Intro

I'm someone who likes games, and particular thinks that engagement fostered by games can show people things that they might not otherwise believe. Worldwide, the populace is cautious, concerned, or anxious of AI and the potential for rogue AI that poses an existential threat to humanity. There are a number of games such as Brenda Romero's Train and Lucas Pope's Papers, Please designed to explore complicity, elicit empathy, and more. There is a 'wargame'-like nature to these games - you are not thinking through a hypothetical, you are acting them out.

Across AI-optimists and -pessimists, there exists a wide variety of outcomes that have been suggested and predicted for the future towards which AI development may develop. This generally includes utopian and more-often [non-mutually exclusive] dystopian scenarios such as:

• 'Seemingly-benevolent' but autocratic AI overlords (a la Jack Williamson's 1948 novel The Humanoids)
• Naive eradication (paperclip maximizer)
• Designed eradication (chemical warfare /The Matrix)
• Bolstrom's Treacherous Turn and deceptive AI
• Tegmark's Life 3.0
• etc etc

Most people here would agree that AI safety and AI alignment is important, largely implying that such work will reduce the risk of p(doom) and hopefully help avoid some of the more treacherous dystopian futures and bring forth better futures. For the non-technical, but concerned folks amongst us, it can feel a bit maddening to watch the pace of development ever-increase while experts' (even including AI-accelerationists) p(doom) creeps higher and higher. Beyond lobbying, and talking about it, it doesn't feel like I can do much about this. However, public opinion and regulation have some potential to help, so why not try to connect to more people in more places and increase society's estimate of p(doom)?

Idea

So, the basic idea is a video game where you are the CEO of an AI research company, largely inspired by the gameplay of business simulation games such as Game Dev Tycoon. The game's goal is to navigate the world to a utopian ending thanks to your stewardship of the company. However, there are many pitfalls along the way - and AI alignment is the [untold] name of the game. 

Players need to navigate the costs of development and raising the associated capital, commercializing (or not) products to build the flywheel of development, politics (export controls? military contracts?), and choosing how much to invest in safety along the way. 

The rewards great - newly discovered cures for diseases; elimination of traffic fatalities; great leaps in energy efficiency and clean energy tech, enabling widespread desalination, and de-desertification.; abundant and ethical food; etc.

But so too do cracks slowly form, did your AI tell someone to hurt themselves? Did you ignore signs that your AI has developed the capacity for deception? For proliferating on its own accord? Did your autonomous vehicle drive into oncoming traffic? Get hacked? Hack a nuclear reactor? Put too many people out of work and caused a revolt against your company or your government? By the way, just opting out and slowing investment won't save humanity: watch as your competitors surpass you, out-raise you, and whoops the world exploded.

There are dozens upon dozens of possible endings, the majority of which are apocalyptic - but importantly I think the game should be 'winnable'. Were utopia not possible, presumably all of our p(doom)s would be closer to 100% than wherever they are today (which may already be 100% for some of us of course). Maybe there is a 'fun' route where you turn into an autocratic military-industrial complex that suppresses all AI development worldwide.

The goal is to be engaging, make players empathize (not sympathize) with decision-making of AI-accelerationists, and also hopefully frighten them into seeing how our current societies approach likely ends in doom, hopefully enough to do something about it.

Outro

Other games have been proposed and funded by EA before, but that's not necessarily what I'm suggesting here. I'm not a video game developer (or even a software engineer of any kind) or designer or writer, nor do I know any, and I wanted primarily to share this idea with the community to share and elicit responses and perhaps inspire some intrepid folks to take a stab at this. Admittedly, one thought came to mind during the course of writing this to get an AI to develop the game for me.

P.S. I think this kind of wargame idea would also work to explain mutually-assured destruction and nuclear safety, where the vast majority of outcomes end in annihilation (some of which exist in the AI game) but carefully tread nuclear disarmament is possible.