Wingman.ai is the latest venture from John O’Malia, the ex-PartyGaming CEO who recently launched his venture VAIX.ai into the world with the aim of leveraging AI technology in message bots and beyond.
Along the same theme, Wingman.ai is an esports-related project which he hopes will design bot assistants that players will effectively train themselves to play alongside them in teams of five versus five competitions in esports such as DOTA 2 and StarCraft.
Appropriately enough in the month when Blade Runner 2049 was released in the cinemas, the project sounds like something out of science fiction. But this is very much a live AI project that O’Malia believes will answer some key questions about how this technology will be deployed in the very near future.
“What we are doing with Wingman is knitting the ability to use natural language to impact the learning of powerful AI agents is actually where the magic is,” he tells GamCrowd.
He says that the aim is to offer esports players a choice of bots or agents which will play with them and be trained by them in strategy and tactics.
“One of the big challenges is that people like to play in teams, but that means they are reliant on playing with four other people, much less agree strategy and stuff like that,” he says.
“So, we’re starting to create agents that will be able to serve and adapt to the payer – that will be their wingman. You can have bit players, assigned different tasks, and they will learn with you interactively.”
He says this plays to one of people’s natural instincts – the desire to teach. “If you just train an agent, it will learn by itself by reinforcement learning. But if you add an extra level, the usability of the agents will go up exponentially, but people will really want to teach in their own way.”
Soapbox racers for the digital age
Where this has a relevancy to the betting world lies in how this ecosystem of human-directed agents might develop over time.
Famously, in the summer the developers at OpenAI, the Elon Musk-backed AI research outfit, announced it had built a bot that could take on all-comers in DOTA 2 in one-versus-one game environments.
It is now working on five-versus-five environments, a challenge which O’Malia believes will be beyond the technology’s capability.
“DOTA 2 was solvable because they played head-to-head and that didn’t need strategy and planning,” he says. “But what about five versus five? How do you generate strategies? One of the feedbacks from the pros was that they wanted to play with it. But give them an AI agent? Where each can play head-to-head? Out of thin air, you can create a space – a soapbox racer for the digital age.”
It opens up the prospect of betting on robots, in a very real sense O’Malia suggests. “We think that angle is fantastic for betting,” he says. “You will have robots that you can objectively observe, whose code will be locked in.”
In order to answer integrity concerns – and general fears over who is controlling what – O’Malia says the whole system would be on the blockchain meaning that the bots would be computationally locked down with no danger of tampering. “You could have bot-cased competitions going on all day long,” he says.
After AI and blockchain, this leads to the other seriously hot topic with regard to gaming and technology – ICO. “At the moment, we’re looking at a broader appeal and putting it on the blockchain and maybe even going the ICO route,” he says. “We want to create a huge open marketplace for these agents. We think open source is the way to go.”
He says the company believes it can get a working ‘alpha’ available within six months. “An ICO would be natural,” he says. “If it does catch on, it would generate a lot of interest.”
That is certainly true. Indeed, O’Malia is something of proselytiser when it comes to AI and open source and believes that developments shouldn’t be left to the big tech companies in the US and China to carve up the opportunities that will open up in the future.
“I think that AI ought to move to the blockchain,” he says. “Having ten companies in the world that control AI research is unhealthy. It’s too important a technology to be corporate captured. The analogy is that all the research is open source, and collaborating.
“It’s like the Tour de France and the peloton. Once you get near the finish line, though, these companies will break out to win the race. We won’t see the stuff published. I worry about the AI peloton. I think AI belongs in the public domain. Independent researchers being able to prosper. You shouldn’t have to join Google.”