Sam Altman, the founder of OpenAI, has debunked rumours of the company already working on ChatGPT-5, just a month after the release of GPT-4. In an interview with AI researcher Lex Fridman at MIT, Altman addressed concerns regarding the open letter demanding an AI pause and shared his critiques.
Altman clarified that OpenAI is not training GPT-5 at the moment and is instead focusing on updates and upgrades to GPT-4. However, he also admitted that there are safety issues with some of the tweaks being shipped, such as the plug-in allowing GPT-4 to browse the internet, potentially leading to data privacy and user manipulation concerns.
Altman attempted to reassure AI sceptics by stating that OpenAI spent over six months training GPT-4 and that the model’s safety is being studied. In addition, he emphasized the need for caution and increasing rigour for safety issues as capabilities become more serious. However, his claims regarding OpenAI’s transparency were met with scepticism, as the company has become increasingly closed-off since its transformation into a for-profit corporation.
Altman’s response to the open letter demanding an AI pause was ultimately deemed a non-answer, as he did not address the specific concerns raised in the letter.
While he clarified that OpenAI is not currently training GPT-5, there are still potential safety issues with the tweaks being shipped for GPT-4. The need for caution and increasing rigour for safety issues in AI was emphasized, but his claims regarding OpenAI’s transparency were met with scepticism.