OpenAI, the company that made ChatGPT and DALL-E 2, announced today that there will be a number of important changes. First, it is releasing developer APIs for ChatGPT and the Whisper speech-transcription model. It also changed its terms of service so that developers could choose not to have their data used to make the site better, and it added a rule that said data would be kept for 30 days.
The new ChatGPT API will use the same artificial intelligence model ("gpt-3.5-turbo") as the popular chatbot. This will let developers add ChatGPT to their apps either as-is or with their own twists. Snap's My AI is an early example, as is the online study tool Quizlet's new virtual tutor feature and the popular local shopping app Instacart's soon-to-come Ask Instacart feature. But the API won't just be for brand-specific bots that look like ChatGPT. It can also be used to power "non-chat" software that could use AI brains.
The price of the ChatGPT API is $0.002 per 1,000 tokens (about 750 words). It also has a dedicated-capacity option for developers with a lot of money who want to use more tokens than the standard API lets them. In February, ChatGPT Plus, a service for consumers that costs $20 per month, was launched. The new developer options are similar to that service.
The Whisper API from OpenAI is a hosted version of the Whisper speech-to-text model that was released in September. Greg Brockman, president and co-founder of OpenAI, told TechCrunch on Tuesday, "We put out a model, but that wasn't enough for the whole developer ecosystem to grow around it." "The Whisper API is the same big model you can get for free, but we've made it as efficient as possible. It's much, much quicker and very easy to use." Developers will have to pay $0.006 per minute to use the transcription API, which will allow "robust" transcription in multiple languages and translation to English.
Last but not least, OpenAI changed its developer terms after hearing from customers who were worried about privacy and security. If a developer doesn't opt in, the company will no longer use data sent through the API to train its AI models for "service improvements." It is also putting in place a 30-day data retention policy and giving users stricter options "depending on their needs" (likely meaning high-usage companies with budgets to match). Lastly, it is simplifying its rules about who owns the data and making it clear that users own the inputs and outputs of the models.
The company will also use a mostly automated system to replace its pre-launch review process for developers. OpenAI said that the change was necessary because "the vast majority of apps were approved during the vetting process" and that its monitoring has "improved significantly." "One of our main goals has been to figure out how to make things easy for developers." Brockman told TechCrunch about this. "Our goal is to really build a platform that other people can use to start their own businesses."
credit - engadget