Consultants calling for a pause on AI growth will probably be glad to listen to that OpenAI isn’t at present coaching GPT-5.
OpenAI CEO Sam Altman spoke remotely at an MIT occasion and was quizzed about AI by laptop scientist and podcaster Lex Fridman.
Altman confirmed that OpenAI shouldn’t be at present creating a fifth model of its Generative Pre-trained Transformer mannequin and is as a substitute specializing in enhancing the capabilities of GPT-4, the newest model.
Altman was requested concerning the open letter that urged builders to pause coaching AI fashions bigger than GPT-4 for six months. Whereas he supported the concept of guaranteeing AI fashions are secure and aligned with human values, he believed that the letter lacked technical nuance concerning the place to pause.
“An earlier model of the letter claims we’re coaching GPT-5 proper now. We aren’t, and gained’t for a while. So in that sense, it was kind of foolish,” mentioned Altman.
“We’re doing issues on prime of GPT-4 that I feel have all kinds of issues of safety that we have to tackle.”
GPT-4 is a big enchancment over its predecessor, GPT-3, which was launched in 2020.
GPT-3 has 175 billion parameters, making it one of many largest language fashions in existence. OpenAI has not confirmed GPT-4’s actual variety of parameters however it’s estimated to be within the area of 1 trillion.
OpenAI mentioned in a blog post that GPT-4 is “extra inventive and collaborative than ever earlier than” and “can resolve troublesome issues with better accuracy, because of its broader basic information and problem-solving skills.”
In a simulated regulation bar examination, GPT-3.5 scored across the backside 10 p.c. GPT-4, nonetheless, handed the examination among the many prime 10 p.c.
OpenAI is likely one of the main AI analysis labs on the planet, and its GPT fashions have been used for a variety of functions, together with language translation, chatbots, and content material creation. Nevertheless, the event of such massive language fashions has raised issues about their security and moral implications.
Altman’s feedback counsel that OpenAI is conscious of the issues surrounding its GPT fashions and is taking steps to deal with them.
Whereas GPT-5 will not be on the horizon, the continued growth of GPT-4 and the creation of different fashions on prime of it can undoubtedly increase additional questions concerning the security and moral implications of such AI fashions.
(Photograph by Victor Freitas on Unsplash)
Associated: Italy will lift ChatGPT ban if OpenAI fixes privacy issues
Need to be taught extra about AI and massive knowledge from business leaders? Take a look at AI & Big Data Expo happening in Amsterdam, California, and London. The occasion is co-located with Digital Transformation Week.
Discover different upcoming enterprise know-how occasions and webinars powered by TechForge here.
