<It's >

friendly economic time for anchors. Today (16th) we will be with reporter Kwon Ae-ri. The new version of Artificial Intelligence ChatGPT came out yesterday.

<Reporter>

Yes. Some of you who are watching the news right now have used the chatGPT that has attracted worldwide attention, and some of you have only listened to the story.

ChatGPT became immensely popular because it was a super-giant artificial intelligence that could convince people and anyone could easily use it online.

I introduced it to you about Friendly Economy a month ago.

ChatGPT, which was released in November last year and became such a global hit, was version 3.5. So there were old versions of 1 and 2 before.

At that time, conversations with people didn't come so naturally, so it didn't attract much attention from ordinary people, but 3.5, which was somewhat demonstrated before the 4 released yesterday, became a big topic.

The 3.5 was also amazing, but I was looking forward to seeing what the level of 4 that I was going to release in earnest would be. I took off the veil yesterday.

<Anchor
>
So it's a little different from the previous version? What do you think?

<Reporter>
In fact,
the biggest change is that it is still difficult for ordinary users to experience it directly.

Unlike 3.5, which only spoke verbally and chat, 4 can react and speak to images such as photos and pictures, since this ability is still closed to the average user.

But overall, you can immediately feel smarter than 3.5.

Especially when I asked in Korean instead of English, the difference became apparent.

I communicate much better in 26 languages, including Korean, as opposed to 3.5, which was very meagre in non-English language information.

I've shown you here that when you asked 3.5, "Who founded the Republic of Korea?" he replied "Kim Jong-il."

I asked 4 the same questions, and this time I got the correct answer.

In the meantime, I learned much more and more accurately.

My memory is about 8 times better than 3.5. The developer says they're 40% more likely to answer the truth.

So, not only did he get a 4 on the U.S. bar exam, which dropped 3.5, but he also scored in the top 10 percent, the developer said.

However, the learning information of ChatGPT stays at September 2021 as of 3.5 or 4.

When I asked how our national baseball team performed in the WBC, the answer was that we could not provide real-time data, so I watched the news.

It seems a little difficult for this version to be as topical as 3.5, but it is paid. You have to pay 26,000 won a month to experience it yourself.

However, other programs using ChatGPT will soon be available for free.

For example, 4 is used in Microsoft's search engine Bing's chatbot 'Bing Chat'.

The Korean chatbot called 'Asukup' that can be found on KakaoTalk also uses chatGPT.

Not yet, but with these apps you will soon be able to try level 4.

<Anchor>

I see. But now I have to pay to get version 4. So Kwon paid and did it all now, right? (I paid for it.) Thank you for making this personal investment for the sake of the article. Finally, if artificial intelligence continues to be developed like this, there will be a lot of talk about how much artificial intelligence will replace humans. But as this becomes more popular, the use of this discussion is becoming clearer.

<Reporter>

Yes, Noam Chomsky, a leading American linguist, wrote in the New York Times the other day.

Artificial intelligence is asserting that it will always be a step that cannot be compared to humans in its current style.

That's because what ChatGPT says sounds surprisingly natural, but in the end, it only comes up with the most probable answer out of the vast amount of data it has put in.

The core of human intelligence is to "judge" and disagree with very little information, but with the current style of artificial intelligence, we will not be able to make any judgments in the future, just find the most probable answer and come up with it as if we were talking to a person for real.

The level of the question given by the questioner and the information you enter determines the level at which ChatGPT answers, and in the end, it becomes a problem for the person who uses it.

Companies are already starting to hire AI copywriters and AI programmers.

This is not a replacement for people, but a tool that saves people their hands and time so that people can do more high-level work, which is the most advanced tool at the moment and can be thought of as such.

[Lee Talc/Head of Technology at AI startup 'Upstage': Because our company also uses ChatGPT as an intern? Why an intern, it's the level of an intern. What he does well If you know and ask them to do it, they're very good at it, but if you do something else, you can make mistakes and say, 'What is this?' If you're an intern, you'll understand. We see AI at that level right now, and we don't think it's going to be able to completely replace the (human) workforce.]