Goal State Launches on Kickstarter November 15th | AI and Games Newsletter 23/10/24
Here's everything you need to know about the AI and Games crowdfunding project. Plus AI testing in AstroBot, A Quiet Place, and a SAG-AFTRA update!
The AI and Games Newsletter brings concise and informative discussion on artificial intelligence for video games each and every week. Plus summarising all of our content released across various channels, from YouTube videos to episodes of our podcast ‘Branching Factor’ and in-person events like the AI and Games Conference.
You can subscribe to and support AI and Games on Substack, with weekly editions appearing in your inbox. The newsletter is also shared with our audience on LinkedIn. If you’d like to sponsor this newsletter, and get your name in front of our audience of over 6000 readers, please visit our sponsorship enquiries page.
Hey all,
here, and welcome back to the AI and Games Newsletter. This week’s issue is a rather selfish one, given it’s almost exclusively about my work and projects on the horizon. Critically, today I announced that my crowdfunding campaign for my online course project Goal State goes live on Kickstarter on Friday November 15th, with an initial funding target of £25,000. So we’re going to dig into what Goal State is all about, what the intention is behind running a Kickstarter for it, and what to expect in terms of the funding target, stretch goals and much more besides.But before all that, we need to catch up with AI in the news, as well as upcoming events. Including that other project I cooked up this year: the AI and Games Conference.
Announcements
Before we dig into Goal State, a quick round-up of important announcements.
AI and Games Conference 2024
The final batch of tickets for the AI and Games Conference went live last week, and sold out within hours. So we are now 100% sold out, and no further tickets will be issued for the event as we have hit our venue capacity. However don’t despair if you’re missing out: we will be recording all of the talks at the event, and will have further announcements on this after the conference has passed.
Plus we ran the final article here on the site giving a breakdown of each of the talks coming to the event. You can also find the full schedule now live on the conference website.
Other Announcements
As mentioned in the last newsletter, I was in France at the time of publication as I was hosting professional training classes on Large Language Models and their applications within games. Thanks to my attendees, and again the team at Gamasté for having me attend. It was a lot fun - and get in touch if you’re keen to have training courses run for you and your teams.
In the meantime, I’m slowly winding down on attending events this year. But a shout-out once again for Next Level, which happens later this week in London. I’ll be giving a very lightweight introductory talk to AI for game development.
Next Level 2024, King’s College London, UK, October 25th.
Konsoll 2024, Grieghallen, Bergen, Norway, October 31st and November 1st
AI and Games Conference 2024, Goldsmiths, University of London, UK, November 8th.
AI (for Games) in the News
Some news stories that caught my eye related to AI for games in the past couple of weeks.
Street Fighter VI has it’s own player-sourced AI bots now? [Kotaku]
I’ve been meaning to check this out: the recent updates to Street Fighter VI has seen the creation of V-Rivals: AI controlled fighters that are generated from player behaviour. This sounds markedly similar to the ‘Shadow AI’ in the 2013 Killer Instinct reboot, which I’ve covered in a previous case study. As a fighting game fan, I’ve been annoyed with myself that my interest in Street Fighter VI waned not long after release. Perhaps this will get me back in.QA on PlayStation 5’s AstroBot was Powered by Machine Learning [GamesIndustry.biz]
One increasingly practical application of AI in games is using it for Quality Assurance (QA) - the process of identifying bugs and other issues in a game and then resolving them prior to and after release. At the recent CEDED event in Japan Hiroyuki Yabe, Yutaro Miynotauchi, and Nakahara Hiroki gave a breakdown of how they trained AI bots to play and interact with their recent smash success AstroBot as part of the QA process. In short, the idea was that by taking actual human playthroughs and QA tests, and then having the AI either fully replicate it, or use it as the basepoint to generate an entire simulated player. This is super interesting, and I would love to run a deeper analysis on it somehow.“Your gaming experience is going to look very different to mine.” [PC Gamer]
Last week I had a lot of people pointing me to this piece over on PC Gamer. The thesis is quite simple: that in the future using real-time AI-driven post-processing filters on a GPU will enable for each player to play the same game, but experiencing it in its their own unique way. No longer just upscaling images like NVIDIA’s DLSS, players would be able to customise the output such that you get to decide the visual style of the game as it comes out the other side. This was inspired by work people have done taking video of games and passing them through what are called ‘video-to-video’ AI tools, where it will take the input and generate an output in a different visual style. So this article unsurprisingly went down terribly with both players and game developers, in large part because it conveniently ignores that games are built with a specific art direction and focus in mind.
Do I think this is something companies will begin to offer as a product for future games? Where an AI post-processor changes the style of the games visuals? Well, everyone seems to have forgotten that Google pitched this exact thing at GDC 2019 for the now defunct Stadia platform. They even had a demo in the showcase, they just never shipped it! Like a lot of things Stadia-related, it seemed to be something that developers, and players, weren’t interested in.The SAG-AFTRA Strike Moves Forward [GamesIndustry.biz]
After my piece discussing the SAG-AFTRA strike a few weeks back, we’re seeing some impact and forward movement in the whole process. The first was a an editorial piece by Rob Fahey highlighting the impact that this strike is beginning to have on live service games, with the recent content update to Genshin Impact having no voice work in it - a notable deviation from their previous work. However, it appears that both sides are returning to the negotiating table today. Fingers crossed we begin to see some progress on this front.
New Games
Last thing before the big story: what games are recently out, or coming soon that have caught my eye.
A Quiet Place: The Road Ahead (PS5, Xbox Series X|S, PC) - Out Now
Giving off big Alien: Isolation vibes, this horror game based on the popular film franchise is perhaps the ultimate stealth game. Make too much noise and you’re dead! Keen to give this a shot.Call of Duty: Black Ops 6 (PS5, PS4, Xbox Series X|S, Xbox One, PC) - Oct 25th
The true litmus test of Microsoft’s whopping $69 billion acquisition of Activision Blizzard. Does putting Black Ops 6 on Game Pass yield the returns that Xbox is looking for? But more importantly to me, is all this talk about cool enemy AI in the campaign of any merit? I’ll find out when I have time.
Horizon Zero Dawn - Remastered (PS5, PC) - October 31
I had a look at the AI of the original Horizon game a few years back, and we’ve got two talks about the sequel at the AI and Games Conference. So yeah, might just see this as an excuse to check in with Aloy all over again.
Goal State Launches on Kickstarter on November 15th
Okay, let’s get back to the big story! Earlier this year on
I announced Goal State, an initiative to provide accurate, accessible and practical learning materials for people interested in the world of artificial intelligence for game development. Given my previous experience as a university lecturer and researcher, and my ongoing work as a professional trainer in the industry, it was a desire to correct an issue I see far too often - of learning resources that are not reflective of how AI is used in the games industry - either not covering the practical skills developers use, or not respecting the reality of how game productions actually run - and therefore placing current and aspiring developers at a disadvantage.Over the past 10 years working on the AI and Games YouTube channel, and now this Substack, I have cultivated an audience who trusts me to inform them accurately. One of the most edifying parts of my work is meeting developers who tell me how influential or valuable my videos have been in their journey into the industry. As such, I have been keen to expand into something more longform that can help deliver something more comprehensive, that is effectively the culmination of my decade of YouTube work to-date. You can read more about my motivations for Goal State in a blog I posted earlier this year.
While I have been building a collection of professional training materials - given I work extensively with businesses on upskilling and general introduction to the subject matter - it was always my goal (no pun intended) to run online courses via a crowdfunding model. The big reason for that, is that this particular series of material is being built specifically for the people who engage with my work online who have supported me over this past decade.
One of the most vital aspects of my early days on YouTube was that the show has been crowdfunded since the early days, and it’s been great to watch as community of people grew to start following what I do, and were willing to put their money behind it. This is the audience that I want to work with to build this content, and by having them come in at the ground level through a crowdfunding initiative, they get to see it develop, give feedback on its overall quality, and get access to the content earlier than anyone else.
So let’s talk a little about what the Kickstarter will fund, and also the overall budget breakdown and stretch goal targets.
What Kickstarting ‘Goal State’ Will Fund
As discussed earlier in the year, to build these online resources will take several months as they build a bespoke set of materials for an online platform. This will also cover time spent writing, recording, editing, polish and final distribution. Plus ensuring that I can work with additional personnel to work in producing this work at high quality, with new audio and art assets. Critically, Goal State needs to be produced at a level that reflects the quality bar AI and Games has set on YouTube for many years, while also ensuring the existing production pipeline is not affected. That means bringing in support in production and editing. As such, this is not about putting money in my pocket, this is about paying for the time of multiple people (some of whom already work with me, others who will be hired in) to ensure we not only maintain our quality bar, but ensure we scale up such that everything else we’re doing at
continues to run smoothly, with new YouTube videos every month written content on Substack each week.I’ve broken up the material into two elements: the theory, and the practice.
The Theory is all of the lecture-style videos as we dig into each topic. This includes breakdowns of core computing and AI fundamentals, specific technologies and approaches, deep dives into how it works, industry relevance and examples, and quizzes and other homework tasks to test your knowledge.
The Practice is when you take all this knowledge and start trying to apply it in a game production. As we write code in different programming languages and game engines so that we can show how this all works when complete.
It was important for me to separate these two elements in a very clear way both in the syllabus, but also in the crowdfunding. Given I recognise there is an audience of people who perhaps only want to learn the theory but aren’t interested in becoming AI programmers for games. Meanwhile the practical parts of each course we build under the Goal State banner will have different permutations, given building these systems in the Unity engine for example, is different from how we would do it in Unreal Engine.
As we’ll see shortly, there are actually multiple courses planned, broken down into both theory and practical streams. But we will start with the Game AI 101 course: this picks up on a lot of the material that I have explored throughout the YouTube series over the years - navigation algorithms, constraint solvers, planning AI, state machines, behaviour trees and more. Critically these are technologies used to build non-player characters and common AI aspects of a games experience. The full breakdown of what is in Game AI 101 is in the article linked below.
I will announce the support tiers in the coming weeks, but in the meantime the key thing to takeaway is that you’ll be able to buy access to the courses, buy premium versions of the courses with extra bonus materials, educational and professional licenses for universities and games studios to use in perpetuity. There will also be early bird sales of most tiers too! But also, if you want to just throw a little bit of money towards it, as little as £10 will give you the inside track on how it’s all coming together.
Budget Breakdown
To make this nice and easy, here’s a fun infographic (‘fun’ is perhaps generous), that gives you an idea of how the funding breaks down.
Let’s go into each item in more detail:
Taxes & Fees (~20%): Sadly first things first is we lose a chunk of money straight out the gate with taxes paid to the relevant authorities, as well as fees related to use of crowdfunding platforms. I’ll be working to mitigate this as much as possible courtesy of my accountant, but not much else to say except big sad face.
Recording (~33%): This handles my part of the process of building the content itself. Including the writing of all videos, curating examples, writing code, and supplementary material. Essentially everything you will see and receive as a student on a Goal State course.
Production Support (~19%): This is the additional production support, including production management and editing, that will is needed to get this material over the line. I cannot do this on my own, and will be working with partners to ensure that quality is maintained while balancing across multiple work projects.
Art/Audio Assets (~15%): I have already spent a bunch of money on putting together the new Goal State brand and relevant assets (see all of the graphics I’ve used thus far). But we will need even more that fit in the brand guidelines, as well as audio and animation.
Updated Recording Equipment (~6%): To ensure video quality is at our desired level, we’ll need to buy in some additional equipment to make that happen.
Hosting and Promotional Budget (~4%): This is to help get the final product out there to end customers. Building and sharing storefronts on the likes of Coursera and Teachable.
Travel Budget (~4%): As part of the overall production, we will on occasion use external recording facilities and not rely solely on our own equipment. This will help cover costs of any personnel participating in that process.
Funding Target and Stretch Goals
When the Kickstarter goes live on November 15th, we will start with the following terms and setup:
The initial funding is to cover all of the production for the theory component of the Game AI 101 course.
30 Hours of video lecture material on Game AI techniques.
This will have a funding target of £25,000 (GBP) / €30,000 (EUR) / $32,500 (USD)
The Kickstarter campaign will run for 30 days.
In the event we’re successful, the campaign will adopt the new ‘Late Pledge’ feature available on Kickstarter, so that people can continue to support the project after the initial window ends.
Stretch Goals 1-3: Game AI 101 Tutorials
With this funding goal reached, this will allow for us to build all of the theory and non-technical homework assignments for the Game AI 101 course. But of course, the thing I want to do next is ensure that people can get their hands dirty in the code and start building stuff. But also people will want to explore it in game engines of their choice. As such, the technical component is broken up into stretch goals for each game engine that we wish to support.
Unity Engine: £31,000 (GBP) / €37,000 (EUR) / $40,000 (USD)
Unreal Engine: £37,000 (GBP) / €44,500 (EUR) / $48,000 (USD)
Godot: £43,000 (GBP) / €51,500 (EUR) / $56,000 (USD)
Stretch Goals 4-7: Course 2 -
Machine Learning for Games 101
Should the funding continue to be successful, then our next stretch goal will be significantly farther off, but critically it will introduce my second course: Machine Learning for Games 101. In our first course, we make little mention of machine learning, given it typically is not a practical approach for game AI. However, machine learning and in-turn deep learning have numerous practical applications in the games industry, both on gameplay side, as well as in production and it is important that we then dig into this some more and cover this in more detail.
Right now I have not published what this course will look like in terms of structure, but it will be based on one my more popular professional training classes that I run with games studios. Expect an update on this course if we pass the initial funding target.
ML for Games 101 - Theory: £60,000 (GBP) / €72,000 (EUR) / $78,000 (USD)
Practicals #1 - Python ML Projects - £65,000 (GBP) / €78,000 (EUR) / $84,500 (USD)
Practicals #2 - Unity: £70,000 (GBP) / €84,000 (EUR) / $91,000 (USD)
Practicals # 2- Unreal Engine: £75,000 (GBP) / €90,000 (EUR) / $98,000 (USD)
Stretch Goals 8+ - Game AI 201 & More
The final stretch goal, which frankly is here in the event that we reach a staggering level of success in crowdfunding is the sequel to the original class: Game AI 201. This would build on the first course, but also expand into new areas that we have not covered. At this point I’m going to get into too much detail on this, but in the event we start hitting that level of interest, then I will give a deeper dive into the structure and format of it all. But also, there is a fourth course I have in mind, but let’s not get ahead of ourselves…
Game AI 201 - Theory: £90,000 (GBP) / €108,000 (EUR) / $117,000 (USD)
Links and Resources
I’ll be continuing to update the Kickstarter page throughout the campaign with more information, but in the meantime the
site is the best place to find stuff.Follow the Kickstarter here.
Read everything posted on this site related to Goal State here.
Wrapping Up
As I wrap up for this week, I’m excited and, full disclosure, a little terrified!
is going out on a high note this year, with the AI and Games Conference just over two weeks away, and then Goal State’s Kickstarter right after it.I really appreciate everyone’s interest and enthusiasm for these projects, and I hope I can count on your support to get them over the line. With that, I’ll get back to preparations, and I’ll see you all next week!