While specialty AI models like OpenAI's Whisper ASR speech recognition and translation target a more technical and niche audience, they offer a powerful lens through which to understand the pitfalls and promise of this new generation of AI tools. As we look to the confidence and hallucination issues of more popularly known tools like ChatGPT, the more limited and specific domain of specialized tools like Whisper allow us to better understand the contours of these challenges and the possibilities of emergent mitigation strategies.
We're very interested in talking to AI researchers interested in collaborating with us around some of the field's greatest challenges, from globalization and debiasing to the creation of new training and testing datasets that better reflect the real world and new evaluation metrics and benchmarks that better reflect the actual downstream applications models are deployed into.
Please reach out if you are an AI researcher with an interest in these topics – we'd love to talk more.