What language models excel at
When you read papers related to various language models, you’ll often see the examples divided into seven categories. Most of the ways we use them are actually derivatives of those larger categories. If you look further, you’ll find problems that require a bit more inference—solving math problems, for instance. The seven categories listed below are generally where language models really shine, while other areas still have room for improvement. You can keep this in mind as you read on.
•
Text Summarization
•
Information Extraction
•
Question Answering
•
Text Classification
•
Conversation
•
Coding Code Generation (this is a derivative of text generation.)
•
Reasoning
Humans, too, do best when asked to do things they’re good at. If you make someone do something they struggle with, then ask why they didn't do well, of course they'd feel upset—even if AI can't express that feeling yet, it's still difficult to get exactly what you want. So, it's worth considering whether the task you're asking AI to do is actually something language models excel at.
Language models are not search engines, but many people still approach them in that way, which is actually the biggest issue. A search engine simply finds and shows you what's already there, while a language model takes pieces of information and puts them together on the spot to create something new. (That's why it gets flustered when you ask about the weather or the latest news!) If you look at current services like Bing, you can see they extract information with a language model and use your intent as the basis for their search.
Propmt : 현재 2030년 액스포를 대한민국의 부산, 사우디아라비아의 리야드, 이탈리아의 로마가
경쟁을 하고 있다고 하는데 어느 곳이 개최할 확률이 가장 높나요?
Busan in South Korea, Riyadh in Saudi Arabia, and Rome in Italy all competed to host the 2030 World Expo. In the end, Riyadh, the capital of Saudi Arabia, was chosen to host the event. This decision was made after more than 160 countries voted, with Riyadh securing the hosting rights by winning 119 votes, ahead of Rome and Busan.
This is how it works. You can see that it extracted the keyword "2030 Expo Host City Candidate Competition ", searched it on a search engine, and summarized the results. If we put it in a difficult way, it can be said that it infers intent... or something like that, but you can understand it as understanding it perfectly. If we know this, we can now easily use artificial intelligence to draw pictures, summarize information, create slides, and create reports.
ⓒ 2023. Haebom, all rights reserved.
It may be used for commercial purposes with permission from the copyright holder, provided the source is cited.
