Moreso, this hellpoint dev really seems to be well beyond the scope of the kinds of games he make. I know a lot about programming, but I don't go and say that advanced technologies which aren't even really ready for widespread implementation or even close to being available on consumer level computers within the next 20 years should be the next thing to focus on, or even what I'd like to see. Sure, it'd be great, but I'm realistic.

While this dev doesn't usually say the things Gamingbolt says they do in context, in this case, he is actually saying he believes that this is going to be achievable with next gen CPU's.

But, to put this in perspective, what he's talking about is more akin to what you see with Forza's Drivatars, but adaptive to a singular player's playing habits. It's not really a neural network in the way he describes it, but rather, a set of potential parameters to change up the way a particular game play option may be available. What he wants is adaptive AI, that adjusts its strategies to the player's playing habits, to help keep things fresh, and the player engaged.

A real neural network which would cause AI to behave more like a human which will adapt in real time, would require a significant amount of computing power. Well beyond what you see in anything available now, and for the next decade at least. Further, the actual AI algorithms themselves aren't there yet. Certainly not ones that are basic enough to run on the limited bus and processing power of modern computers. Cloud could help, but that's a hell of a lot of data to crunch in real time, and ultimately, is actually a waste when there are much simpler methods to achieve this, and have been implemented in games already to some degree. The only reason they don't seem more advanced now, is because devs don't take the initiative.

The thing about game programming is its not about using whatever fancy processing power or algorithms may be the next big thing. it's about finding ways to do the same thing in a much simpler way. Best example is physics processing. People keep saying how Physics needs to be better in games. I agree. But you don't do that by having complex physics algorithms, because games themselves use about ten core physics calculations, which are the same perceivable calculations that we experience in real life. Things like gravity, mass, acceleration, velocity, etc. You don't need to have quantum mechanics or theory of relativity level of physics to get things done in games.

Most of the time, Gamingbolt takes his quotes out of context for their headlines. This time, they didn't. But I do think this hellpoint dev often talks well beyond the scope of the kinds of game he makes. Furthermore, while I respect him as a dev, I think he should come to realize that gamingbolt does not have his best interest in mind, and his constant quotes about what he wants to see, or would like to see, allow gamingbolt to make him look foolish.

There's nothing wrong with him wanting to see things in the future. We can all be optimistic or excited for new tech. But as a dev, there is a fine line between being excited for something, and promoting such things well before their time. Saying things like this sets up expectations for the future which aren't obtainable, particularly in the time frame that this dev posits.

@Gumby

I have that same policy, but in this case, I wanted to know what he was talking about when he said neural networks, because that could be a few different things. In this case, he might be a bit easier to take seriously if he used the proper term, which would be adaptive AI, which is already obtainable and implemented to some degree, and probably will become more advanced, and may be used in the way he states. In what he talks about, it's not really a neural network, as its not in real time, nor is it a learning AI. It's just a set of parameters to change things up based on if certain criteria are met by the player. It's a large database of potential scenarios to assign different variables.