Today, OpenAI, the startup behind ChatGPT and DALL-E 2, announced a number of important modifications. Initially, it will make developer APIs available for ChatGPT and the Whisper speech-transcription paradigm. It also modified its terms of service to allow developers to opt out of having their data used for improvement purposes, as well as introducing a 30-day data retention policy.
The new ChatGPT API will make use of the same artificial intelligence (AI) model (gpt-3.5-turbo) as the well-known chatbot, enabling developers to integrate plain or customised versions of ChatGPT into their applications. Snap’s My AI, as well as a new virtual instructor function for the online study tool Quizlet and an impending Ask Instacart tool in the popular local-shopping app, are early examples. The API, however, will not be confined to brand-specific bots that resemble ChatGPT; it may also power “non-chat” software experiences that may benefit from AI brains.
The ChatGPT API costs $0.002 per 1,000 tokens (about 750 words). Furthermore, it is providing a dedicated-capacity alternative for wealthy developers that anticipate using more tokens than the normal API enables. The new developer choices complement ChatGPT Plus, a $20-per-month subscription that launched in February.
The open-source Whisper speech-to-text model that OpenAI debuted in September is now available as part of its hosted Whisper API. “We provided a model, but it was not enough to compel the entire developer ecosystem to build around it,” said OpenAI president and co-founder Greg Brockman on Tuesday. “The Whisper API is the same huge model that you can obtain open source, but we’ve optimised to the utmost. It’s significantly faster and more convenient.” The transcription API will cost developers $0.006 per minute and will enable “robust” transcription in many languages as well as English translation.
Lastly, OpenAI updated its developer agreements in response to user comments concerning privacy and security issues. Unless a developer opts in, the firm will no longer utilise API data for “service enhancements” to train AI models. Moreover, it is implementing a 30-day data retention policy while also offering harsher retention alternatives “based on customer needs” (likely meaning high-usage companies with budgets to match). Lastly, it is explaining that users own the input and output of the models by reducing its data ownership conditions.
In addition, the corporation will replace its developer pre-launch approval process with a mostly automated approach. OpenAI explained the adjustment by noting that the “overwhelming majority of apps were approved during the vetting process,” and that its monitoring has “significantly improved.” “One of our biggest focuses has been figuring out, how do we become super friendly to developers?” Brockman told TechCrunch. “Our mission is to really build a platform that others are able to build businesses on top of.”
