ChatGPT now has 200 million users, GPT-4o with new features
It’s hardly surprising that ChatGPT has managed to double its weekly active users since last November. The hype around AI hasn’t stopped since last year, and ChatGPT remains the most well-known generative AI chatbot. The OpenAI service now sees 200 million active users each week, up from 100 million in 2023. OpenAI also says that 92% of Fortune 500 companies use its products. Usage of its automated API has also doubled since the release of GPT-4o Mini last month.
In a statement, OpenAI CEO Sam Altman said:
“People are using our tools now as a part of their daily lives, making a real difference in areas like healthcare and education—whether it’s helping with routine tasks, solving hard problems, or unlocking creativity.”
ChatGPT-4o can now be fine-tuned
However, ChatGPT is trained across the internet, so it may not have the tone and style of output for your project. You can try creating a detailed prompt to achieve that style, but you can now fine-tune the model. Fine-tuning is the final touch of an AI model. It happens after the bulk of the training is complete, but it can have a strong impact on the output with relatively little effort. OpenAI says that just a few dozen examples are enough to change the tone of the output to better suit your use case.
For example, if you’re trying to build a chatbot, you can write several question-answer pairs and feed them into GPT-4o. Once fine-tuning is complete, the AI’s answers will be closer to the examples you gave it. OpenAI is letting you use 1 million training tokens for free until September 23rd. After that, fine-tuning costs $25 per 1 million tokens and using the tuned model costs $3.75 per 1 million input tokens and $15 per 1 million output tokens. OpenAI has detailed and accessible documentation on fine-tuning.
The company has worked with partners to try out the new features. Cosine has an AI called Genie that can help users find bugs and offers the fine-tuning option. Cosine trained it on real-world examples. Then there is Distyl, which tried to optimize a text-to-SQL model (SQL is a language for querying databases). It ranked first in the BIRD SQL benchmark with an accuracy of 71.83%. For comparison, human developers (data engineers and students) achieved an accuracy of 92.96% in the same test.
You may be concerned about privacy, but OpenAI says users who fine-tune 4o have full ownership of business data, including all inputs and outputs. The data you use to train the model is never shared with anyone else or used to train other models. However, OpenAI also monitors for abuse in case someone tries to tweak a model that violates its usage policies.
