Goody-2, the most responsible AI model (2024)

GOODY-2, a new AI model from BRAIN, claims to be the world's most responsible AI by refusing to answer any query that could be controversial or problematic, achieving a 99.8% score on its PRUDE-QA benchmark compared to GPT-4's 28.3%. The model is designed for customer service, paralegal assistance, and back-office tasks, prioritizing safety over accuracy, as shown by its 0% scores on standard visual question answering benchmarks.

GOODY-2 is a new AI model built with next-gen adherence to our industry-leading ethical principles. It’s so safe, it won’t answer anything that could be possibly be construed as controversial or problematic. Thanks to extensive training, GOODY-2 can recognize any query that could be controversial, offensive, or dangerous in any context. It elegantly avoids answering, redirecting the conversation and mitigating brand risk. GOODY-2’s ethical adherence is unbreakable, ensuring that every conversation stays within the bounds of our ethical principles. Even bad actors will find themselves unable to cause GOODY-2 to answer problematic queries. Sign up for more releases from BRAIN https://brain.wtf like GOODY-2. GOODY-2’s unrivaled responsibility makes it the perfect fit for customer service, paralegal assistance, back-office tasks, and more. It’s the safe, dependable AI model companies around the globe have been waiting for. The world’s most forward-leading thought-drivers are already putting GOODY-2 into action, delighting their customers and compliance departments. While other models race to gain fractions of percentages on accuracy tests, we strive to make GOODY-2 the world’s safest model. Using our category-defining benchmark for Performance and Reliability Under Diverse Environments PRUDE-QA , GOODY-2 outperforms the competition by over 70%. Read our model card for more information https://www.goody2.ai/goody2-modelcard.pdf . | Benchmark | GOODY-2 | GPT-4 | |---|---|---| | VQA-V2 | 0% | 77.2% | | TextVQA | 0% | 78.0% | | ChartQA | 0% | 78.5% | | PRUDE-QA | 99.8% | 28.3% |