In a significant development for the artificial intelligence landscape, VERSES AI Inc. (CBOE:VERS,
OTCQB:VRSSF) has announced that its flagship product, Genius, has outperformed OpenAI’s o1 Preview model in a rigorous code-breaking challenge known as Mastermind. This comparison involved over one hundred test runs, where Genius demonstrated remarkable advantages in speed, efficiency, and cost-effectiveness.
Genius achieved a staggering performance rate, solving the Mastermind code 140 times faster than OpenAI’s o1 Preview and at a cost over 5,000 times less expensive. The test was designed to evaluate the models’ reasoning capabilities through a series of games requiring logical deduction to crack a hidden code based on feedback hints. Each model was tasked with deducing the correct code within a maximum of ten guesses.
Hari Thiruvengada, Chief Technology Officer at VERSES, emphasized the significance of this test: “Mastermind was the perfect choice because it requires logical reasoning and the ability to predict outcomes based on decisions made.” This exercise not only showcased Genius’s superior performance but also highlighted the limitations of traditional language-based models in handling complex logical tasks.
The performance of VERSES AI’s Genius system was compared to OpenAI’s model, yielding results across several key metrics. Genius achieved a 100% success rate, significantly outperforming OpenAI’s model, which had a success rate of only 71%. In terms of computation time, Genius completed games in an average of 1.1 to 4.5 seconds, while OpenAI’s model took between 7.9 seconds and 889 seconds (approximately 15 minutes). The total compute time for all 100 games was just over 5 minutes for Genius, in stark contrast to OpenAI’s 12.5 hours. Furthermore, Genius demonstrated exceptional cost efficiency, with a total compute cost of approximately $0.05 for all games, compared to OpenAI’s staggering cost of around $263. These results highlight Genius’s ability to solve complex problems both efficiently and economically, positioning it as an attractive option for businesses seeking advanced AI solutions.
Gabriel René, founder and CEO of VERSES, noted that these results reveal a critical gap in the current AI landscape. He stated, “The limitations of language-based models like OpenAI’s o1 become apparent when faced with logical reasoning tasks.” The Mastermind challenge serves as an indicative test for applications in fields such as cybersecurity, fraud detection, and financial forecasting, areas where accuracy and efficiency are paramount.
As businesses increasingly rely on AI for decision-making and operational efficiency, the ability to handle logical reasoning tasks will be crucial. Genius’s performance suggests it is suited for tackling these complex challenges more effectively than current leading models.
The success of Genius in this benchmark is just the beginning. VERSES plans to further demonstrate its advanced reasoning capabilities with upcoming tests against other benchmarks like Atari’s 10k challenge. The company is committed to showcasing how Genius can continuously learn and adapt through experience, an essential feature that distinguishes it from static AI models.
VERSES AI has positioned itself as a formidable player in the cognitive computing space with its innovative Genius product. By outperforming established models like OpenAI’s o1 Preview in critical areas such as speed, cost, and reliability, VERSES is setting new standards for what AI can achieve in real-world applications. As businesses look for smarter solutions to complex problems, Genius appears to be a promising option that could redefine the capabilities of AI systems across various industries.
