GPT-4: Impressive but Imperfect.

OpenAI's GPT-4, the latest text-generating AI model, has improved upon its predecessor, GPT-3, in key ways, including multimodality, allowing it to understand images and perform actions. However, GPT-4 still has serious shortcomings, such as making basic reasoning errors and "hallucinating" facts. OpenAI's co-founder and president, Greg Brockman, acknowledges these limitations but emphasizes the model's new mitigatory steerability tools, including an API-level capability called "system" messages, which act as guardrails to prevent GPT-4 from veering off course. OpenAI is also testing a version of GPT-4 that can "remember" roughly 50 pages of content, which Brockman believes will lead to new, previously unexplored applications, particularly in the enterprise.
- Interview with OpenAI's Greg Brockman: GPT-4 isn't perfect, but neither are you TechCrunch
- GPT-4 has arrived. It will blow ChatGPT out of the water. The Washington Post
- OpenAI Unveils GPT-4, Months After ChatGPT Stunned Silicon Valley The New York Times
- 10 Ways GPT-4 Is Impressive but Still Flawed The New York Times
- Commentary: OpenAI's GPT-4 has some limitations that are fixable — and some that are not GeekWire
Reading Insights
0
12
6 min
vs 7 min read
92%
1,397 → 106 words
Want the full story? Read the original article
Read on TechCrunch