Source: Giktime
It took a while, but it’s here: OpenAI, the company behind ChatGPT and DALL-E 2, announced some interesting and significant updates, including: the launch of developer APIs for ChatGPT and its speech transcription model Whisper, which will allow developers to integrate these impressive capabilities into their applications , which have become mega-popular in recent months. The new ChatGPT API, by the way, will use gpt-3.5-turbo, the same model on which the popular chatbot is based, so that the capabilities, which will be implemented from now on in the various applications, are expected to be the same as the results you get in the service itself.
The API price of ChatGPT is $0.002 per 1,000 tokens (something like 750 words) and the company explains that it is a tenth of the cost of the existing plans, but if you need tokens wholesale – but really wholesale – say in the region of 450 million per day – you can purchase Dedicated instances, that is Azure computing resources, the cloud infrastructure on which OpenAI’s various products are built. Microsoft has invested billions of dollars in OpenAI.
The first to use the new API interface are Snap with its new My AI feature; The American EdTech company Quizlet, which offers quizzes and games for learning; and Instacart, the American sales site that recently introduced a tool called Ask Instacart.
Apart from the chat models, the company also released an API for Whisper, the company’s open source transcription model, which it launched in September – a model that we will mention also supports Hebrew. The company explains that for the purpose of releasing the API, the model was optimized to make it faster and more convenient. Using the Transcription API will cost developers $0.006 per minute.
The new API interfaces join ChatGPT Plus, the company’s premium service launched in Israel in February, which is aimed at private consumers and offers several advantages for $20 per month.
Integration in Python:
import openai
completion = openai.ChatCompletion.create(
model="gpt-3.5-turbo",
messages=[{"role": "user", "content": "Tell the world about the ChatGPT API in the style of a pirate."}]
)
print(completion)Upgrades in privacy
OpenAI explained that it was important for them to be “developer friendly” and therefore, in addition to the launch of the API, the company announced some additional changes that are relevant to the development processes. First, the approval process for developers, which until now was manual, has been replaced by a largely automated system. Second, the company changed its terms of use for developers, who use its various products. Among other things, the company announced that it will no longer use the data sent through the API for the purpose of “improving the service” or in other words to train its AI models. Those who would still like to allow the company to train their models on their queries and data will be asked to proactively join the program (Opt-In). In addition, the company announced a change in the data retention policy, which will be limited to 30 days while providing stricter retention options “according to the user’s needs.” A recent update simplified the terms around data ownership, making it clear that users own the model outputs.
All the technical details and documentation await you here and here