GPT-5 may arrive this summer as a “materially better” update to ChatGPT

When OpenAI launched its GPT-4 AI model a year ago, it created a wave of massive hype and existential panic over its ability to mimic human communication and composition. Since then, the biggest question in AI has remained the same: When will GPT-5 come out? During interviews and media appearances around the world, OpenAI CEO Sam Altman is often asked this question, and he usually does Gives a shy or evasive answersometimes combined Promises Of amazing things to come.

According to a new report from Interested in tradeOpenAI is expected to release GPT-5, an improved version of the AI ​​language model that powers ChatGPT, sometime in mid-2024 — and likely during the summer. Two anonymous sources familiar with the company revealed that some enterprise customers have recently received demos of GPT-5 and related improvements to ChatGPT.

One CEO who recently saw a release of GPT-5 described the release as “really good” and “materially better,” as OpenAI demonstrated the new model using use cases and data unique to his company. The CEO also hinted at other unreleased capabilities for the model, such as the ability to launch AI agents being developed by OpenAI to perform tasks automatically.

We asked OpenAI representatives about the GPT-5 release date and Business Insider reported. They responded that they had no specific comment, but included an excerpt from the text of Altman's book Final appearance On Lex Friedman's podcast.

Lex Friedman (01:06:13) So when is GPT-5 coming out again?
Sam Altman (01:06:15) I don't know. This is the honest answer.
Lex Friedman (01:06:18) Oh, that's the honest answer. Blink twice if this is the year.
Sam Altman(01:06:30) We're launching a great new model this year. I don't know what we'll call it.
Lex Friedman (01:06:36) So this goes to the question, how do we edit this thing?
Sam Altman (01:06:41) We're going to be releasing a lot of different things in the coming months. I think this would be very cool. I think before we talk about a GPT-5-like model called that, or not called that, or a little worse or a little better than what you'd expect from GPT-5, I think we have a lot of other important things to get out there first.

In this conversation, Altman seemed to imply that the company is ready to launch a major AI model this year, but whether it will be called “GPT-5” or seen as a major upgrade to GPT-4 Turbo (or perhaps an incremental update) like GPT remains to be seen. -4.5) In the air.

Like its predecessor, GPT-5 (or whatever it will be called) is expected to be a Large Language Multimedia Model (LLM) that can accept encoded text or visual input (called a “prompt”). Like GPT-4, GPT-5 will be a next token prediction model, meaning it will produce the best estimate of the next most likely token (part of a word) in a sequence, allowing for tasks such as completing a sentence or writing code. When configured in a certain way, GPT templates can run chatbot applications like ChatGPT.

OpenAI launched GPT-4 in March 2023 as an upgrade to its major predecessor, GPT-3, which appeared in 2020 (with GPT-3.5 arriving in late 2022). Last November, OpenAI released GPT-4 Turbo, which dramatically reduced the inference (running) costs of OpenAI's best AI model, but was plagued by accusations of “laziness” as the model sometimes refused to answer prompts or complete programming projects as required. OpenAI has tried to fix the laziness issue several times.

Master's degree holders such as those developed by OpenAI are trained on large datasets pulled from the Internet and Licensed by media companies, enabling them to respond to user prompts in a human-like manner. However, the quality of information provided by the model can vary depending on the training data used, and also based on the model's tendency to aggregate information. If GPT-5 can improve generalizability (its ability to perform new tasks) while reducing what is commonly called “hallucinations” in the industry, it will likely represent significant progress for the company.

According to the report, OpenAI is still training GPT-5, and after that is complete, the model will undergo internal safety testing and further “red teaming” to identify and address any issues before its public release. The release date may be delayed depending on the duration of the safety testing process.

Of course, the sources cited in the report may be wrong, and GPT-5 may be released later for reasons other than testing. So, consider this a solid rumor, but this is the first time we've seen a potential release date for GPT-5 from a reputable source. Also, we now know that GPT-5 is complete enough to undergo testing, which means its main training run is likely complete. More improvements are likely to follow.