Summary

  • GPT-4 is set to be replaced by GPT-4o in ChatGPT at the end of April, marking the end of an era for AI language models.
  • When it launched in March 2023, GPT-4’s capabilities caused widespread excitement and alarm, with some observers concerned about the implications of increasingly sophisticated AI.
  • GPT-4 combined data from a range of human knowledge sources into a set of neural networks with a combined total of 1.76 trillion parameters.
  • Despite OpenAI CEO Sam Altman estimating that it cost over $100 million to train, and required over 20,000 high-end GPUs to develop, the model’s impact began before the official launch.
  • Microsoft integrated an early version of the model into its Bing search engine, creating a chatbot that drew controversy over its responses to some media articles.
  • Following its retirement from ChatGPT, GPT-4 will still be available via API for developers.
  • The replacement, GPT-4o, is purportedly more accurate and refined, providing responses that are fresher and based on more up-to-date information.

By Benj Edwards

Original Article