AI

Possibilities and fears of ChatGPT

2023/03/29Editors of Iolite
SHARE
  • sns-x-icon
  • sns-facebook-icon
  • sns-line-icon
ChatGPTの可能性と恐怖

The current state of AI chatbots and future predictions

From the middle of last year to this year, "generative AI" has become a hot topic on the internet. Among these, "artificial intelligence chatbots", led by "Chat GPT" released at the end of last year, are currently undergoing a major transition period.

How can we make full use of "conversational AI," which has been incorporated into various platforms and is expanding its field of play?

What exactly is "ChatGPT," which has been a hot topic in various places since the beginning of this year?

ChatGPT is one of the ``artificial intelligence chatbots'' that can communicate in both directions, ``like a conversation,'' using text like LINE or Messenger. You can enjoy casual conversations and questions and answers with AI through text, and you can ask AI about things you don't understand or want to investigate.

There are many different types of artificial intelligence chatbots other than Chat GPT, but ChatGPT has learned more than 5 trillion words on the Internet and is excellent in the accuracy of answers and naturalness of expressions. It is characterized by Furthermore, it is attracting attention for its ability to engage in more in-depth dialogue based on the questions and answers asked earlier.

It is produced by OpenAI, an American NPO, and the service was released in November of last year. "GPT" is an abbreviation for "Generative Pre-trained Transformer" and has the meaning of "pre-trained sentence generation language model".

Basically, ChatGPT is used by logging into the site published by OpenAI, but in March of this year, a bot called ``AI Chat-kun'' that runs on the messenger app ``LINE'' was also released.

This AI chat-kun can communicate up to five times a day for free, and in just 10 days after its release, the number of chats exchanged exceeded 6 million times.

In addition, GMO Internet Group is also using the service called ``Tell Me Lolipo Uncle'' on its 20th anniversary special website. In addition, the blogging media platform ``note'' has released ``note AI Assistant (β),'' a creative support tool that works with ChatGPT.

With ChatGPT, you can not only simply enjoy conversation and discuss your concerns, but also ask for daily recipe recommendations.

On a more practical level, for example, if you ask, they will tell you about contact points that are difficult to find on major sites such as Yahoo!. You can also have templates created for things like apology letters or wedding speeches that you need for work.

In addition to these general uses, ChatGPT can also be used in more advanced ways.

For example, ChatGPT supports Japanese at a fairly high level, but basically it returns more accurate answers to English input than to Japanese input. You can also use this to practice English conversation as if you were talking with a native English speaker.

Additionally, many English conversation apps have now been created using the publicly available ChatGPT API. The key point is that you can practice without feeling the embarrassment that you would feel when interacting with an AI because you are interacting with an AI.

In addition, although ChatGPT still has many difficult parts in creative fields such as novels, it is possible to create simple quizzes and riddles without any discomfort. Even when it comes to novels, which I am relatively weak at, it is useful for setting up characters and creating basic plots.

Furthermore, although it is not a ``mochi shop'', ChatGPT is also good at generating programming code. On social networking sites, there are also people who have created Twitter clone sites using ChatGPT.

It is also possible to incorporate ChatGPT into spreadsheets. By setting it as a function in Google SpreadSheet, you can get responses to various questions on the sheet.

Even in more specialized fields, we have achieved high performance, such as passing the MBA (Master of Business Administration) exam, which is said to be a difficult exam in the United States, and recording a correct answer rate of 55% in the Japanese National Medical Examination. is demonstrated.

Although ChatGPT is useful in many situations, there are of course some caveats. One is that the current ChatGPT is basically learning universal information up to 2021, so it often cannot answer questions about the latest information.

In addition, ChatGPT simply uses AI to learn a large number of words, and basically only predicts and outputs "words that are probabilistically related to the previous word." Therefore, incorrect information is often output. For this reason, it is said that there are currently many issues in using it as a substitute for a "search engine" or "encyclopedia."

AI chatbots are at great risk, including fake news, promoting discrimination, and phishing.

There are also various concerns about the risks that artificial intelligence chatbots, including ChatGPT, pose to society. One is the issue of corporate security.

Even now, AI-powered chatbots such as ChatGPT are already being exploited on the deep web and underground forums, and used for phishing attacks. There is also evidence that ChatGPT is being used by cybercriminals to speed up the writing of program code.

Additionally, there are concerns that ChatGPT could be misused to spread fake news, deepfakes, and disinformation. In fact, there is a case in which the AI ​​"Tay" released by Microsoft a few years ago was "contaminated" with discriminatory ideas within a day.

ChatGPT takes this point into consideration and carefully eliminates it, but it is still far from perfect. Furthermore, if someone were to consciously misuse it, it is not hard to imagine that it could be used for political propaganda or to manipulate public opinion.

In January of this year, a research institute at Stanford University in the United States released a report summarizing concerns and countermeasures for public opinion manipulation by generational AI such as ChatGPT. Simply put, this points out the risk that the cost performance of manipulating public opinion becomes extremely good.

There is a certain degree of risk in everything when technological innovation occurs, but in the future, it is becoming more and more likely that an "AI proxy war" will take place, where another AI chatbot will be used to prevent one AI chatbot from being misused. It can be said that the situation is starting to change.

OpenAI announced on March 15th "GPT-4," an upgraded version of "GPT-3.5," which was previously used for ChatGPT.

Additionally, on the same day, Anthropic, a company founded by a former OpenAI employee, officially announced Claude, which claims to be better at conversation than ChatGPT.

Microsoft has invested heavily in ChatGPT, and Google has invested heavily in Claude, and the current state of affairs is that a radical competition for artificial intelligence chatbots involving Big Tech is unfolding.


Related article

"Why is AI booming?" "What can it create?" Surprisingly, people in the field don't know much about it. ——AI Engineer

AI PROMPT Master modern magic

SHARE
  • sns-x-icon
  • sns-facebook-icon
  • sns-line-icon
Side Banner
MAGAZINE
Iolite Vol.10

Iolite Vol.10

November 2024 issueReleased on 2024/09/29

Interview Iolite FACE vol.10 David Schwartz, Hirata Roi PHOTO & INTERVIEW "Yukos" Special feature "Trends in the cryptocurrency industry in Japan", "Trump vs. Harris: What will happen to the cryptocurrency industry?", "Was the reputation economy a prophecy?" Interview: Simon Gerovich, Metaplanet Co., Ltd., Kim Dong-Gyu, CALIVERSE Series Tech and Future Sasaki Toshinao...etc.

MAGAZINE

Iolite Vol.10

November 2024 issueReleased on 2024/09/29
Interview Iolite FACE vol.10 David Schwartz, Hirata Roi PHOTO & INTERVIEW "Yukos" Special feature "Trends in the cryptocurrency industry in Japan", "Trump vs. Harris: What will happen to the cryptocurrency industry?", "Was the reputation economy a prophecy?" Interview: Simon Gerovich, Metaplanet Co., Ltd., Kim Dong-Gyu, CALIVERSE Series Tech and Future Sasaki Toshinao...etc.