Chat GPT artificial intelligence passed the MBA examination. It could adjust its answers by benefiting from the tips that humans gave. That gives an interesting vision of the possibilities that AI has.
Of course, it is possible, that some people try to make their examination by using AI. In this scenario, the person will put the AI to make all work. And then enjoy the results. But there is a possibility that other AI-based systems will uncover straight loans.
There is the possibility that in the future, AI can make many things like data analysis and other kinds of work for humans or helps humans very much. There are needed humans. But their role is to check the results that the AI gives. That means AI can make mistakes.
The AI as a data analyst is as good as the data sources are. If the AI uses uncertain sources, or those sources involve false data, the AI will not realize that, which makes it erratic.
If the AI uses let's say, ten home pages. It uses the first home pages that it finds. And in the worst case, the data that the AI uses involves things like souvenir guides. And in the thesis, that kind of thing is not very impressive.
If the thesis should handle things like the economic situation in the Mediterranean area, the AI should remove things like souvenir guides and fictional novels. The thing, that causes problems for AI is. How to confirm the information that it collects? AI is a very good tool for analyzing pre-processed data.
But if the AI should search the data from the net, it should make a difference between fact and fiction. The problem is that the AI can collect data. But it doesn't understand what that data involves. The Ai knows that the data it collects fills certain parameters. But otherwise, it doesn't know the meaning of the data.
AI is an impressive tool. But the problem is that it doesn't realize the data that it collects for users. It uses certain parameters for selecting texts. And one of them is this: the AI or its user selects. As an example five home pages. And then. It connects paragraphs from those pages. But then there is the possibility that those homepages involve data that has no connection with work that the AI should do.
The error correlation, like how many times a certain world repeats on the homepage can help the error correlation. Still today, there are needed humans who check the information that the AI inputs or outputs. The human must control the data that load into the system. Or the users must check the processed information.
https://www.nbcnews.com/tech/tech-news/chatgpt-passes-mba-exam-wharton-professor-rcna67036
Comments
Post a Comment