OpenAI Announces GPT-4 AI Language Model

Facts

  • OpenAI has announced the release of GPT-4, the latest product in the line of AI language models that already powers the likes of ChatGPT and Bing's chatbox.1
  • GPT-4 became available on Wed. to OpenAI's paying users via ChatGPT Plus, and developers can sign up on a waitlist to access the API. The AI is able to generate text and accept image and text inputs, in contrast to its predecessor (GPT-3.5) which only accepted text.2
  • OpenAI's CFO, Sam Altman, described GPT-4 as a 'more creative' and 'less biased' than the previous model. Altman also claimed that the new AI was capable of passing the legal bar exam and that it could 'score a 5' on several AP exams.3
  • OpenAI stated that GPT-4 was also 82% less likely to respond to requests for disallowed content and scores 40% higher on certain tests of factuality.4
  • Despite this, Altman measured expectations by claiming that GPT-4 was still 'flawed' and 'limited'. The company warned that individuals should still take care when using the AI, especially in 'high-stakes contexts.'5
  • Microsoft has currently invested $10bn USD into OpenAI. ChatGPT first launced in November 2022 and has been used by millions of people. Accessing ChatGPT Plus currently cost $20 per month.6

Sources: 1Verge, 2TechCrunch, 3Business Insider, 4Reuters, 5Independent and 6BBC News.

Narratives

  • Narrative A, as provided by Pcworld. GPT-4 is a profound improvement from its predecessor and potentially opens the door to major advances in AI technology. The model has a longer output, is simply smarter and safer than OpenAI's previous models. While a lot will need to be proven over time, it seems that ChatGPT and Bing have just received a major upgrade.
  • Narrative B, as provided by Wired. While GPT-4 scores highly on a range of tests and OpenAI has forced many to rethink the potential of artificial intelligence, the language model still inherits the faults of its predecessors. The tendency to 'hallucinate' incorrect information and exhibit social biases remain. Despite catching the public's attention, there is still a long way to go.