OpenAI CEO Sam Altman says ChatGPT's GPT-4 'kinda sucks'

By
Web Desk
Sam Altman speaks at the Wall Street Journal Digital Conference in Laguna Beach, California, U.S., October 18, 2017. —Reuters
Sam Altman speaks at the Wall Street Journal Digital Conference in Laguna Beach, California, U.S., October 18, 2017. —Reuters

OpenAI CEO Sam Altman spoke very frankly about the shortcomings of GPT-4, their newest model, during his dialogue with Lex Fridman, Gizmodo reported. 

Altman said GPT-4 “kinda sucks” adding that that they need to keep on improving themselves to meet future expectations. Although Fridman first recommended GPT-4, all his attention was shifted in the direction of GPT-5 which he predicted to make a much bigger leap forward in AI capabilities.

Acknowledging the achievement of GPT-4 as well as calibrating expectations, Altman stated the continuous growth in the development of AI. He underlined the necessity of keeping mind the future advancements, while being able to critically analyze current tools' capabilities.

Altman's remarks bear out the common feeling of AI burnout existing where the promises of revolutionary technology are far from the reality. He explained that GPT-4's main thing was in brainstorming and gave an example of the problem that is too complex to be solved step by step.

The OpenAI community is riddled with rumors about future releases such as GPT-5, GPT 4.5 and GPT-4.5 Turbo. Yet, Altman implied a state of fuzziness, meaning that no deadline is certain. On the other hand, Sora, OpenAI's text-to-video generator, is highly anticipated with claims of realistic outputs.

In spite of the shortcomings of GPT-4, OpenAI does not stop succeeding with 100 million users per week and $80 billion valuation. Altman is determined and is looking forward to the arrival of GPT-5 which will accomplish the dream of AI supporters.