Edge magazine interviewed Geiba Newella (Gabe Newll) and now lays out pieces of conversation. At some point, the conversation with the co-founder Valve reached the topic of artificial intelligence – in the opinion Newella, In a few years, the development of AI will lead to a revolution in single games.
As Gabe recalls, for Dota 2 Already trained AI Openai, which surpasses the best e -sportsmen of the planet. But to win a person is not such a difficult task as entertaining a person. Scientists have not yet bought this puzzle yet.
By estimates Newella, Somewhere in nine years, AI will create, who knows how to do everything that a smart person can do. Crazy money will be spent on creating the first “silicon person”, but then the technologies will begin to quickly become cheaper – up to the point that dozens of extremely capable AIs will live on your computer.
Such artificial intelligence will change single games: they will be able to develop as MMO, royal battles and other multi -user projects are developing now. “A single game that can be played 20 hours a week, and it will only become richer and richer, and even after 400 hours I will be as interesting as the first 20 hours”, – describes the idea Newell.
As Gabe predicts, something similar in the https://nodepositbonuscasinos.co.uk/review/seven-casino/ industry will begin to appear in five years. And when the developers are achieved, in video games (including single), a real “tectonic shift” will happen.
Other comments Newella You can read from fresh interviews here and here.
March 21, 2020, sweep the floors and crumbs of monsters – Isabella from Animal Crossing made a partner for a classic doom
March 22, 2020 in May to Monster Hunter: World – Iceborne will fly a fire -breathing black dragon
The best comments
According to Newell’s estimates, somewhere in nine years will create AI, who can do everything that a smart person can do.
Artificial intelligence and non -heating character are not necessarily identified, it depends on the specific game already. What has so far been accepted to evaluate as “artificial intelligence” is just a set of algorithms, sometimes quite complicated to convince the player of a certain share of “rationality” what these algorithms are attached to-some bots in F.E.A.R., They are on the highest difficulties in Quake 3 / Unreal Tournament, bots in Starcraft 2 (which are called AI right in the game, but are not such). The main property of the “real” AI should be, I quote the ability to interpret external data, extract lessons from such data and use the acquired knowledge to achieve specific goals and objectives using flexible adaptation. Specific goals and objectives are set in the framework of the environment in which the AI is inscribed-for example, in some single role-playing game, this could be applied to his partners who would remember the player’s actions, analyze his methods of the game and offer his additions from their skills, some advice, and maybe even something completely different to the player and he believes even more in the “reasonableness” of these painted children. This is quite possible in 5-9 years, because remaining within the framework of specific games and the entertainment industry the spectrum of the required analytical tasks will be primitively minimal (and nevertheless, it will be enormous expensive). See this as that is Openai, there is DeepMind, and they are quite “reasonable”, the latter in Starcraft even came up with, if I may say so, to oversaturate the mineral economy – but these great silicon brains are useless outside their domain learning, they will not even be able to distinguish an apple from the orange, they do not even suspect what apple and oranges are an apple and oranges.
But II, who “knows how to do everything that a smart person can do,” in abstract terms-indeed, in 5-9 years will not appear. This is if you look at an attempt to simulate human consciousness through an imitation of neural ties on supercomputers. In my opinion, humanity has not advanced further than the mouse brain in this matter. If you look at the fantastic “sudden self -awareness”, then it, as in fantasy itself, is not predicted – if at all possible. As Heinline wrote, “self-awareness is cut through somewhere in the path of development from macromolecules to the human brain. Psychologists assure that this happens automatically when the brain accumulates a fairly large number of associative chains. In this case, I do not see any difference, protein is the goals or platinum."
Depending on what is meant by AI for the computer. Game, t. e. NPS most likely.
And what exactly is needed from NPS in games and what could be improved in them?
I can’t imagine this
AI studies by “fitting” to some answers, so it is easy to train the neural network to win.To. There is a marking on winning and losing moves
But I xs how to take markings for the "interesting" games)
Gabena doesn’t happen much – I have been waiting for thirty years and three years, now you can read an interview after a violation of the “radio monitor”
