EVE Fanfest | The Infinite Game – Fireside Chat

Añadido: 2026-05-22

[00:00:01][applause] >> So, it's always wonderful to be here, but uh it is especially wonderful right now because uh this is the first time we host Fanfest as Fenderis Creations.

[00:00:16]And uh yeah, cool.

[00:00:19]Uh and uh I'm going to be joined here for uh for a while with Adrian Bolton, uh who is from Team Mind. Everyone.

[00:00:29]>> [applause] >> Um and of course, people naturally have a lot of questions of like why are we building a partnership with one of the leading [snorts] AI companies in the world.

[00:00:43]Um and that is very much because I believe building partnership to tackle this big topic is something that every responsible society needs to do.

[00:00:53]And we're here to be the janitors of EVE Online, and we will have to figure this out like all other countries.

[00:01:01]Uh and we thought it was very fitting uh to get help on that front to kind of frame it in the kind of in the in the right arena uh to get the President of Iceland to uh to give us her thoughts. Um this is also her first time at Fanfest. She was elected between Fanfests, so she will now maintain the tradition of part of the job of the President of Iceland is to talk to the uh Capsuleers of EVE Online.

[00:01:35]So, uh I am sure she will do a fantastic job. So, please join me in welcoming on stage Halldóra Tómasdóttir, the President of Iceland.

[00:01:45]>> [applause] >> Wow.

[00:01:52]>> [applause] >> Congratulations again.

[00:01:59]Wow, wow, wow.

[00:02:01]It's a little humbling to be here. You are a bigger country or community than all of Iceland.

[00:02:07]Um but I'll gladly take on the role to be um your president as well, if you allow me.

[00:02:14]I understand that in some of the worlds, you will need to kill somebody or I will need to kill somebody to uh stay in control. So, I will go to the universe where you will democratically elect, I hope.

[00:02:27]But really welcome everyone. Welcome to Iceland and welcome to Fanfest. What an incredible community this is. I don't believe there are many places in the world where thousands of people travel across oceans united by an online universe to discuss spaceships, virtual economies, diplomacy, strategy, and even the future of humanity itself.

[00:02:51]What an incredible universe this is.

[00:02:54]And EVE Online has never really been just a game.

[00:02:58]What began in Iceland in the 1990s with Hilmar and an incredible team was an ambitious entertainment and business venture, but eventually became something much, much larger.

[00:03:13]A living universe shaped by millions of human decisions, relationships, alliances, conflicts, and stories, just like the world.

[00:03:24]This is an extraordinary achievement because in many ways, the developers of EVE understood early on something that much of the world is only now beginning to fully grasp.

[00:03:41]Technology is never only about technology.

[00:03:46]Behind every line of code, every innovation, and every technological leap are human choices.

[00:03:57]Who is included?

[00:03:59]Who is excluded?

[00:04:01]What is amplified?

[00:04:03]And what is left behind?

[00:04:05]Technology is about people.

[00:04:10]But also about power.

[00:04:12]About creativity.

[00:04:14]And perhaps most importantly, about trust.

[00:04:19]It is about the systems we built together.

[00:04:23]And even more importantly, and sometimes forgotten, the values we embed within them.

[00:04:30]And that is why your community matters.

[00:04:36]We are entering a new era in the world shaped by artificial intelligence.

[00:04:42]Technology that is transforming how we work, create, communicate, govern, and imagine the future itself.

[00:04:50]It's changing everything.

[00:04:54]And naturally, much of the conversation we're having in the world, and probably here today, focuses on capability.

[00:05:00]What can AI do?

[00:05:02]How fast will it evolve? How powerful may it become?

[00:05:07]But I believe there are more important questions that need to be talked about as well.

[00:05:13]What kind of humanity will AI help us shape?

[00:05:18]Will these technologies deepen trust or division?

[00:05:25]Expand creativity or diminish it?

[00:05:30]Help people flourish or only optimize efficiency?

[00:05:37]On what we will focus will determine the outcome. And these are not just technical questions. They are fundamentally human questions.

[00:05:46]Perhaps the most challenging questions we have faced in our lifetimes.

[00:05:52]And communities like yours have an important role in answering them. And I hope you take that on.

[00:06:00]Because virtual worlds are not escapes from society, at least not only that.

[00:06:07]They can be powerful testing grounds for society.

[00:06:12]Places where people experiment with cooperation, leadership, economics, ethics, identity, governance, and human resilience.

[00:06:23]Places where we learn that systems matter, they do, but people matter more.

[00:06:31]For more than 20 years, the EVE community has demonstrated the extraordinary power of imagination and technology placed in the hands of people who are determined to build something greater than themselves.

[00:06:44]This is worth celebrating.

[00:06:47]This is a worthwhile purpose.

[00:06:51]And it is so fitting that Iceland is a place for a conversation about this.

[00:06:57]How we build this new world, how we imagine it, and how we build it. We are a small country, smaller than your universe, but we have long shown that creativity, curiosity, and human connection can have global impact.

[00:07:14]And in a time of rapid technological change, such values matter more than ever.

[00:07:22]The future will not be shaped by technology alone, although its impacts have probably never been more profound.

[00:07:29]The future will be shaped by the courage, wisdom, and imagination with which humanity chooses to use it.

[00:07:39]So, dear friends, I thank you for visiting Iceland and allowing me to be your president even if only for a couple of minutes.

[00:07:49]And I thank you for helping us shape a future where technology serves humanity, not the other way around.

[00:07:58]Thank you and I look forward to the conversation.

[00:08:02]>> [applause] [applause] [clears throat] >> So, Adrian, uh that was quite the kickoff. That's >> [laughter] >> That's pretty amazing. Pretty amazing.

[00:08:23]So, uh maybe if we just start a little bit uh about like who are you?

[00:08:29]Uh yes. Yeah, I'm uh I'm Adrian Bowden.

[00:08:32]I'm a senior director at Google DeepMind and um I run the uh the inception team at Google DeepMind. So, we do breakthrough experiences, the the early-stage research. And uh picture on on the screen, this is uh this is DeepMind's in its early days. So, this is um 2012. That's me at the front looking very young.

[00:08:52]Uh Okay, so I mean this picture is black and white. I thought it was maybe from the '60s, but uh I see it's also in color. Well, yeah. And we And we actually love this culturally we love this. Uh this is a great example of the team. We're very close team initially and the first picture is actually our official picture for uh for for DeepMind um just before we moved office. We never We never actually take pictures. We don't do any of these things. You can see You can see I think the office was for 14 people. I think there were 35 of us in there at that point. It was getting pretty wild. And um the black and white one is our official picture and actually the photographer who we never We never normally get. The photographer came in and kept taking pictures. And actually the color one I love the color one because this is after the official picture and you can just see how uh the connected the team are actually and how happy they are. And how Demis is, you know, giving us bunny ears, which is very helpful. But but it's a fantastic example really of the culture of the team. Yeah. This guy is scratching his ears. He's probably like no regretting his choice. Exactly. Why would he sit next to Ben?

[00:09:58]Sorry, Ben.

[00:10:00]And actually loads of these people are still here. So So yeah, so we we we do a lot of the very early stage stuff and actually more importantly I think for our conversation my background is I'm a gamer for life actually and I've had a career of 33 years now I think in game development and and and now I I research. And along the way we've worked in I've worked in a lots of built lots of creative teams in in sort of adjacent technologies as well, game streaming very early on and and motion capture and game production and web and all these areas. So we've had a very long sort of background background in this space. What What was it like back then? You say it like this picture is from 2012 and Demis is wanted what? 2010? 2010. So >> that must have been like some like bat crazy stuff back [laughter] then.

[00:10:58]Well, what was interesting so I got to work with Demis originally in 1999 and I went to join him at Elixir Studios which was his games company as you know and we were then really using AI research in games. We were doing a lot of simulation and and sandbox games and you know, there's a lot of work in and around there. Demis's background is in black and white with you know, which is heavily RL and Theme Park and and Syndicate [snorts] and then we made Republic the Revolution there and and even then as a games company, Demis was then talking about he's going to build an AI company and do do real AI research. And then in 2010, he decided this was his perfect opportunity really to build DeepMind and build AI responsibly for the benefit of humanity as a as a mission statement.

[00:11:54]And it was a very interesting time because actually, if you know the history of of AI, everybody just thought it was impossible actually. In 2010, we were in what is defined as an AI winter and everybody thought AI was impossible and we were mad, which was actually in retrospect very useful because it allowed us to in in stealth collect together lots and lots of people into the core team that we wanted to to to build build DeepMind.

[00:12:24]And actually, we went and collected a lot of the the the game developers and creative creatives and artists and designers that we'd worked with in the past through our sort of sort of our companies. And we collected a lot of research academics that were interested in AI at the beginning who were crazy enough to join at that point.

[00:12:43]And it was yeah, it was an amazing time actually. And that sort of blend really set us up and this is an interesting it's an interesting component of DeepMind that AI had a and and is so heavily influenced our research by games and the and the game development community. And interestingly, I looked at this when we were we were talking previously and probably 30% of the people in here are ex-game developers of some description. And so it had a massive influence really on how we progress. Yeah, that's definitely been an observation in our collaboration so far, how many game developers there And they're still in DeepMind. Yeah, very much so. So So I mean, I can definitely connect to the craziness when we were making like a single-charted MMO in Iceland that had never made a game in '99.

[00:13:31]People thought we were like crazy doesn't even like capture it. That was I think we were. I mean, in fairness, we were.

[00:13:38]>> Yeah, I mean, obviously we were. That's how you achieve these things. We're here drawing blood from people to write poetry. Uh obviously not very >> [laughter] >> Exactly.

[00:13:49]Exactly.

[00:13:50]Uh so, I mean uh maybe if I talk a bit about like this kind of partnership idea.

[00:13:58]And as we dive deeper into it, so here's kind of a a picture I've been showing at FanFest.

[00:14:05]This picture is actually from FanFest 2017.

[00:14:08]And and uh there it's kind of like, okay, we started kind of in single-sharded supercomputers, then we were kind of exploring further immersion with VR.

[00:14:19]Then we had this moment of like discovering uh the friendship machine of EVE Online. What really what holds it together.

[00:14:27]Then obviously we uh were planning on using blockchain, and then we had AI there. And this is from 2017, and obviously talking about blockchain and AI in '17 was kind of actually crazy.

[00:14:39]But it's kind of interesting as I was reviewing it, because every step of the way we were kind of we built quite a bit of partnerships.

[00:14:46]So, to build the single-sharded computer, we worked a lot with IBM and a company called Texas Memory System. And we had like battery-powered SSDs to make the whole thing work. We're getting like bleeding-edge blitz from Microsoft on SQL Server. And at the at the time that EVE comes out, uh it actually does rank on the supercomputer list. Um uh it isn't officially ranked because they you needed to use like a special type of interconnect to qualify for the benchmark, but if I just added it all together, we were very much high on the list. Obviously, we're very low on the list now, thanks to you guys and your AI machinations.

[00:15:25]And then, of course, we actually did work with Google on VR. We worked with the Daydream team, which was a lot of fun, even if VR is not yet ready and and all that. And it's probably needs to attack it more, at least, then still the immersion effect of VR is very much there.

[00:15:41]And in a way, we partnered with Eve players to come up with the friendship machine. There was very much conversation, research into these kind of fundamental things of Eve.

[00:15:52]And we have Sveinn here in the audience.

[00:15:54]So, Kevin there and okay, they're here front row.

[00:15:57]Um and we partnered with them uh on Project Eve to build Frontier there. And it was always kind of this, okay, to a bit, what do we do on AI? And obviously now, that answer is and it's interesting. I was taking a I found this picture. So, uh there uh Demis is giving me a lesson on how AI works in the DeepMind office. And it's from like March 17th. And I was thinking, I probably put AI because Fanfest takes place months later.

[00:16:31]And it was probably this conversation with Demis from March 17th that, okay, I should put AI there. Eventually, this will be ready and we were like >> I remember him talking about this meeting. I wasn't there on from the other side and how enthusiastic he was about it.

[00:16:44]>> Yes, we were very enthusiastic about like building Alpha Eve or some like strong >> And this one on the left, this was taken when?

[00:16:52]So.

[00:16:53]>> [laughter] >> This picture The time dilation on that one is slightly off. Yeah, uh uh yeah, unfortunately, I didn't take a picture of myself with them. I should have done that. But this is a composite the the the uh government TV station of Iceland made.

[00:17:10]>> [laughter] >> Um And here is actually here we are again.

[00:17:15]This is December last year. Yeah. That was quite a journey.

[00:17:19]>> a while to get together. Yes, it took a while. It only took like Well, less than a decade. Yeah, yeah, that's pretty I guess it was 9 years or something.

[00:17:28]And I mean, obviously, CCP maximum cuts is there waving his hands talking about some Who will inherit the world? Our descendants or our creations? Kind of vibes.

[00:17:40]Um But, um Maybe this element of like Like, why why video games? Like, um Yeah. Why why have you guys used games and video games so much?

[00:17:54]Yeah, it's um there are I think in reality, when you're doing scientific research of any description, you ideally want lab scenarios. You want to You want to test your hypothesis in in limited labs, and you want to remove distractions or anything that could do training and testing. And actually, for for AI, virtual environments and games are the perfect low um you know, high safety uh uh high complexity scenarios. And we've used them um throughout uh throughout our work. So, we brought in a lot of experience, obviously, in game development and and and building them.

[00:18:31]And we used games for um for AI research, and they're very good for things like safe exploration. So, so you [snorts] can do an awful lot of training and testing on very early experimental uh AI research without impacting the real world. Um so, you can think things like um uh self-driving cars. So, a lot of the work in self-driving cars was done purely in simulation to start with.

[00:18:58]Billions of miles of of driving from, you know, the very earliest stage research uh to to higher levels. And that And you know, and you can test edge cases that these systems can never see. And so, they will have they will have done an awful lot of simulation and effectively game work before ever getting in a real car and, you know, inter- interacting with the real world and and real drivers. And that that strengthens that safety and security. It's very good for for research. And then, you know, we use them a lot for scalable challenge because we're actually game developers.

[00:19:32]Um it allows you to dial up and down the complexity of of the training and testing environment you want for AI. So, we use them a lot.

[00:19:41]As example, we use them in Atari a lot.

[00:19:43]So, in in our Atari work originally, you can sort of look at it on the face of it, it looks like we're just teaching uh an AI to play an Atari game. But actually, what we're doing is we're building we were building an AI brain to learn a general system so it could learn across a a lot of games. So, we were building a learning system. And so, there was 50 games in the in the suite of Atari games, and um they were, you know, all the classics. And there was boxing and racing and Space Invaders and all these things. And then, we would give the, you know, in terms of complexity, you would give all the the challenge, you would give the agent a stream of pixels. So, it doesn't actually see the screen. It's uh doesn't see the screen. Doesn't know what it's looking at. Doesn't know the game state.

[00:20:27]Doesn't know anything about the game.

[00:20:29]And you give it a bunch of um actions, buttons it can press. Um but in terms of scalable complexity, all the buttons do different things in different games. So, it has to learn a general understanding of of how to maximize its score across all these games. And so, video games are a fantastic space where we can add we can add a bunch of games to that or we can remove them or add complexity or not. Um so, it's a very scalable challenge. And then, thirdly, really what they're really good for is measurable progress. So, you can tell how good your AI is because um games are designed for us. And so, we can set benchmarks in games Uh and we can see the progress of our AI agents at various capabilities. So I can play Space Invaders and and score 100 points. Probably should have scored more than that, but I could score 100 points and you can put an agent in and it scores four points and you know you know very clearly that it's got work to do. It can hill climb too. And you can set very clear objectives for it to hill climb to to a a credible score. So very good to see the range. So one thing I mean as we have been talking about this that really I found interesting is because I mean there is a Demis holding a chess piece and and playing Go. I'm not sure if it's if he's good at Go. I mean I know >> pretty good now. Okay. Yeah, unsurprisingly actually. Yeah.

[00:21:55]>> [laughter] >> I mean but apparently he was very good at chess. He's very good at chess. So because this element of like when you hear okay some AI is going to learn how to play and what does that really mean and then you can think okay as per our president's comments about like what about the human factor? Yeah, absolutely. But as I was kind of thinking about this because when Deep Blue wins Kasparov in chess you would think okay chess is over.

[00:22:23]Nobody's going to play chess anymore because the computer is best at it.

[00:22:26]But frankly today chess has never been bigger. It's bigger and bigger. Yeah.

[00:22:29]Yeah. If you if you go and look at the MAU of chess.com I mean it might be the biggest game in the world because people playing on chess.com Yeah, and it was solved. It's a solved game. It's a solved game and but what I I found out obviously by kind of tracking it and and looking at it is that the fact that AI could learn chess it in no way elevated chess. Chess chess became an elevated game and there were more dimensions to explore as a consequence and I I've heard that happened quite a bit with Go So, move 37 to like now the kind of the understanding of what Go is has elevated again and more people are playing Go than ever. Absolutely. And I just had a big big event to celebrate the the 10 years of of AlphaGo and I missed it all in in Korea. And yeah, it's it's it's a it it it is a elevated area and you know, it's not a solved game.

[00:23:24]But it's but it there's there's a lot of people learning now actually from from AlphaGo and AlphaZero and how it plays and the moves it plays and the new techniques it's brought to the game really. It's such a big um you know, it's played for between 2,000 and 3,000 years ago.

[00:23:42]And so there's an awful lot of human knowledge and human experience has been passed down and the complexity. But it's unlike chess, the reason we were so interested in Go is it's it sort of set as the next complexity. It you can't brute force Go. You can't just calculate every move into the end, you know, as they did with with chess. You can't calculate the the perfect response to everybody's moves. There's too many potential moves. And so you need a you need new architectures and new systems that are much more human-like in estimation and and confidence in you know, moves. You're estimating the board and then when a move is played you you recalculate who's winning, which isn't always very clear in Go.

[00:24:25]And so you build an algorithm that estimates who's winning and and how good a move is.

[00:24:30]And it's so it's it's those new techniques have definitely become building blocks that we've ended up using in other research that we do across the board including you know, AlphaFold and all those kind of things.

[00:24:42]They are direct building block links to those.

[00:24:44]I I find that very interesting. The fact that Okay, Deep [clears throat] Blue masters chess to some extent. AlphaGo and AlphaZero master Go.

[00:24:54]Both games became bigger as a consequence. They were elevated, understanding greater, etc. And well, I mean, I can't draw maybe like a a very clear line, but intuitively the you could kind of see something in that way and happening with Eve. I mean, right now, I mean, as we think of the partnership, when first we're going to make like an offline version of Eve.

[00:25:16]Yeah. And I mean, it takes a while just to figure out all the do the basics in Eve. Yeah. And [snorts] then at some point we'll discover something and we're actually doing something similar in Eve where we're building a kind of a safe space for new players.

[00:25:30]Not really an offline version, but it's kind of like we've we've sanded off the edges a little bit for new players in the beginning. So, it's kind of interesting to think about like, okay, here's an offline version for an AI to learn and here's a safe space for new players to learn.

[00:25:45]I've heard like you've now stopped playing Slay the Spire.

[00:25:50]I know you're playing Eve.

[00:25:52]How's that going?

[00:25:54]Yes, I've I Yeah, I've started playing Eve. And yeah, I am beyond the the learning cliff, I think, beyond the learning curve.

[00:26:04]Vertical cliff.

[00:26:05]>> [laughter] >> Well, okay.

[00:26:07]I'm I'm I'm I'm enjoying it enough to think I'm beyond that.

[00:26:11]>> Yeah, okay. I will find out. But yeah, it's it's it's it's it's an interesting challenge to get get started. And you know, and I'm interested from from Aura's point of view and from from Aura's guidance point of view. And um and and interestingly, I mean, we talked about it, but I ended up It's a plug.

[00:26:32]It's a horrible plug for our own technologies we use, but we use them every day.

[00:26:36]Actually, it's terrible, isn't it? But in Gemini app is a camera version, which is Gemini live. And it turns out that there is an awful lot of public knowledge about Eve over the years. And the model knows a lot about Eve. And so, in particular places where I embarrassingly lost my my quest objects in my first ship for a different quest.

[00:26:59]And so I went to do my quest and they're gone and and the game tells me you don't have these objects anymore and I don't even remember what they were. I don't remember what they were. I don't remember where to get them.

[00:27:10]And so so yeah, Gemini live told me about the marketplace and and and where it is and now I can buy the replacement object I need for this particular quest. And and so that was very useful and then I used I used it to strengthen my ship for the next run. I had some money and and and it's amazing it's amazing um it's amazing the depth of these models actually. So so I wanted a new shield. I want to make sure I didn't die this next time.

[00:27:42]And um uh and it and it uh it detailed the module I wanted the shield I wanted but it also also said actually there's a cheaper version that will do 90% of what you want. Okay. Um [snorts] so I went to the marketplace and this was true. There was an expensive version which was the high line and then there was a cheaper version which did enough to get me through. And how are you doing that? Is it just like pointing the thing?

[00:28:04]>> point it and just and because it's a whole camera and it can see and it's seen so many you know public videos of of Eve and yeah it knows what the what the errors are. Fantastic. That's kind of cool.

[00:28:18]Well, I'm mostly playing Crimson Desert.

[00:28:20]I have not uh used AI modes but I've certainly used a lot of Google to figure out where all the all the stuff is. Mostly Google and YouTube.

[00:28:28]So I'm in that kind of maybe brings us to like why why are you guys so excited about Eve?

[00:28:36]Like it's very kind of flattering and humbling but why?

[00:28:41]I think you know we sort of talked about we've used games for for AI research for a for a long time. You're um you know, and it's about it's it's about really about three things. It's about the the the team and and the working relationship. It's about the complexity of the game. You guys know it's a very complex game. And you know, and about the community and about all of you actually. And for us um the team we got to work in on collaborate directly on features um previously and and that's the sort of combination of research and um and game development has a lot of overlap actually. You know, there's a lot of pressure on multi-disciplinary teams pushing on exploration in areas.

[00:29:32]And with research you'll you have your hypothesis and you're pushing hard on this hypothesis. And with games you're trying to build something exciting and entertaining and interesting. But ultimately both of them get measured at some point in the future. Uh your hypothesis whether it actually works.

[00:29:48]Um and your and your game whether it's actually fun and entertaining. So you have these great ideas, but in reality until you try these things you never know. And and those that sort of long-term relationship and iteration and and exploration and and development we found a lot of that in our partnership which was you know, the team came back raving about the about working with you. And and so you know, and you you have so many team members that have been with you for so long. It builds that kind of level of trust and incredible ability to explore new ideas.

[00:30:24]And we have the same. You don't actually We don't get to work with many companies that actually have people that worked on them for longer than we do.

[00:30:31]So so it's lovely and that means there's real strength in your team to be able to try things and experiment. And obviously the the your support for um science and the and the background work and the community and the fiction is fantastic. So the So the team um dynamic is is amazing for us. The complexity of the game, I don't need to talk to you too much about how complex your game is.

[00:30:56]Um but it's particularly unique uh and and particularly challenging for for a lot of the AI areas that we would like to focus on.

[00:31:06]Um and so, you know, obviously it's persistent and it's running whether you're involved or not. It's player-driven heavily uh and uniquely to a level that's that's incredible, really.

[00:31:17]Um there's actual agency for players. You know, so so you can you can you can change the world and it stays changed.

[00:31:26]Uh and there's a lot of people playing.

[00:31:28]You were the you know, there's massive dynamics and ecosystems and movement of um the dynamics of the game, which make that a very special um uh experience for for key kind of bits of research that we're interested in.

[00:31:41]So, you can long-term planning, um memory, uh continual um uh continual learning. These are key pieces of research that you want uh uh a great environment to learn in. And then, you know, ultimately the passion of this, coming here and seeing you all here and and you coming here every year to celebrate and uh this universe and this experience and this entertainment and and your engagement is is quite incredible. So, having a community that is that bought in, but not just um not just the company and the and the you know, the players, but actually the the the massive link between all of you and how much you you care about this this uh universe. It's It's hugely um we got massive respect for it. I mean, that maybe brings us into this the responsible way, like uh our president was opening up. I mean, the uh uh we're all trying to figure out I mean I guess all aspects of technology. I think we're just uh given the speed of change when it comes to AI, we're just we're forced to figure it out very quickly. Whereas usually we have more time. I remember when we're making EVE we're kind of in the actually the end of the internet bubble. We're kind of making EVE and we had had a good time to figure out the internet.

[00:32:59]I mean I logged on to the internet I think in '92 and then a bank in Iceland is on the internet in 2010.

[00:33:08]So I mean that took like uh I mean over a decade from first contact to like it's diffused.

[00:33:15]So [snorts] uh a bit on that. We actually have a QR code here up on the screen um and you can uh scan it and uh put in questions. So we might have time for a few questions at the end.

[00:33:29]Uh which are burning for everyone here.

[00:33:32]But uh I mean obviously we're very much in the in the in the mindset of we're here as caretakers / janitors of of EVE Online and uh and and we are very much contented with okay this is coming at us at like an amazing speed and everyone is kind of scrambling and uh some people are taking shortcuts. I mean uh for sure some gaming companies are trying to just um do the same things quicker and things like that and and I would just [clears throat] want to emphasize to everyone we're not to very not at all thinking about it like that.

[00:34:07]I mean this has this is almost as much an exploration of understanding of EVE Online as it is about understanding AI capabilities because I mean EVE Online is a marvel and we're still figuring out I myself I'm still learning about what it really is. And it's not the mechanics of the game which I'm also constantly learning and forgetting all day long but uh what it is really. And I And then when I emphasize this element, of course, we will begin with an offline version. So, I mean nothing immediately will happen or anything like that.

[00:34:42]Um then, I mean, we'll probably next take few of those things to EVE Frontier, where we're doing crazy experiments anyway. So, adding on top of that >> Yeah. to kind of test the edges of things. And uh but still, I mean, we we have been able to kind of bring to EVE I mean, uh already a a beneficial thing. And uh I mean, I can talk about it now freely, but it's really Gemini that uh plays a part in making order guidance uh give guidance without like going all over the place with the guidance.

[00:35:14]And it's interesting to hear that I mean, you just using your phone and Gemini live were getting some guidance.

[00:35:20]So, I think that even pets a path we could probably increase the the capabilities of order guidance. And I mean, just like you were talking about how um AlphaGo Zero played a part in opening up Go to more people. Mhm. Um I mean, uh then I mean, order guidance very much has the potential to open up um >> for a new safe zone. Yeah, and uh coupled with the safe zone and order guidance, we might be able to get more people thinking that have mastered the learning cliff like you have.

[00:35:52]Just hold that.

[00:35:52]>> it might not be a cliff one day. It might not be. It might not be a cliff.

[00:35:56]It might not be a cliff. Um And and of course, I mean, uh what we are celebrating here at Fanfest is the 10-year anniversary of uh of Project Discovery, where EVE players through their collective efforts have contributed to sciences in a in a meaningful way. From like astrophysics to medical advancement and and even uh bio uh molecular science and whatnot. And I always brag about this thing that EVE is the only fictional cover on on nature and I know like nature cover are kind of a kind of a thing in your business.

[00:36:33]>> They're kind of a thing.

[00:36:34]>> Yes, I I I saw you like you were you have like six or something like that.

[00:36:39]>> Oh, no, it's yeah. Or maybe you have more.

[00:36:41]>> won't talk about how many we've got.

[00:36:42]Okay. Okay. Yeah.

[00:36:43]>> [laughter] >> Lots.

[00:36:44]>> Well, you have more than one. But but but yes.

[00:36:46]>> Well, okay, but we have one and maybe we'll we'll and and actually project discovery when we were doing the actual planet discovery contributed to Michel Mayor getting a Nobel Prize. So, maybe we will can get them as another Nobel Prize.

[00:37:01]>> Yeah, I know. He's He's only got one now. Exactly, he only got one.

[00:37:04]I think maybe economy is a good for that. Yeah.

[00:37:08]>> [snorts] >> Uh so, um what are we excited about the future?

[00:37:16]Like Uh well, I mean, this is this is an amazing time to be a gamer or to be an AI researcher, frankly. So, I'm obviously in my my element, but with games I can I we touched on it, but we've got a lot of game developers actually in Google DeepMind and we are I think about as AI native as it gets now. We've been playing with this technology for for many many years and we have artists and designers and and QA and and and engineers and we have built lots of prototypes of brand new game play experiences have never been built before and could not be built built without AI. So, we're super excited and I'm um I'm old enough uh I'm old enough to remember the internet having an impact on games and you know, and adding a massively new dimensions. Obviously, Eve plays into that a lot, but being able to play with people all over the world is amazing.

[00:38:20]And you know, and and mobile provides, you know, gaming experiences or new gaming experiences to people all over the place. So, and and and both of them and and others um really expanded gaming in general and all the ways we can do it. And we think, I mean, we see the tip of the iceberg of some of these experiences that are creative and we're using these tools creatively new. And um there are some fantastic game play experiences we are coming, we think. And we think AI native will add another dimension in the same way. It doesn't change everything else.

[00:38:57]People, as you say, people are taking, you know, utilizing these things, but we think brand new game play experiences will be amazing, actually.

[00:39:06]Yeah, I mean, I I think what makes me excited um is this element of how AI, in a way, mastering other chess or Go, elevated both games, both how they were played, but also how many people are playing them. Like, finding out >> [snorts] >> some deeper understanding or a different way to look at what are Eve's issues and opportunities.

[00:39:37]Obviously, we have a expert game designers, we have counselors telling management, we have all e-players, etc. give us a lot of feedback. Then we have a lot of data from the game there where we can have a lot of insights. And we're in a way spoiled for like opinions, feelings, and and facts about what is going on in the game.

[00:39:59]And I think it is in many ways like every society is like full of facts about themselves. Um I think we are particularly living that now right now in on the economy front. Um all countries seem to be I mean, rather poor time of out economies. I mean certainly here in Iceland, Iceland is tackling like an interest environment and money supply post COVID etc. etc. And and it's sometimes like figuring out what is the right answer.

[00:40:31]Um just with like classical approaches with computers and economists etc. I mean of course that will continue but having some additional vector if we if we were to like realize some great insights about like what are really the the the stumbling blocks of learning EVE online.

[00:40:55]I mean in your own way. I mean you did the very common mistake of forgetting the mission item. Who hasn't undocked from a station and left the stuff you should be delivering behind?

[00:41:06]Very common EVE mistake.

[00:41:07]And and often these kind of issues maybe just they almost they they kind of just kind of tune out.

[00:41:15]They've been there for such a long time and we we often get more kind of drawn to other things mainly because our veteran community is maybe not like really thinking about that. And doesn't forget these things.

[00:41:29]Yeah but they're still there and and and finding some like new way to really kind of distill like for example for new players what are really the problems and if they were a distill like some insights about like okay how can we make the EVE universe more alive Yeah.

[00:41:47]living breathing etc. Um that that seems like a kind of a great Atlas motivation of like how can we elevate EVE by using extremely powerful tools to figure out what it is and realize this kind of EVE forever mission we have of like where we where we find every possible way to kind of underpin that vision.

[00:42:09]Yeah you're even talking briefly about the three points of release that we that's just come out and maybe that's something of interest to us. Yeah, of course. Yeah. So, I was maybe going to throw it out to the audience.

[00:42:23]Um if we have the ability uh okay.

[00:42:28]Okay, here are questions. Um >> [snorts] >> Okay, um Okay, would it be possible for this partnership to assist with repairing and replacing the spaghetti code?

[00:42:41]>> [laughter] >> Surely your your code's just perfect now. Surely. Surely. Yes, possible. Not a problem at all.

[00:42:50]Um [laughter] So, um I mean I I can even speak to this. I mean already coding agents have been extremely helpful in identifying issues in our code base because uh our code base has been now written over a period of 30 years.

[00:43:07]And a lot of the context uh of the code >> Yeah. is lost because I mean people have moved on or forgotten what they were doing at the time etc. and everything is of course like always poorly documented.

[00:43:21]And uh and what we have found like the rediscovering context from old code um coding agents are extremely helpful in that. I mean we have found a lot of like memory issues and I mean now increasingly they're finding security issues etc. Yeah.

[00:43:40]Um so, I mean I I I would say yes, they they are they're definitely assisting with repairing and replacing spaghetti code. That is already happening. I mean it's mostly humans that actually do the replacement, but discovery of like uh where the problem areas are uh that's definitely already occurring.

[00:44:01]Another question here, can AI win EVE Online?

[00:44:07]And and and so, winning EVE Online is a Yeah, exactly. What is winning? It has a certain meaning here, but um >> [laughter] >> Are are some people here winning EVE Online right now?

[00:44:20]Um what what do you think about that?

[00:44:23]Uh can you win?

[00:44:26]Uh Uh I think I think um we can we can get uh agents that are good at playing. I think the interesting thing for for for us uh is that you know, can a single player win? Yeah. Uh is a is a question. I think the exciting thing for us is that actually collaboration and communication and all these team and social dynamics you you if we want to build an agent to win, then then you got to want to use it and got to want to play with it. Um so I think uh that's the real sort of challenge for us. It's like can you build an agent that is actually um positively you know, you positively want to use uh and collaborate with. Um Uh Okay, what parts of EVE are going to be the main focus of Yeah.

[00:45:14]>> near steep dive partnership? Uh deep dive DeepMind?

[00:45:19]There's uh yeah, I think um well, I think we are we are very interested in in lots of areas. Actually, there's a lot of research teams inside uh Google DeepMind who are very interested in collaborating at various levels from the science teams um uh and uh all the way through to continual learning and all these things. So, the three things that we've sort of talked about are um uh long-term planning. So, this idea that you can picture goal far into the future and we are very good at stepping back from that and deciding what the steps are to get to it. Um but most systems at the minute are quite greedy and they will pick the first um uh rewards they can get. And so, you want to build long-term planning and an understanding of where you're going.

[00:46:03]And we're very interested in memory and not necessarily remembering things.

[00:46:06]Computers are very good at remembering, but we need to get systems that are that are performing in knowing what to forget, actually. And so you can get compression in all the It's actually more about forgetting than remembering.

[00:46:19]>> about forgetting. What's What's the right thing to forget?

[00:46:21]>> saying forgetting things is a superpower?

[00:46:23]>> It's a We are very good at that. Yeah, yeah, it turns out we are very good at that.

[00:46:28]Uh and then continual learning. And so, you know, it leads These models are incredible. The language models that are available now, they can do some fantastic things.

[00:46:35]But they're baked systems. You You take your data, you build your model, [clears throat] you train your model, you fine-tune your model, you ship your model, and people can use it, but but the use isn't is having no effect on the model. This is baked. And really what you want is is um use uh people using it to actually uh help uh the model learn at the same time. But but deciding what is important use and what isn't and and one of the techniques we talk about is catastrophic forgetting. You know, you can find that if you if you do continual learning, so if you allow constant learning, you forget previous skills.

[00:47:14]And so, these are very big um gnarly questions for us and things that we are from from the um uh from the games for AI research side of things, from our research, this is what we're super interested in. But there's lots of there's lots of interest around the other way. So, you know, agents and NPCs and and uh uh and explorations from from the sort of AI research for games component.

[00:47:40]Yeah. It's actually one of the questions is like can AI be used for NPC functions to make them more alive?

[00:47:46]It's actually So, the agent system that does most of the agent missions in Eve already was written, I think, 24 years ago, and it's a lot sleeping there just as it is and and it is a system that tries to procedurally generate variety in the missions.

[00:48:08]Obviously not very sophisticatedly but pretty good for those standards. So I mean the fact that you are in a way in Eve doing a mission for an agent. I mean it's it's almost like in the narrative itself. It fits very nicely so I mean all these elements of field more alive I think are very >> explore that. Yeah, very exciting things to explore. How can we just make the game feel more alive because the NPCs and the factions they obviously are statically there. Can we make breathe life into them is is exciting.

[00:48:44]Another question is can AI be used to reduce tidi in Eve online?

[00:48:49]Um I mean that sounds like an AI problem to me. Um Let's put some like compute to work on reducing tidi and then final question Adrian, what's your in-game name? No reason.

[00:49:04]>> [laughter] >> Uh Don't give it up. Don't give it up.

[00:49:08]Don't like trust me on this. Well, just just just to plug then then I've applied to Eve University but I haven't had my application agreed so you know if you could just let in the last few people that applied. That's how tough it is. You're not even allowed in the university yet.

[00:49:25]>> [laughter] >> Damn.

[00:49:27]Yeah, all right.

[00:49:28]But Adrian is here at Fanfest so I mean I'm sure you can track him down.

[00:49:34]Maybe we'll lure him on to Pop Girl and then somebody might let him into a university for Christ's sake.

[00:49:39]>> [laughter] >> Well, thank you Adrian for talking about this here with all of us.

[00:49:47]Thank you everyone here. Thank you.

[00:49:54]Thank you [applause] guys.

[00:49:57]>> Thank you.