Future of ChatGPT

A "Weird" Name and a Scrappy Start

ChatGPT was effectively born in December 2021 with an OpenAI project dubbed WebGPT: an AI tool that could search the internet and write answers. The team took inspiration from WebGPT’s conversational interface and began plugging a similar interface into GPT-3.5, a successor to the GPT-3 text model released in 2020. They gave it the clunky name "Chat with GPT-3.5" until, in what Turley recalls as a split-second decision, they simplified it to ChatGPT.

The name could have been the even more straightforward "Chat," and in retrospect, he thinks perhaps it should have been. "The entire world got used to this odd, weird name, we’re probably stuck with it. But obviously, knowing what I know now, I wish we picked a slightly easier to pronounce name," he says. (It was recently revealed that OpenAI purchased the domain chat.com for more than $10 million of cash and stock in mid-2023.)

Scaling Up

Only days after Turley tells me this, ChatGPT did get a new $200 price tag for a pro tier that includes access to a specialized reasoning model. Its main $20 Plus tier is sticking around, but it’s clearly not the ceiling for what OpenAI thinks people will pay.

ChatGPT and other OpenAI services require vast amounts of computing power and data storage to keep its services running smoothly. On top of the user base OpenAI has gained through its own products, it’s poised to reach millions more people through an Apple partnership that integrates ChatGPT with iOS and macOS.

The Future of ChatGPT

For Turley and others, this is where the stakes will get particularly steep. Agents could make AI far more useful by moving what it can do outside the chatbot interface. The shift could also grant these tools an alarming level of access to the rest of your digital life.

"I’m really excited to see where things go in a more agentic direction with AI," Kim tells me. "Right now, you go to the model with your question, but I’m excited to see the model more integrated into your life and doing things proactively, and taking actions on your behalf."

The Goal of ChatGPT

The goal of ChatGPT isn’t to be just a chatbot, says Fedus. As it exists today, ChatGPT is "pretty constrained" by its interface and compute. He says the goal is to create an entity that you can talk to, call, and trust to work for you. Fedus thinks systems like OpenAI’s "reasoning" line of models, which create a trail of checkable steps explaining their logic, could make it more reliable for these kinds of tasks.

The Road Ahead

Turley says that, contrary to some reports, "I don’t think there’s going to be such a thing as an OpenAI agent." What you will see is "increasingly agentic functionality inside of ChatGPT," though. "Our focus is going to be to release this stuff as gradually as possible. The last thing I want is a big bang release where this stuff can suddenly go out and do things over hours of time with all your stuff."

Conclusion

By ChatGPT’s third anniversary next year, OpenAI will probably look a lot different than it does today. The company will likely raise billions more dollars in 2025, release its next big "Orion" model, face growing competition, and have to navigate the complexity of a new US president and his AI czar.

Frequently Asked Questions

Q: What is the goal of ChatGPT?
A: The goal of ChatGPT isn’t to be just a chatbot, but to create an entity that you can talk to, call, and trust to work for you.

Q: What is the future of ChatGPT?
A: The future of ChatGPT is to be an entity that can perform complex, multistep tasks autonomously, such as making phone calls and placing food orders.

Q: Will OpenAI create an "agent"?
A: No, OpenAI does not plan to create an "agent." Instead, it will release increasingly agentic functionality inside of ChatGPT.

Q: What is the timeline for ChatGPT’s development?
A: By 2024, OpenAI will likely raise billions more dollars, release its next big "Orion" model, and face growing competition.