AI Made Friendly HERE

ChatGPT’s Knowledge Base Finally Extends Beyond 2021

OpenAI is updating ChatGPT’s knowledge base so that it can finally respond about things that occurred after September 2021.

The change is part of a slew of improvements OpenAI is making to its AI program, including the ability for anyone—including people who can’t program—to create custom ChatGPT models. 

During its first annual developer day, OpenAI CEO Sam Altman introduced GPT-4 Turbo, a new large language model for ChatGPT, which currently runs on GPT-3.5 or GPT-4 models.

Sam Altman

(Credit: OpenAI)

GPT-4 Turbo is designed to address “many of the things” that developers and users have been asking for, Altman says, including more up-to-date information.

“We will try to never let it get that out-of-date again,” according to Altman, who says the GPT-4 Turbo model updates the chatbot’s knowledge base up to April 2023. “And we will continue to improve that over time,” he adds.

To further improve ChatGPT’s ability to talk about current topics, the program is adding a new “retrieval function” that’ll let users upload documents or databases to the chatbot. 

Another notable change is how GPT-4 Turbo supports up to “128,000 tokens” of input. This means a user can submit about 300 pages of text to ChatGPT, and it’ll be able to come up with the appropriate response. Previously, GPT-4 was restricted to text inputs at 8,000 tokens to 32,000 tokens in length (or about 5,800 to 23,400 words per input). 

For developers, the company is also lowering the API price to build apps. GPT-4 Turbo promises to be about 2.75 times cheaper than GPT-4, or $0.01 per 1,000 input tokens and $0.03 per 1,000 output tokens. 

Sam Altman

(Credit: OpenAI)

For now, the company is rolling out the GPT-4 Turbo model to developers as a preview through the company’s API. It’s also unclear if Turbo will be restricted to ChatGPT Plus users. But the company’s blog post says a stable production-ready model will arrive in the “coming weeks,” presumably through the ChatGPT interface. 

Creating Custom Chatbots

During the developer conference, OpenAI also previewed the future of the company’s technology. Altman views chatbots as one day becoming smart enough to act as “agents” that can perform any task a human requests.

“We know that people want AI that is smarter, more personal, more customizable, can do more on your behalf,” Altman said. “Eventually, you’ll just ask a computer for what you need and it’ll do all of these tasks for you.”

To reach that goal, OpenAI is focusing on letting anyone create custom chatbot models, known as “GPTs” that can go beyond the standard ChatGPT program for bots dedicated to a specific area or task. For example, you could create a GPT focused on teaching computer science in schools or a GPT devoted to letting consumers create marketing logos and posters using a text prompt. The GPTs can be used privately, shared publicly, or exclusively deployed at a company. 

GPT builder

(Credit: OpenAI)

The bot customization will arrive through a new GPT builder app. The effort builds on the plugin function the company introduced for ChatGPT in March. OpenAI’s goal is to offer the GPTs on a digital store, and share revenue with the developers who make them. The so-called “GPT Store” is set to launch later this month. Meanwhile, users can start creating their own GPTs today, a company blog post says.  

Sam Altman

(Credit: OpenAI)

In a demo, Altman also showed that a user can build a custom GPT without any programming skills. The user merely has to type out the requirements for the bot and GPT will do its best to comply. “You can in effect program a GPT with language just by talking to it,” he said. 

Altman indicated he views GPTs as a stepping stone to even more powerful AI programs that the company is already working on. “Over time, GPTs and assistants, our precursors to agents, are going to be able to do much, much more,” Altman said. “They’ll gradually be able to plan and to perform more complex actions on your behalf.” 

“We really believe in the importance of gradual, iterative deployment,” he added. “We believe it’s important for people to start building with and using these agents now to get a feel for what the world is going to be like as they become more capable.”

Get Our Best Stories!

Sign up for What’s New Now to get our top stories delivered to your inbox every morning.

This newsletter may contain advertising, deals, or affiliate links. Subscribing to a newsletter indicates your consent to our Terms of Use and Privacy Policy. You may unsubscribe from the newsletters at any time.

Originally Appeared Here

You May Also Like

About the Author:

Early Bird