you want to make sure people are on creative synthetic
03:04 creators, that's bad.
03:05 But the other thing is, let's say you're a big advertising
03:09 platform like a Facebook or Google,
03:10 and you want to make sure people aren't spreading
03:12 misinformation, especially around political ads.
03:14 Something that's very timely now.
03:15 And these are not-- for us, it's interesting
03:17 is we couldn't have predicted this five years ago
03:19 because generally AI is fairly new.
03:21 But this is what's, I think, interesting about businesses,
03:24 there's always new challenges that will come up,
03:25 and we keep building new products with that.
03:27 Yeah.
03:28 And you alluded to it a bit there,
03:29 but what is the business case for hiring or using
03:32 Hive technologies from those other multiple different kinds
03:36 of platforms perspective?
03:38 Why are you guys such a good investment for them to make?
03:41 Yeah.
03:42 I think it's like buying any other service.
03:44 And there is the trade-off of, do you really
03:46 want to do it yourself?
03:47 Or should you, in a sense, outsource it
03:50 to another party that is going to be better and ultimately
03:53 even cheaper potentially to perform that product?
03:56 And so for us, it comes down to--
03:58 even for our very large customers,
04:00 we work with Reddit and Snap and Spotify.
04:02 These are businesses that have amazing engineering
04:04 teams internally.
04:05 But it turns out that they actually
04:07 do not necessarily have the same access
04:09 to the wide range of data and the deep experience
04:10 we have to build these models.
04:11 So even these very technical companies will say, look,
04:13 at least I have a machine learning problems,
04:15 we're going to use Hive as our cloud provider.
04:17 We're going to trust them the same way that they made
04:18 a decision, for instance, to use an Amazon Web Services
04:20 or as a Google Cloud back in the day
04:22 for a lot of other compute.
04:24 We're not going to take all the ML work away
04:26 from internal teams.
04:27 But for the narrow things that we've done really well,
04:30 it makes all the sense to use us rather than
04:32 try to do it yourself.
04:33 Yeah.
04:33 You guys have also gone through pretty extensive fundraising
04:36 and really high evaluations with the company.
04:39 Where exactly are you guys now in terms of how much money
04:42 you have raised?
04:42 I know it's in the multiple 100 million.
04:45 And most recently, a $2 billion valuation.
04:48 Where are you at with that right now?
04:50 And why do you think you are so attractive to investors?
04:53 Yeah, I think, look, AI in this current day and age,
04:56 it's as people would view this as the next wave.
04:58 So it's a little mobile now as AI.
05:00 I think what needs to happen in this for any sort of, I guess,
05:04 industry shift to occur, you need
05:05 to be able to show that there can be
05:06 in very businesses built too.
05:08 And I think what's unique about us is everything we sell
05:10 and build is AI models.
05:11 But we also have a business model behind it.
05:13 We have real customers.
05:15 We have an efficient business model.
05:17 Our goal in the next few years is ultimately
05:19 to go public and be an independent business.
05:22 And I think that's really attractive to investors.
05:24 I think this is something that hopefully there
05:27 will be more companies like us rather than ones that are just
05:29 being built for, say, an acquisition.
05:31 I think it's actually important for there
05:33 to be more options out there and to give customers
05:35 more choice ultimately.
05:37 But yeah, we've had great success the last few years.
05:40 Raising capital is one benchmark of that.
05:42 But really, from my perspective, it's more about,
05:44 are we building really high-quality products?
05:46 Are we innovating?
05:47 Do customers love using us?
05:49 And then ultimately, yeah, I think our end state is to say,
05:51 can we become a cloud provider of AI models
05:54 talked about in the same language
05:56 as people would talk about in Amazon services or Google
05:59 Cloud?
05:59 Yeah.
06:00 And the Hive platform has multiple different use cases,
06:03 multiple different offerings from a marketplace
06:05 to NFT sites, things like that.
06:09 Who are the majority of your customers today?
06:12 Yeah, we focus primarily on companies
06:13 that have a lot of content.
06:14 So it's social platforms, marketplaces,
06:17 like you mentioned, media companies even, too.
06:21 But more recently, there are new customer types
06:24 that we couldn't have imagined before.
06:25 So for instance, we've had several different insurance
06:27 companies on board with us in the same time
06:29 in the last few months because they've
06:30 noticed a massive uptick in insurance claim fraud.
06:33 People would upload an image of, say,
06:35 their car bumper with a scratch on it.
06:36 And it turns out that scratch was AI generated.
06:39 And they noticed this because in the month of August,
06:41 the number of claims for bumper damage
06:44 increased by 100x or something.
06:46 So it's kind of interesting to observe
06:47 that our customer base, too, will naturally
06:50 expand as these challenges enter into different domains.
06:53 Totally.
06:54 And as consumers start using these technologies as well,
06:57 you guys have to be even better.
06:58 It turns out this is just human nature.
07:00 If there is a vector through which they could abuse
07:03 or take advantage of it in some way, they will find that path.
07:06 And in an argument, that's where we exist,
07:10 is to provide the tools to companies
07:12 to understand that problem and combat it.
07:14 Yeah.
07:15 What are some of your biggest expectations or predictions
07:18 in terms of AI and Hive's use case
07:20 as more and more of these technologies pop up,
07:22 more people are using them in so many different ways?
07:25 Where do you see your biggest opportunity or expectations
07:27 in the next 5, 10 years?
07:29 Yeah, I think it is definitely the case
07:31 that how we engage with the world
07:34 is going to change very materially
07:35 over the next few years.
07:36 What we're seeing with these general models right now
07:38 is kind of precursor to that, but the concept
07:40 of general intelligence is coming.
07:42 And in that process, it's not going to be suddenly--
07:46 you snap your fingers, now the world is different.
07:48 It's going to change fairly gradually.
07:50 And every one of those little changes
07:51 will result in some sort of adaptation
07:53 that we have to do as humans.
07:55 And I think in our business, Hive,
07:57 we exist to help us in that journey.
08:00 We want to provide tools, again, primarily companies,
08:02 to be able to, in some cases,
08:05 counteract the negative impacts of these things.
08:07 So one tangible example I'll give you
08:08 is I think AI-generated content is going to be
08:11 the majority of content you consume on the internet
08:13 in the next 10 years.
08:14 And if that is true, there are many consequences of this.
08:17 And at minimum, we would say, at least you should be able
08:19 to have a label that says, all right,
08:20 was this thing made by a human or was it made by a model?
08:23 So at least let you, as a consumer,
08:25 have some ability to make a better judgment call.
08:28 I really believe that there are very few companies out there
08:30 that can provide that accurate kind of metadata.
08:33 And that responsibility will fall on companies like us to do.
08:36 - Yeah, in a previous interview you did,
08:38 I know you mentioned, as you've grown as a leader
08:41 in this company, having to realize that other people
08:43 might be better than you at given tasks or jobs
08:46 and just the importance of outsourcing
08:48 and building a team that has comparable skills,
08:52 but in very different areas.
08:54 What are some other lessons that you've learned
08:56 as a young founder that you've had to kind of
08:58 maybe stumble upon or really be conscious of
09:01 as you've been growing your company?
09:03 - Yeah, I think absolutely organization building is,
09:06 businesses and scaling, I think that becomes more and more
09:09 of a present problem.
09:11 It is true, in the early days when there was 10 of us,
09:14 I could do a lot of things myself.
09:16 As a business, we're at 100 and now 300,
09:18 you have to scale, right?
09:19 And it comes to finding the right folks that can,
09:22 like you said, do things better
09:24 than I could have done myself.
09:26 And I think that's held true today,
09:27 whether it's technical product development,
09:29 whether it's sales and business development, recruiting,
09:32 these are all functions.
09:33 Really, the goal is to hire folks
09:34 that will be able to take ownership of that
09:37 and scale the business up.
09:38 And then I think that is something
09:40 that as a younger technical founder
09:42 and where I did a lot of things myself
09:43 to overcome that was challenging.
09:46 But today, I've totally internalized that.
09:48 - Yeah, totally.
09:49 I wanna ask too, in terms of the under 30 community,
09:52 you're an under 30 alum,
09:54 what is your biggest either piece of advice
09:56 or something that you would advise young founders today
09:57 to look out for as they're growing their own companies,
10:00 given the market that we're in,
10:01 the constant changes that we're seeing
10:03 with different technologies?
10:04 What is your advice for them
10:05 and how to kind of move forward?
10:07 - Yeah, I think there was a period a few years back
10:11 when it was tempting to raise a lot of money
10:14 because it was easy.
10:16 I think maybe these were founders
10:18 that hadn't seen tougher times.
10:20 Myself, you could have included in that batch as well.
10:23 Times are changing, right?
10:24 I think it turns out that it's maybe not easy
10:26 to get money anymore.
10:27 You have to be a little more disciplined in how you spend
10:29 and very deliberate in how you build, right?
10:31 And so I think that's probably the biggest change I've seen
10:35 for a lot of these younger founders
10:36 are now going through this current time period.
10:38 They don't realize that it's not always so easy
10:40 to build a business.
10:41 In some cases, you really have to,
10:42 and you get a lot of traction early on,
10:43 you have to do more with less, right?
10:45 I think that that mindset is,
10:47 the reality is it's gonna,
10:48 that austerity mindset, I think,
10:50 will have to last longer than most people realize.
10:51 I don't think we're gonna go back to those one times,
10:54 anytime soon.
10:56 - What is the biggest risk you see
10:57 in the AI industry right now
10:59 as someone who's really growing a business in it?
11:02 - I think our single biggest concern
11:03 is that all innovation just consolidates
11:05 with the existing incumbents.
11:07 I think that's really scary for the world, honestly.
11:09 I don't think it's good for Google, Amazon, Microsoft
11:12 to basically be the single controller of all this technology.
11:15 And I think even for a lot of startups
11:17 that you read about in the news,
11:18 it's a little odd to me that they're all, in a sense,
11:21 either raising large amounts of capital from,
11:22 or in some cases, even being controlled
11:25 by these big tech companies.
11:27 And so I think that's, again,
11:28 that's why we've gone our path here,
11:30 which is we haven't raised money from them.
11:32 We're not even hosting that infrastructure.
11:33 We build all of our hardware ourselves.
11:35 We really are trying to present an alternative.
11:36 And I would hope that there were more companies
11:38 that will do that too,
11:39 because they're getting something that's really important.
11:41 And this is unusual in that in every other
11:43 kind of major industry shift the last few decades,
11:47 there are always a lot of startups
11:48 that were independent and competing,
11:50 maybe not all of them succeeded,
11:52 but at least there was a chance.
11:53 I am concerned in this current dynamic
11:55 how it just seems that all these companies are kind of,
11:59 it's a little bit consolidated right now
12:01 into those three businesses.
12:02 And I hope that there's more alternatives out there.
12:04 - Yeah.
12:05 As an up-and-coming company,
12:06 how are you continuing to differentiate yourself
12:08 and encourage customers to come to Hive
12:10 as opposed to some of those incumbents?
12:12 - Yeah, I actually think it's because of the fact
12:14 that they know that we are truly independent.
12:15 That active health in some ways, as an example,
12:18 I don't think a company like Snap, for instance,
12:20 would feel very comfortable sending all their content
12:22 to an entity owned by Facebook.
12:25 And so the fact that they know
12:26 that we are truly not associated with big tech,
12:28 we exist primarily to serve best-in-class models for them
12:33 without any ulterior motives.
12:35 That's actually really refreshing for them
12:36 because every other company they work with,
12:37 they have to wonder, okay,
12:38 they took a lot of money from Amazon or Google,
12:40 are they really independent
12:41 or are they kind of subsidiary now?
12:43 And so I think our story is actually clean for our customers.
12:46 I actually think it helps a lot.
12:47 And that is, with itself, a value prop for them.
12:49 - Totally makes sense.
12:51 The last question I have for you,
12:52 I know that growing up you played chess
12:54 and for a while you thought that might be your career path.
12:57 What are some of the through lines that you have noticed
12:59 or similarities or major differences,
13:01 I'm sure there are a lot,
13:02 between chess and being a company founder?
13:05 - Yeah, I know chess was a big part of my life
13:06 and I think it's funny.
13:08 It's one of those things where at the time
13:12 I didn't necessarily appreciate the game
13:14 at the level of how people could be.
13:16 Now I can trust more again, it's fun.
13:18 But how it permeates everything we do,
13:20 it's about planning,
13:22 it's about thinking a few steps ahead
13:23 and wondering, okay, if I do this
13:25 and our competitors will do this move
13:27 five years down the line, what would that mean?
13:28 And I think an easy example I'll give is five years ago
13:30 when we pivoted into this business,
13:33 we made the usual move of building our own infrastructure
13:36 and not hosting on a public cloud.
13:38 And our investors were stunned by this.
13:40 Winning and prevailing wisdom that was out there.
13:43 The model then and now has always been
13:45 you raise venture capital and you funnel it
13:47 into a public cloud to host.
13:48 No one builds on infrastructure.
13:49 But I told them, look, you got to think
13:50 a few moves ahead here.
13:52 If we're really trying to compete
13:53 with an Amazon or a Google in the long run
13:55 and we're hosted on infrastructure,
13:56 how will we ever win?
13:57 It just didn't seem, if you just plan it out
14:01 and you do the numbers behind this,
14:03 you will inevitably have a more expensive product.
14:06 And so we made the choice to build on our own hardware.
14:09 And our investors at the time,
14:12 they weren't that pleased with it, I would say,
14:13 but they accepted it.
14:15 I think now they would agree that that was probably
14:16 one of the single greatest decisions we made.
14:17 And I think that's an example again,
14:19 where we had to really think far ahead
14:21 and really calculate out that this was best for us,
14:23 even if this is a more unusual move at the time.
14:26 - Yeah, definitely.
14:27 And what would you tell the under 30 community today?
14:30 Is it to play chess before they start a company
14:31 or some other piece of advice that you think
14:33 that the under 30s should really use
14:35 as their guardian light?
14:36 - I think what I would say is,
14:40 first is there is going to be a huge amount of opportunity
14:44 in AI, what I call application businesses.
14:47 And I think this is something that,
14:49 on the one hand, it's interesting.
14:50 It's almost gonna be,
14:51 there's gonna be too many people going into it,
14:53 but at the same time, I think there's also gonna be
14:54 not enough people building very useful,
14:57 like kind of vertical specific applications.
14:59 And I think that is where there's gonna be a lot of value.
15:02 And I think that's what I would say
15:03 is if you're a younger founder now,
15:05 looking for something that could be a little bit more
15:07 called capital efficient,
15:08 maybe you can build on top of other APIs,
15:09 whether it's companies like OpenAI or even Hive,
15:11 we have a number of these products too.
15:13 We wanna make it easy for developers
15:14 to build applications on top.
15:15 I think this is something that you're gonna see
15:17 kind of like in the early days of mobile
15:20 and apps that emerged,
15:21 something like this is gonna happen for AI.
15:23 I think there's a lot of exciting opportunity there.
15:24 - Cool, well, thank you so much for chatting today.
15:26 I'm so excited to have talked about the business growth
15:28 and everything that you've done.
15:30 - Sure, happy to chat, thanks.
15:31 (upbeat music)
15:34 (upbeat music)
15:36 [BLANK_AUDIO]
Recommended
0:40
|
Up next
Be the first to comment