AI and Games Q&A

One of the most fundamental aspects of any game is in its Artificial Intelligence. How players interact with NPC's, take on enemies and indeed how these NPC's react to the player and their environment is essential in immersing you in the world. The funny thing about AI is that it is at its best when it's not noticed. No one will ever bring up how an NPC walked into a room they were in, sat down in a chair and acted like they were busy writing. Nor will they mention an enemy attacking relentlessly until their health was low and then breaking for cover or making a run for it.

What they will bring up, consistently I find, is how an NPC couldn't figure out where that door was and instead glitched out and forever walked into the corner next to that door. Or how enemy pathing borked and instead of shooting you they unloaded into the wall in front of them. How it totally broke the immersion and looked a bit silly. This is not to say that it's unreasonable or even wrong to point out, just that when everything is working nobody bats an eye until the years of hard work that are put into these games breaks for just a moment and the veneer of the world is broken.

Enter Tommy Thompson and his YouTube channel AI and Games. A channel dedicated to breaking down the various AI and behavioral systems implemented in some of the most influential games. From his very first foray detailing the Goal Oriented and Action Planning AI of F.E.A.R to the Secret AI Testers in Tom Clancy's The Division 2, AI and Games has helped grow this humble casual (and literally thousands of others) understanding of these systems and more over helped develop a new found respect and understanding for the effort involved in creating these worlds.

Not one to bury the lead here, Tommy graciously took some time out to speak with us here at CGC and I couldn't be happier to share his story, so without further adieu let's dive right in!

Tell us about yourself, why you do what you do etc. What was the moment that made you want to start making these videos?

I come from a background in academia and work as a university lecturer. I studied a PhD in artificial intelligence and upon graduating in 2010 I decided I wanted to focus on teaching. That didn’t pan out straight away as the job market was a mess thanks to the recession at that time, but after working as a software developer for a couple of years I became a lecturer in Computer Science in 2012.

I wound up leading a games programming degree and during that time I taught a class on game AI. Given my PhD was in AI research in video games, this was an exciting opportunity for me. In amongst all the ‘core’ lectures discussing AI theory and such, I ran extra classes where I looked at some famous video games (notably F.E.A.R, Halo 2, Batman: Arkham Asylum) and broke them down to explain how the AI works. The students liked it, but it was difficult trying to schedule in these more ‘fun’ sessions around all the material I needed to deliver for the class.

So, I decided I would make some sort of online series that I would share with my students - and everyone else! At first it was going to be a blog series, but then I realized video might make more sense. So having zero experience in making a video (or how you even uploaded them to YouTube), I set about making a video on the AI of F.E.A.R. and I’ve now made over 50 episodes covering different games and topics – and I think they progressively got a lot better too.

AI is often critiqued in reviews and talked about in communities, but it can in the same breath be overlooked. That is, it's the invisible hand that is only noticed when it doesn't do its job right. What has become more apparent to you as you've continued your breakdown and analysis?

While AI is often overlooked, it is also the thing that really sells an experience given it adds another layer to the core game systems and mechanics. AI for games is what I often refer to as performance theater: it’s designed to set the stage and welcome the player into that world. To make it feel alive and interesting. It’s maintaining the fiction that you’re in this world and your actions have consequence. It’s always been interesting to me how the games industry is driven by visual fidelity – because fancy graphics sell and it’s easier to showcase that in screenshots and trailers – but then a title can suffer because once you get past the fancy packaging the inner workings of the game are a little broken. Your gorgeously rendered scene means nothing if the character standing in front of you starts walking into the wall, it completely removes you from the experience and that’s where AI is often exposed more than anything else. AI is typically the first thing people complain about when it doesn’t work as intended. Googling ‘<GAME NAME> bad AI’ will almost always pop up Reddit threads or YouTube videos on edge cases where character AI doesn’t work as intended. But it’s seldom celebrated when it works right. How many games can you name where people talk about how great the AI is? It’s often the same games listed every time. The ‘exceptional’ cases are often mentioned, but that’s only because somebody noticed it. I’d argue 95% of the time game AI that achieves its goals and does what it needs is never noticed by the player. Either because the technical challenges are overcome to a point that nobody outside of the development team would even know or it’s just that the final product is so satisfying, we don’t really think about it. But even when a technically competent and well-executed AI system is in the game and satisfies the developers design goals as best as possible, it does not always mean players will respect that. Reading the comments in my videos over the years – yeah, I know, I shouldn’t read the comments – has helped me understand how many people develop their own perceptions of how an AI works in a game, because they observe the behavior without understanding the inner workings.

One of the things I hope my videos can achieve is allow a viewer to connect the dots between their perceptions of how these complex systems work and how they actually work under the hood. Interestingly, there are times the viewer will reject what the video tells them – the comments that call me a liar are funny, the ones that write an essay telling **me** how the AI works are hilarious – because they refuse to reconcile their perceptions with the reality. I think there is a lot of parallels here with many aspects of the real world and modern society, but perhaps we can discuss politics another time! One of the first of many good examples on his channel is his breakdown of the Last of Us. Tommy dives deep into the AI architecture for each enemy faction, the human hunters and hostiles survivors, the infected runners and clickers to the adaptive partner AI that is Ellie herself.

Since starting the channel, has its growth in popularity affected your life personally? Even perhaps how you play games in your own time, not strictly done for analysis?

The impact on my day-to-day life has been quite drastic. I suffered from exhaustion and burnout from my academic work in 2017 and decided to step into a part-time position and I still work as a part-time lecturer today. This meant I could focus more time and energy on the YouTube channel, which is something of a requirement given some episodes can take days if not weeks to make. However, AI and Games is also now a consultancy business. Over the past 2+ years I have worked either as an AI programmer or a design consultant for some prototypes that hopefully get further funding as well as some in-development games. Plus, also working as a technology consultant to some larger firms.

In my personal life, the only real issue is finding time to play games. I schedule periods of my working week (and sometimes weekends) to record footage for an episode. And as weird as it sounds, you don’t play a game the same way when you know you’re recording it for a YouTube essay versus a Let’s Play. You have to approach it in a different way. I have around 4TB of footage from games recorded over the past 3-4 years that I lift from when I need to show game ‘X’ for 5 seconds in a video.

The knock-on effect is I sometimes do not have time to play the games I want to play! It is a ridiculous situation and I cannot complain given I am extremely fortunate to be able to do this. The reduced time to play games I want to play, combined with the fact that I have to wait for research materials to become available, means I rarely buy or play games in the year they are released. That said, I’m doing a little better in 2020 as I managed to pick up both DOOM Eternal and Animal Crossing: New Horizons.

Given the depth of the content that is contained in any one video, outside of editing, a lot of research and time goes into each project. With the internet being such a vast well of knowledge sifting through all that information just to get what you need must be daunting and I wanted to get a glimpse of the process.

For creating just one of these videos, you've mentioned trawling through textbooks, developer interviews and more. Without giving too much away how intense can the research be? Has this become easier over time?

An episode can take anywhere from two days to a week to research as I pull from many different sources. Sometimes those sources are hard to come by, so an episode can really start 6-9 months earlier as I slowly accumulate research materials. Even finding mods to crack games open or source code released years afterwards. Once all the research is done, the hardest part is collating all of that stuff together in a way that makes sense for the audience and minimizing the technical complexities that might be really confusing if you don’t have a background in AI and/or game development. It gradually gets easier simply through repetition, figuring out what are the fun facts that my audience will be interested in. Plus knowing where to find the material that I need to do my work.

But it also requires a certain amount of content to be in place before I can happily say we have enough information to make a video. I want my audience to come away from a video having a good idea of how that game works and hence accuracy is key. In fact, there are several topics I have started researching only to shelve them because I simply don’t know enough to do it justice (Middle-Earth: Shadow of Mordor is a good example of that) or slowly work on them as more information gradually comes to light.

My video on the DOOM reboot is a good example. I had a lot of notes on different parts of the game, but not enough to say ‘this is how it works’ And it was thanks to a developer reaching out to me who I later interviewed that I was able to bring it all together into what that video is now. And that really ties into the biggest change in this process, in that the parts of the games industry have started to recognize my work.

2019 was a watershed moment for the channel. In the summer I released a four-part series on the AI of Rare’s Sea of Thieves in which I interview half a dozen programmers. I had met some of the programming team at a conference before the game came out and after chatting via email, they invited me to hang out with them for a day at the studio and it was a great experience. Then my episode on the Hitman reboot was vetted by IO Interactive who offered to review it for me and my video on Halo Wars 2 has in-engine footage provided by Creative Assembly who also reviewed that episode as well.

That’s a big deal, given it means any mistakes on my part can be quickly corrected and it’s nice to see that they understand how important accuracy is to my work. I have since worked with Massive Entertainment on a three-part series on the AI of Tom Clancy’s The Division 2 released earlier this year. Ubisoft reached out to me about working on this project together and I was really pleased with how seriously they approached it and gave me freedom to tackle it in the way that I wanted to. I am immensely proud of that series and everyone involved from Ubisoft and Massive were great to work with.

Changing tracks a little, it's always seemed to me, outside looking in, that YouTube can be a minefield for content creation. From channel take-downs and the questionable AI (see what I did there?) they employ for copyright strikes to de-monetization practices a lot of work is involved in just maintaining a YT channel and making it as consumer friendly as possible, without stepping on company toes so to speak. A lot of content creators employ patreon to help supplement their income as a result, just to make sure they can feasibly continue their work.

I understand that patreon is how these videos continue to exist, with the constant writing and rewriting of YouTubes policies, has this made your process more difficult?

A few years ago the big issue was copyright claims as take-downs were issued all the time. This meant fending off false claims or dealing with issues where 20 seconds of uninterrupted game-play would result in a claim on a 20 minute video – and some AAA publishers would pursue this aggressively. Hence you often need to have your video online a day or so before public release so the copyright systems can do their work and flag any issues. Now I am often happy to defend and argue a point and have won more copyright infringement claims than I’ve lost, but that’s largely because the work I do is covered under fair use in copyright law: it’s a trans-formative piece made to critique and analyse a copyrighted piece of work. Which is much easier to defend than say a Let’s Play. Thankfully, it’s nowhere near as big a problem as it used to be as false claims died down and publishers seem to be a lot less trigger happy these days.

The more recent changes as a result of COPPA in the US (the Children's Online Privacy Protection Act) have not affected me at all, but then that’s because my channel is aimed at older audiences. Only maybe 3% of my audience is under the age of 18, while the 25-34 age bracket is my largest demographic. COPPA is having a bigger impact on a lot of Let’s Play creators largely because they target a younger audience. Each time you now upload a video you can self-certify your videos against the new guidelines, but they’re also vetted automatically, so if you try to subvert the system it’s a mark against your channel. My channel recently received an update saying that by default YouTube is now more willing to use my self-certifications because they’re consistently accurate and that requires us as creators being honest about what we’re making, identifying content (be it language, violence etc.) that could be harmful to minors and acknowledging it during certification. I can appreciate what it sets out to do, given YouTube and its creators like any video distribution platform need to be held accountable and make content that is appropriate for their target audiences. That said, it’s also very heavy-handed in parts and a lot of legitimate efforts to make content aimed at children are now suffering because their videos are no longer allowed to be added to the recommendation algorithms or participate in monetization, which could effectively kill them off overnight.

Meanwhile a lot of the other changes made to YouTube monetization systems in recent years don’t really hurt channels like mine that already have a head-start. However, for others starting out it’s a huge inhibitor and you really need to have built a small audience before you can start to make any money from it. Despite that, it’s a headache whenever YouTube decide to make changes to policy, given they seldom expressly communicate why a change is being made and how best to work within the new guidelines. Not to mention the fact that most changes are told to YouTube creators as and when it gets shared with the rest of the platform. I usually only find out about this stuff when it’s reported on either in YouTube itself or on social media and gaming sites.

So with all that said, my Patreon community have been fantastic and often one of the main motivators to keep me going when I feel a little deflated or exhausted from the process of making videos. Patreon and other crowd funding platforms are so important for creators such as myself, given I receive more in donations from my community than I make in YouTube revenue.

I think it’s important point to address that even now - after millions of views and hundreds of thousands of subscribers - my channel doesn’t make enough money to live on. I think many people see channels like mine and think we are laughing all the way to the bank. In truth, it really is far from that and it’s why I only work on the videos part-time, given I rely on my other jobs to pay my bills. What my Patreon community provides is the ability to comfortably take maybe spend a week or so in a given month to work on videos uninterrupted and without worrying about money and that’s a really big motivator when a video project is taking up a lot of your time.

By the way, congrats on the recent Creator Award you received, how does it feel to pass 100k subs?

Thank you! It’s all a little surreal, if I’m honest. The first video was six years ago and in the first three years the channel only gained around 4000 subscribers. I still remember it being a small niche thing that a handful of people paid attention to. Now my subscribers could fill a sports arena. That’s nuts.

That said, it’s also very rewarding and humbling to see such a large audience stick around to check out my work. I occasionally receive emails from viewers who talk about how a video helped influence a paper or project they wrote for college or how they used it for a game they made. Heck in a handful of cases it led to people studying game development and/or AI, which is amazing and I wish them all the very best.

I’ve had viewer emails sent in from the US, the UK, all over Europe and as far afield as South America, Australia, China and the Middle-East. Outside of the numbers going up on the YouTube Analytics Dashboard, that is the real sign that what I do has a positive impact.

Does passing this milestone change anything for you with the channel moving forward? Right now, it’s business as usual: videos are in the production pipeline that will keep me busy for the next 4-5 months and right now Patrons are helping me find content that’s a bit broader than usual. Recent episodes – Alien: Isolation, Splinter Cell, The Last of Us – have been very stealth and combat focused, but upcoming videos are moving in a different direction and I’m excited to work on them.

However, I am looking to the future and what different directions the channel could take. I have thought a lot about more tutorial-style videos or even a lecture series (an entire AI class perhaps?), but they need to fit the style of the channel better than a more traditional tutorial or lecture video. AI 101 will continue to grow as we cover more and more topics and hopefully create a good learning resource. Design Dive is my own personal series where I talk about stuff I want to talk about and more recently I’m becoming more confident talking about issues beyond ‘I (dis)like game X’, such as my episodes on lootboxes and war games, and I have some more thorny issues I want to tackle, but waiting until I know I can do them justice.

Arguably the biggest thing on my mind is expanding to other languages. The vast majority of my audience come from countries where English is the first language, but then it has slowly developed an audience in South America and Central Europe, plus around 15% of my audience watch with the subtitles on – though I’m sure for some it’s just a means to get around the Scottish accent. I am keen to find a way to better provide to these audiences: be it to re-release episodes with dubs in other languages or provide subtitles in each episode. That’s a larger endeavor that will need some investment to pull it off, but it’s something I am currently exploring.

Bringing it all back around again to gaming, Tommy recently collaborated with Ubisoft on a series devoted to breaking down the AI structures of Tom Clancy's The Division 2. Talking about how Massive Entertainment use AI to secretly test the game in a live test environment along with standard QA testing practices, the first of which you can see below.

The recent collaboration with Ubisoft for the division series, is this something you'd like to do more of in future?

Absolutely. It really helps when I can speak directly with someone who was actively involved in development and ask some of the more technical questions I have. It really helps ensure that accuracy and for me to celebrate the hard work of everyone involved at that studio. Also it’s just really nice to meet the people working at the studio. Everyone in both the Sea of Thieves and Division 2 series were excited to sit and chat with me. It’s not often a developer gets to share stories like this.

Does having the backing of the development team help put these videos together more quickly or does this process bring with it, its own set of challenges?

The one new challenge this presents is that by virtue of that collaboration the episodes are vetted by the studios prior to release. This means I must factor that into the process, and it influences how I make my videos. The Sea of Thieves series was a big learning experience as those episodes went through several drafts, largely because we had not discussed in advance what each video would be about and how I approach it. This meant there was some information that the developers happily shared with me during our interviews, but the studio later requested I remove it – largely because it could lead to exploits in the live game. Sadly I only received that feedback after the videos were largely complete and it meant a lot of work trying to address those issues and re-editing it all to suit.

It is really all about communication: what I want to explore, what studio is happy talking about and how we will go through the review/vetting process. My time working on Sea of Thieves heavily informed how I approached the Division series and it was a much more streamlined experience. We ran all the interviews; I did all my script writing and then I sent rough cuts of each episode for review. Only once those workprints were approved did I then go away and prepare the final versions for another review pass. But all of this was agreed in advance, so everyone knew what we were doing and how we were going about it.

Lastly, what's next for you and the channel? (I'll try rephrasing better but if there's anything you'd like to specifically shout out it's here!) Right now I have a stack of episodes of AI and Games in some stage of development, while I’m keeping some of them a secret, you can expect to see the Command and Conquer remaster as well as Cities: Skylines appearing on the channel soon. Plus, expect another collaboration with a game studio later this summer. This time I’m working with a smaller indie team and taking a look not just at their breakout game from a couple years ago, but also going behind-the-scenes of the impending sequel which is currently still in development. Should be fun!

And there it is, possibly one of our most in-depth Q&A's yet on the channel but a fascinating look in. If this has piqued your interest in some small way I'd very much encourage taking a look at the channel, you'll find links below, including socials of course! See you guys next time.

Links and Stuff

All images generously provided directly by Tommy for this article, thanks dude!

Be sure to sign up on the site for all things Casual, and create your own Casual Game Community profile.

Join the discussion on Facebook with our Casual Game Community, and Twitter to keep things even MORE Casual (get it?)!

Have a great content idea that would be perfect for CGC? Contact us here and we’d be happy to chat!