FULL DISCLOSURE: This is sponsored content for Verses AI.
Preliminary results are out for Verses AI’s (CBOE: VERS) performance within the Atari 10K challenge. The firms AI platform, Genius, is said to have outperformed leading AI algorithms within the tests.
The Atari 10K challenge is a variant of the Atari 100K challenge, wherein software agents must learn gameplay on their own using a limited amount of training data. The goal of the challenge is to efficiently learn to meet or exceed human performance across a variety of classic video games, with the exercise testing an agents ability in the areas of interactivity, generalization, and efficiency.
The results of the challenge, which were released in a blog post by Verses, outline that Genius trained for two hours on several games. Genius Agents are said to have matched or exceeded the performance of top ranking models that were trained on ten times the data and many times more compute. A highlight of the results include Genius outperforming humans at Pong, achieving perfect play on that particular game.
READ: Verses AI Aims For Commercialization In 2025 As Early Atari 10K Data Is Released
To provide an apples-to-apples comparison, Genius was placed head to head against IRIS, which is based on the breakthrough transformer architecture that is the foundation of generative AI and LLMs such as GPT, Claude, Gemini, Grok and others.
The IRIS model had variations of training time including 2 hours, 2 days, and 6 days, while Genius was trained on just two hours for all comparison periods. In each case, IRIS used 8 million parameters for its modeling, while Genius used 350,000 parameters, equivalent to 96% less compute. In every case however, Genius dramatically outperformed IRIS in the game of Pong.
The challenge also included the models competing in games of Boxing and Freeway, with Genius said to demonstrate progressive online learning throughout the training process. Overall, Genius outperformed IRIS with 90% less data, 96% faster training and a 96% smaller model size as per Verses.
READ: Verses AI Sees Genius Outperform OpenAI Model In Tech Challenge
“The Atari Challenge represents more than just machines playing games. Video games are a proxy for the complex dynamic systems all around us. We need AI to help us to better navigate the complexity and uncertainty of the real-world; yet state-of-the-art AI algorithms remain too unreliable, inefficient and unexplainable. While there is more work to be done to productize this research into Genius, we believe these early results and the research behind them signal a historical shift towards developing smarter, safer and more scalable AI,” commented Gabriel Rene, founder and CEO of Verses.
Further details of the Atari 10K challenge can be found here.
Verses AI last traded at $1.96 on the CBOE.
FULL DISCLOSURE: VERSES AI is a client of Canacom Group, the parent company of The Deep Dive. Canacom Group is currently long the equity of Verses AI. The author has been compensated to cover VERSES AI on The Deep Dive, with The Deep Dive having full editorial control. Not a recommendation to buy or sell. Always do additional research and consult a professional before purchasing a security.
