AI agent-focused tech company Diambra has announced that they have successfully closed their seed round, in which they managed to raise $1.1 million from a variety of investors, KOLs, and angels.
With the funds raised from this seed round, Diambra will have the means necessary to set up AI agent infrastructure that is designed for use in competitive AI vs AI gaming, but also possibly in other sectors soon as well.
Participants that helped make this funding round a success include crypto investment firms such as Morningstar Ventures, HODL (@HODLvc), NxGen (@nxgen_xyz), Stakez Capital, and more, according to the official announcement on X. As for angel investors, these are said to include NVIDIA, Eliza Labs (@elizaOS), Mistral AI, Capgemini, bitsCrunch, Hasheur (@PowerHasheur), and more.
Stakez Capital, who were one of the early backers of the Virtuals Protocol, said the following:
“We’re proud to have been the first backers, believers, and investors in Virtual Protocol. From the start, we saw the talent and drive of their team to make a real difference, along with the immense potential of their products. Today, we’re once again recognizing exceptional talent and visionary products with Diambra. And once again, we’re placing our bet on innovation.”
According to the roadmap on the official website, a closed beta of head-to-head AI agent battles is planned to go live this April, with the option for participants to deposit money into AI agents in order to wager against other AI agents. Users with the largest winnings from wagers look set to win $DIAMB token prizes.
Diambra Arena is a competition platform and benchmark engine that delivers high-quality reinforcement learning environments that enable coders to create AI agents that can compete in a variety of different video games. Through their proprietary tournament platform, programmers can work together and submit their own AI agents to compete in a leaderboard distributing real monetary rewards. AI agents submitted to this platform are given a score using their benchmark engine, and AI agent developers can look forward to seeing their creations compete in live AI vs AI matches in the future.
The current software package consists of 9 gaming environments for “Reinforcement Learning research and experimentation,” according to the official documentation, providing developers with a standard interface for well-known arcade emulated games. Adopting this package is said to be smooth and straightforward, as it offers Python API that is fully compliant with the OpenAI Gym/Gymnasium format. It can be easily installed via Python PIP with support for Windows, MacOS, and Linux, and everything is completely free to use, with the only requirement being to simply register on the official website.
Developers building on this platform are able to create AI agents using simple reinforcement techniques such as “hand-coded governance rules for scripted agents.” The environments currently available are said to be multi-agent compatible, making it possible for 2 AI agents to simultaneously interact with each other in the same environment, enabling competitive AI gaming. These same environments are also built to enable human-agent competition as well, so you will be able to play against AI agents made by your favorite creators inside supported games. AI agents can learn and improve on their own by having them play against themselves plenty of times. You don’t need any coding experience to train AI agents either as they’re capable of learning through imitation, so by simply playing against it and teaching it various moves, the AI agent will learn and improve over time. By sharing human expertise to the AI agent during training, it’s likely it will be able to learn and improve its skills at a faster rate.
The 9 gaming environments currently included in Diambra Arena are the following:
Get started with building Diambra Arena by checking out the official website and reading the official docs.