A few days ago, Google and MS shared an article about using AI technology in search engines.
The Era of Large AI Search Engines Begins: MS and Google Start
https://www.soonsoon.co/post/dev/10499/
And finally, after a long wait, I was able to try out Bing AI search. Also, I was able to try the Open AI service provided by Azure relatively recently.
So, before I try it out in detail, I want to leave a brief review.
I tried entering the question "Tell me the difference between Bing AI search and Bard search." into Bing Chat Search.
To summarize the answer, Bing AI search seems to have the advantage of connecting to source-based information because it cites sources. I expect Bard to be largely different in this regard as well; otherwise, it would just become a wiki like KakaoWiki because you wouldn't know the source of the information.. Since I haven't been able to use Bard properly recently, it seems to be giving answers like the one above.
I applied the same question to ChatGPT, which is known to use a similar technology, GPT-3.
Huh? That's a weird answer? Is Google's Bard based on GPT-3?
As is widely known, ChatGPT seems to provide incorrect information about things it does not know well. It answers by creating sentences for incorrect information and nonsense quite carefully.
The answer content is written in English just in case, and the answer is translated into Korean. (Even if written in English, it tends to talk nonsense.)
It is known that the technology used in GPT Chat is based on GPT-3, and this content is based on a training model created using data collected up to 2021 to provide answers, but it seems that is the case because there is no proper information about Bard.
So, I asked Bing AI search about this part.
Although the veracity is unknown, Bing AI is said to reflect information within the last hour. Therefore, it appears to provide more accurate information compared to ChatGPT.
Since Bing Search provides answers while citing the sources of information, it appears to enable more correct usage.
Personally, ChatGPT excels at answering with fluent, sometimes offbeat responses, whereas Bing AI seems to show more exemplary answers.
Since MS is directly providing a chat-based AI search, it seems to give much more defensive answers.
And now, with Azure, a cloud service provided by MS, you can also use the OpenAI API. I wrote a similar question in English and asked it. (You cannot input Korean in the Azure OpenAI playground. Or maybe I just don't know how to do it.)
Translated into English, it is as follows.
Bing AI Search is a service that allows you to search for AI content using Google Search. This is different from Google Bard Search, which is a service that allows you to search for AI content on Google Search.It answers quite simply.
I asked for a more detailed answer to the question.
The answer was as follows.
“Bing AI is a search engine that is run by Microsoft. Google’s AI search service is run by Google. Both services are designed to be used on computers. Microsoft’s AI search service is not meant to be used on computers. Google’s AI search service is.”
Bing AI is a search engine operated by Microsoft. Google's AI search service is operated by Google. Both services are designed for use on computers. Microsoft's AI search service is not for use on computers. It is Google's AI search service.What does this mean..
It gives answers that are too obvious.
It is like the answer of the Blade from the legendary Kid Gang comic.
Looking at the Azure OpenAI documentation, it says this friend was trained only up to 2019. ChatGPT, which is known to have been trained up to 2021, seems to provide more fluent answers.
On the other hand, since Azure OpenAI is a service provided by Microsoft, I wonder if it was designed to be more cautious about generating false information. The developer documentation also seems to explicitly mention such points.
By briefly using Bing AI, ChatGPT, and Azure OpenAI Chat, we can directly and indirectly see how Microsoft uses artificial intelligence.
Bing AI appears to focus on the search engine aspect, being able to search for information up to an hour ago while ultimately citing sources.
OpenAI's ChatGPT is a service that focuses on leading the conversation itself based on models trained up to 2021, so ultimately, to utilize it well through conversation, one must likely make good use of the prerequisite process of teaching correct information or the application process known as the Fine Tune methodology.
Finally, the Azure OpenAI service is aimed at individuals and enterprises wanting to use OpenAI's solutions on Azure, so it feels much simpler and less defensive. It feels like there is only substance, while MS's efforts to prevent it from being used in unexpected (?) fields seem to show through.
This concludes a brief review of the AI-related chat services I have used.
I will play around with it a bit more in the future and also leave a post.