OpenAI to give users more control over generative AI system

"We'll launch more things soon that give users additional control on the system to behave this way or that way," says CEO

By
Reuters
|
A response by ChatGPT, an AI chatbot developed by OpenAI, is seen on its website in this illustration picture taken February 9, 2023. — Reuters
A response by ChatGPT, an AI chatbot developed by OpenAI, is seen on its website in this illustration picture taken February 9, 2023. — Reuters 

OpenAI, the creator of the buzzy chatbot ChatGPT, will release tools to give users more control over the generative AI system, while improving the models for both general and specific use cases, its CEO Sam Altman said Thursday.

Speaking to investors at a Morgan Stanley conference, Altman said the AI company will focus on building a platform that sells APIs to others and creates killer apps like ChatGPT.

Since its launch in November, ChatGPT's popularity has surged as traffic to the site hit more than 1 billion visits, up from 616 million in January, according to Similarweb estimates. OpenAI has launched a subscription tier of ChatGPT where users can pay $20 per month for more reliable services.

The Microsoft-backed company is working with enterprise clients to train its models in particular domains and has effectively reduced hallucinations, and incidents when an AI system confidently gives a response that is factually incorrect, according to Altman.

Management consultancy Bain & Company, has struck a global services partnership with OpenAI, enabling Bain to embed AI in its client operations.

Enterprises that work with OpenAI can use their data and make a copy of the model to alleviate data safety concerns. Coca-Cola, for example, is working with OpenAI and Bain to use OpenAI's ChatGPT and DALL-E platforms to create personalised ad copy, images, and messaging.

Altman, a veteran entrepreneur and investor, said the company should be valued by investors as a firm to achieve the goal of general artificial intelligence.

Individual users should also have more control over how the AI works, Altman added. The company said last month it is developing an upgrade to its chatbot that users can customise to address concerns about bias in artificial intelligence.

"We'll launch more things soon that give users additional control on the system to behave this way or that way."

Altman acknowledges the AI system cannot achieve 100% accuracy, and he said he expects applications including AI doctors and AI lawyers to emerge on people's phones soon.