OpenAI is reportedly making more money from its chatbots than Microsoft. While Microsoft initially partnered with OpenAI to sell access to their models together, the terms changed when Microsoft invested more money into OpenAI. OpenAI reached $1 billion in annualized revenue from selling model access, surpassing Microsoft’s Azure OpenAI Service. Microsoft has cut prices for Azure OpenAI Service to compete with OpenAI. However, it needs to be clarified how much Microsoft earns from Copilot and how much OpenAI earns from ChatGPT for enterprises.
OpenAI has developed CriticGPT, a tool designed to help human trainers identify ChatGPT code output errors. This tool aims to improve OpenAI’s AI models by providing trainers with an AI assistant to spot inaccuracies and write more comprehensive critiques. However, CriticGPT has limitations, including being trained on short ChatGPT answers and the potential for generative AI to cause hallucinations and lead to mistakes. OpenAI acknowledges that evaluating extremely complex tasks or responses may still be challenging even with model assistance.
OpenAI has announced a delay in launching its “advanced Voice Mode” feature for ChatGPT voice assistants. The company needs more time to improve content detection, user experience, and infrastructure scalability. A small group of users will test the feature before being launched to all paying ChatGPT customers in the fall. The release of new voices and a camera understanding feature has also been delayed.
Google Translate is expanding its language support by adding 110 new languages, making it the biggest expansion in the company’s history. The addition of languages was made possible by Google’s PaLM 2 AI language model, which excelled in learning related languages. The newly supported languages include Cantonese, which has been highly requested, and a significant number of languages from Africa. Most of the new languages are spoken by at least one million people, with some spoken by hundreds of millions.
Why do we care?
OpenAI is the rocketship of an energized startup. Microsoft will be the steady, wise investment over time. You can pick your risk tolerance level – both will do well.
But let’s focus for a moment on that Google Translate enhancement. AI did that. And that’s the trick – we’re looking for outcomes less than the technologies themselves. I’m far more interested in CoPilot’s success than the specific LLM because that’s where the benefits to the business and the employees are. We’re tracking earlier to gauge what is coming, such as CriticGPT enhancements to CoPilot. But the value is in the use cases.