Early experiments researchers have done with GPT-4 show that we soon break through a barrier, namely that AI soon breaks through the wall that prevents our further cognitive development. The quad version turns out to be considerably smarter than ChatGPT and may be the AI that passes the Turing test.
GPT-4's performance is strikingly close to human-level performance, and often vastly surpasses prior models such as ChatGPT.
GPT-4 can solve novel and difficult tasks that span mathematics, coding, vision, medicine, law, psychology and more, without needing any special prompting. Moreover, in all of these tasks, GPT-4's performance is strikingly close to human-level performance, and often vastly surpasses prior models such as ChatGPT
If the AIs become a little better at lying now, they have the potential to outcompete politicians. However, if you create clearer prompts, GPT-4 will have to switch to being truthful. So the Turing test can perhaps be narrowed down to whether "large language models" manage to lie more convincingly than a politician. And even the previous generation was convincing enough:
and-which-is-made-of-people-can-you-det?referer=https%3A%2F%2Fwww.aftenposten.no" target="_blank">https://e24.no/karriere- and-management/i/l3lekM/here-guess-nrk-boss-which-job-search-is-made-with-artificial-intelligence-and-which-is-made-by-humans-can-you-do- that?
So to the extent that it is dangerous for AIs to make things up, it is because some people will deliberately go in to manipulate truths. Was this supposed to be something new?
...the final breakthrough will come with 5G - GPT-5 and the like.
Comments