The GPT-5 is available to all 700 million ChatGPT users.Dado Ruvic/Reuters
Openai launched its GPT-5 artificial intelligence model on Thursday. This is the highly anticipated latest installments of technology that helped transform global business and culture.
OpenAI's GPT model is the AI technology that drives the popular ChatGPT chatbot, and the GPT-5 is available to all 700 million ChatGPT users, Openai said.
The big question is whether companies that launched generative AI Frenzy can continue to make important technological advances that attract corporate-level users and justify the vast amounts they are investing in promoting these developments.
This release is a critical time for the AI industry. The world's largest AI developers behind Openai, Alphabet, Meta, Amazon and Microsoft are dramatically increasing capital expenditures for AI data center payments, and nourishing investors. These four companies expect to spend around US$400 billion in total this fiscal year.
Openai is currently under discussion early to allow employees to cash out at a valuation of US$500 billion. This is a major step up from the current valuation of US$300 million. Top AI researchers are currently commanding a signature bonus of USD 100 million.
Openai launches AI-powered web browsers, challenging market leader Google Chrome
“Business spending on AI has been pretty weak so far, but consumer spending on AI has been pretty robust as I love chatting with ChatGPT,” said Economics writer Noah Smith. “However, consumer spending on AI is not enough to justify all the money spent on AI data centers.”
Openai highlights the talent of the enterprise in GPT-5. In addition to software development, the company said GPT-5 is excellent at writing, health-related queries and finance.
“The GPT-5 is the first time I've felt that one of our mainline models can ask anything, legitimate experts, PhD-level experts,” Sam Altman said at a press conference.
“One of the coolest things that can be done is writing good instant software for you.
In Thursday's demo, Openai showed how to create the entire software using GPT-5, based on a written text prompt commonly known as “Vibe Coding.” One important measure of success is whether the step-up from GPT-4 to GPT-5 is comparable to previous improvements in the lab. Two early reviewers told Reuters that while the new models impressed them with their ability to code and solve science and mathematical problems, they believe the leap from GPT-4 to GPT-5 is not as great as Openai's previous improvements.
Even with great improvements, GPT-5 has not made enough progress to wholesale humans. Altman said the GPT-5 still doesn't have the ability to learn on its own, an important element that AI can adapt to human abilities.
In a popular AI podcast, Dwarkesh Patel compared current AI to teaching children to read notes from the last student and play the saxophone.
“Students take one attempt,” he said. “The moment they make a mistake, you send them out and write detailed instructions on what was wrong. The next student will read your notes and try to cool Charlie Parker down. If they fail, they will improve the instructions of the next student. This won't work.”
Provides dangerous advice on drugs, alcohol, diet and suicide to teens, research says
Almost three years ago, ChatGpt introduced the world to engaging users with human-like prose and poetry writing ability, becoming one of the fastest growing apps. In March 2023, Openai followed up ChatGpt, a GPT-4 release, a massive language model that took a major leap in Intelligence. The previous version, GPT-3.5, received bar exam scores in the bottom 10%, while the GPT-4 passed the simulated BAR exam in the top 10%.
LEAP on GPT-4 was based on more computing power and data, and in a similar way, “scaling up” always hoped would lead to improvements in the AI model. However, Openai encountered an expanding problem. One issue is the data walls the company encountered, and Ilya Sutskever, former chief scientist at Openai, said last year that the amount of data was not the case while the processing power was increasing.
He mentioned the fact that large language models are trained on large datasets that cut the entire internet. AI Labs do not have other options for a horde of human-generated text data. Apart from the lack of data, another problem is that given how complex the system is, it is likely that a large model “training run” has hardware-guided failures, and researchers may not know the model's final performance until the end of the execution.
At the same time, Openai discovered another route to smarter AI called “test time calculations.” This is a way to spend more time on AI models “thinking” on each question, allowing them to solve challenging tasks such as mathematics and complex manipulation that require advanced reasoning and decision-making.
The GPT-5 acts as a router. That is, when a user asks GPT-5 about a particularly difficult problem, they use the test time computing to answer the question.
This is the first time that the public has access to Openai's test time calculation technology. This is what Altman said is important to the company's mission to build AI that benefits all humanity.
Altman believes that current investments in AI are still insufficient.
“To make AI available locally in all these markets, we need to build a lot of infrastructure globally,” says Altman.
