"True Artifical Intelligence" (sic) "coming to Space Engineers".... Really.

NickWiggill

First off, some background from earlier this year. In short, Google is using large computing clusters to develop AI that learns to play games using neural networks that train on the fly from scratch, using the framebuffer (computer vision).

Now we have claims of "TRUE ARTIFICIAL INTELLIGENCE COMING TO SPACE ENGINEERS". Really? Hm.

It's a laudable goal to create a general purpose AI that isn't based on neural networks, which are opaque / hard to understand the dynamic structure of, once they are trained... Q-learning (in conjunction with specialised planning algorithms) has shown promise for a while now, which is part of the approach taken here. But this guy Rosa is claiming they will be running "true" (does that mean strong?) AI on nothing more than a desktop in the near future. I can't see this achieving much more than a very limited form of general purpose AI at best.. which, granted, will be super for all of us, but is also not the strong AI he twaddles into in this piece, which really reduces any technical credibility.

I see this article -- not least the obviously smug vanity shot at the top -- as being either pure ego or little more than a way to attract interest / further funding to add to the (claimed) $10M already in hand... another attempt at hype-generation like Euclideon's patently untrue "unlimited detail" claims a few years back. I'm also blown away by the degree of apparent ignorance about the broader AI field... maybe he is pandering to the gamer crowd who don't know that for instance car-driving AI is already well along the way to real world use?

Your views? (aside from the fact that this is exactly the sort of publicity they wanted)

EvanGreenwood

Just regarding the claimed 10M already on hand. Given how successful Space Engineers has been, that figure is quite likely (if he hasn't earned that in profits, then that kind of money is certainly accessible to him now).

From what I understand he's setting up a research team to pursue his dream of developing general AI. And some of that AI development might trickle into his games (though the AI project isn't about making games, and it's being run as a separate business).

From my skimming of his blog post, I don't get the impression that he's especially qualified to advance the field of AI. His main qualification seems to be his wallet and his passion (which are still useful qualifications to have).

This probably doesn't hurt for marketing his games though. Space Engineers has a large programmer/engineer following. So good AI in is games is going to please his player base. It's a player base that probably is following developments in AI anyway (like car driving AI).

Full disclosure, I know Space Engineers, but I know very little about the field of AI beyond how cool the advancements in car driving and such are :)

Thanks for linking that article about Google's AI game playing advances :)

hermantulleken

Aside from the exposure they get from the game, I don't think it matters much whether it will in fact be "real" AI (by whatever definition). People have been making grand claims about AI since the beginnings of the field. I am always a bit surprised that when grand claims are not just shrugged off by most people. Until it's happened, who cares!

I think @BlackShipsFilltheSky has nailed it on the head; the man is pursuing a dream. Dreams by definition are not technically credible. And there may be another important reason why they are trying to get some spotlight. They seem to be recruiting, and for that I'd imagine they want as many people to know about it as possible, so that they can increase their pool of picks.

((If I ever had a huge surplus of cash, I would spend it all on infeasible projects that promise near infinite amounts of computing power. My dream is that we can all write our programs using O(n!!!!!!....!) algorithms, never have to worry about GC or foreach, and use the most abstract languages that we can come up with :) ))

dislekcia

So there's a Polish doctorate team working on AI that plays Desktop Dungeons. They've been going for nearly 2 years, their bots are pretty smart by now. Does that mean that we're a cool dev team too?

I think they should be publishing soon...

NickWiggill

@BlackShipsFillTheSky Agreed on all that. My AI knowledge is also fairly superficial. Glad to supply relevant news!

@dislekcia Ah, cool... roguelike bots have been around since the early-mid 90's at least, the Angband "Borg" comes to mind. I played more Moria / Angband than any other game growing up. I remember my old man joking with me after purchasing a shiny new Pentium-2 266, "Son, I buy you this fast machine with lots of RAM and all you do is play ASCII games on it? Why did I bother?" Good old days. Admittedly the machine was primarily for Quake II, TA & Angband. :)

@hermantulleken We all have different ways of addressing marketing, but I don't believe in taking people's money based on frivolous promises. Do what you say you are going to do, nothing less, and promise only what you can certainly deliver. I found the article in question, in a discussion between two non-developer contacts on facebook suggesting they'd like to "find out how to invest" because of the proposed AI, which is the point at which alarm bells rung for me and I commented on their conversation. Your dreams are your dreams, and may or may not come to fruition; business OTOH is something we do with other people. There's already been some discussion in the broader industry around empty game design promises, c.f. Peter Molyneux... though at his level, it scarcely matters. For the rest of us, I think it does matter what we promise... even idle spoken thoughts can end up being taken as commitments by gamers who hold a future stake in our work, and the backlash when these "promises" aren't fulfilled... let's just say that in this age of crowdfunding, PR matters.

hermantulleken

NickWiggill said:

We all have different ways of addressing marketing, but I don't believe in taking people's money based on frivolous promises.

Ah, OK, I understand your position better now. It's a tricky thing taking other people's money. People giving money based on promises should understand that those promises are likely to be untrue (even in formal investment scenarios you are more likely to lose than to win... you can only come up top if you are lucky, or can play enough times so that your wins offset the losses). But of course people often don't understand that, and then insincere promises are not so cool. (Not saying they are insincere, just to be clear, but if they were).

NickWiggill

@hermantulleken I should qualify my statement further, your remark made me think, and I'd say my view has changed a bit. I'd say there is a moral obligation when you are dealing with individuals who are willing to part with their hard-earned money to directly support what you do. That's the scenario I'm addressing here: indie / self-published / crowdfunded / funded by donations etc. Dealing with corporate investors who "can take care of themselves", and make their decisions following an informed business analysis and solid deliberation is an entirely different world. They take calculated risks to benefit off your resource... and as a developer, you are doing exactly the same thing in regards to them.

Thing is, the distinction can be blurred. Individuals who want to formally invest in a publically listed company -- does it matter if they're sufficiently-informed before committing? I suppose it does; shareholders at any rate should be kept aprised of major doings of the company though I guess the finer details may or may not be relayed to them. The point I'm trying to make is, I think we can all agree that such a scenario is quite different to an ordinary Jane/Joe spending $20 on kickstarter to see their favourite project come to life. Developer obligations in this case are direct-to-customer. That's where I'd apply the moral view liberally, by avoiding potentially unkeepable promises.

Does it apply where consumers are paying for their game through a larger publisher? I'm not sure. I think the dynamic there is very different and creates different pressures on developers than the direct-to-consumer model does... pressures that can perhaps justify a game not keeping all of its promises.