According to this fresh-out-of-the-oven paper, "#GPT4 demonstrates impressive capabilities in a wide range of tasks, such as reasoning, content generation, problem solving"
154 pages of use cases worth checking out 👇
https://arxiv.org/pdf/2303.12712.pdf?

The paper highlights the model's "fundamental limitations due to its 'linear thinking' as a next-token prediction machine" and its "substantial flaws, inherent to the next-word prediction paradigm that underlies its architecture."

But researchers still do not fully understand how these abilities arise from the basic building blocks that the model is made of.
Their questions bring to mind the #NMT black box problem.