Sam Altman hints at GPT-5, says GPT-4 ‘sucks’
To succeed in life, one must be honest about their flaws yet aim to improve. This characteristic is especially true for leaders, and fortunately, Sam Altman seems to exhibit these qualities.
OpenAI’s CEO hinted that his company is working on a new AI model, which likely means he’s working on GPT-5, the next large language model. Yet, he admits that GPT-4, the latest ChatGPT LLM, “kinda sucks.”
READ: OpenAI files trademark for GPT-5
What did Altman say about GPT-4 and GPT-5?
Here's my conversation with Sam Altman (@sama), his 2nd time on the podcast. We talk about the board saga, Elon lawsuit, Ilya, Sora, GPT-5, $7 trillion in compute, open source, and AGI. This was a truly fascinating conversation.
— Lex Fridman (@lexfridman) March 18, 2024
It's here on X in full, and is up on YouTube,… pic.twitter.com/h2ANXZpcrz
Business news website Inc. reported on Sam Altman’s remarks during a recent episode of The Lex Friedman Podcast. He and the host discussed generative AI following ChatGPT’s success.
Friedman called GPT-5 “historically impressive” and “amazing.” Also, he described the evolution of artificial intelligence as “a historic, pivotal moment” in the world.
In response, Altman surprisingly said, “I think it kinda sucks.” He compared AI innovation to how some people might reminisce on past versions of the iPhone and see they’re obsolete compared to current models.
“I think it is an amazing thing,” he said, giving OpenAI credit for the free version of ChatGPT running GPT-3. Nowadays, GPT-4 offers better features that cost $20 monthly.
“At the time of GPT-3, people were like ‘this is amazing, this is [a] marvel of technology,’… and it was. But now we have GPT-4, and you look at GPT-3 and you’re like ‘that’s unimaginably horrible,’ ” Altman explained.
Then, he addressed the next version of ChatGPT’s LLM, which would likely be called GPT-5. “I expect the delta between 5 and 4 will be the same as between 4 and 3,” the CEO said.
“It’s our job to live in the future and remember that our tools are going to kind of suck looking back at them,” he added.
Techradar reported another of Altman’s statements in the podcast. He said that OpenAI excels at multiplying “200 medium-sized things together into one giant thing.”
Altman specified to Lex Friedman that this trait is true “especially on the technical side.” Still, the tech innovator says he’s excited for the next GPT version to “be smarter.”
Friedman asked Altman to “blink twice” if GPT-5 would come out this year, but Altman didn’t. Instead, he said OpenAI will release a new, different model.
