OpenAI is making its ChatGPT API and Whisper API available to companies so they can incorporate the AI chatbot and speech-to-text capabilities into their products – Instacart, Snapchat and Spotify are all early adopters.
ChatGPT API terms and conditions meanwhile emphasise that users can't deploy the ChatGPT API to power adult entertainment, generation of malware, nor indeed "high risk government decision-making"...
The company (which hit 100 million monthly active users in January just two months after launch, making it the fastest-growing consumer application in history, according to a UBS study) is moving swiftly to try to monetise this huge user base as it faces huge computational bills for running its substantial NLP models.
The ChatGPT API will be priced at $0.002 per 1000 tokens. (OpenAI has a tool that estimates how a piece of text would be tokenized by the API, and the total count of tokens in that piece of text available here.)
The company is also making dedicated compute instances available for users who want “deeper control over the specific model version and system performance” saying requests by default run on shared Azure servers and “with dedicated instances, developers will pay by time period” for reserved single tenant servers.
And it has promised to stop using data submitted to ChatGPT for service improvements (including model training) unless organisations specifically opt in. (Automated systems using the API must also disclose to users that they are interacting with an AI system and products that simulate another person must either have that person's explicit consent or be clearly labeled as "simulated" or "parody" OpenAI's terms emphasise...)
OpenAI API users wanting dedicated instances will get “full control over the instance’s load (higher load improves throughput but makes each request slower), the option to enable features such as longer context limits, and the ability to pin the model snapshot” OpenAI said, adding that this may “make economic sense for developers running beyond ~450M tokens per day [and] enables directly optimizing a developer’s workload against hardware performance, which can dramatically reduce costs relative to shared infrastructure.”
Most users who have experimented with ChatGPT will have often found it at capacity.
OpenAI admitted it was working to improve this even as it looked to onboard commercial clients, saying “For the past two months our uptime has not met our own expectations nor that of our users. Our engineering team’s top priority is now stability of production use cases—we know that ensuring AI benefits all of humanity requires being a reliable service provider. Please hold us accountable for improved uptime…”