ChatGPT（Chat Generative Pre-trained Transformer）

Glossary

「ChatGPT（Chat Generative Pre-trained Transformer）」

This glossary explains various keywords that will help you understand the mindset necessary for data utilization and successful DX.
This time, let's take a look at "ChatGPT," which is expected to be used in future business data and AI utilization.

What is ChatGPT (Chat Generative Pre-trained Transformer)?

ChatGPT (Chat Generative Pre-trained Transformer) is a conversational AI service developed and provided by OpenAI. It is based on OpenAI's GPT, a large-scale language model (LLM), and is classified as a type of generative AI.
When ChatGPT was released in November 2022, it became a global hit due to its groundbreaking user experience, which allows users to use it as if they were chatting with a human through its conversational interface, and its unprecedented ability to perform a variety of tasks by providing instructions in natural language.It is expected to be widely used not only in business but also in society in the future.

A chat using the large-scale language model "GPT"

ChatGPT is an AI that can be used with a conversational interface, developed based on OpenAI's in-house large-scale language model "GPT" (Generative Pre-trained Transformer), specifically "GPT-3.5." It is also classified as "generative AI," and is often considered a prime example of the recently popular "generative AI."

The technology underlying ChatGPT is the large-scale language model GPT, which was first introduced by OpenAI in 2018. GPT-3, which was released in 2020, is a formidable technology that can generate sentences so natural that they are indistinguishable from those written by humans, and has become a hot topic of discussion worldwide among technology experts.

GPT itself has existed for some time and has been a hot topic, but ChatGPT was developed by adjusting GPT to enable it to be used as a chat service and offered to external parties. ChatGPT was launched as a prototype service by OpenAI in November 2022.

For more information on large-scale language models (LLMs), please see this article.

⇒Large Language Model (LLM) | Glossary

The unprecedented boom that followed the arrival of ChatGPT

At the time, most people had never heard of GPT or OpenAI. There was no promotional activity, and even though it was free to use, the registration process was not easy to understand. Normally, even a well-designed service would have difficulty attracting users.

Despite this, the service experienced an unprecedented boom, with over 100 million active users just two months after its launch. This was an exceptionally fast growth compared to general-purpose cloud services that are offered as easy-to-understand, user-friendly services and actively promoted. It took Instagram two and a half years and TikTok nine months to reach 100 million active users.

In addition to its technical significance and what can be achieved through its use, ChatGPT is unprecedented in the magnitude of the impact it has had on society and the unprecedented attention it has received from society.

The feature of "interactive interface"

One possible reason ChatGPT has become such a hot topic is its conversational interface. For example, iPhone's Siri (compared to ChatGPT) does not have very high conversational abilities and can only respond to keywords and provide other services, but the fact that it can converse with users is impressive and has made a strong impression on them. Amazon's Alexa has a similar feature.

Furthermore, even in an era when computers did not have sufficient processing power, "chatbots" that acted like human conversations were developed and became a hot topic.

A famous example is ELIZA, developed in 1966. It had no understanding of what humans were saying at all, and was only capable of making appropriate responses or using phrases entered by the other person to reply without understanding them, using "simple string processing only." In other words, it was merely a trick to create the illusion of conversation.

However, many people believe that ELIZA is listening carefully to them and that they are being given advice as a psychiatrist, and something surprising happens: some people even confide their worries to ELIZA and ask for her advice.

If you think about it, even when people are talking to each other, they may not be listening to what is being said or understanding it at all, but they may still nod along or repeat what the other person said, and somehow end up having a conversation.

Also, I think that sometimes people call the iPhone's "Siri" "Siri-san." On the other hand, they rarely call it "iPhone-san." Humans have a special reaction to and empathize with the being they are talking to.

That's probably how special a "conversational entity" is to people and their brains, and the emergence of ChatGPT as a "conversational AI" matched "human needs" that IT had not been able to meet until now, which is why I think it has become such a hot topic.

It made programming in natural language possible

In the past, if you wanted a computer to do something, you basically had to program it. Using an application, you could do many things within the scope of pre-provided functions, but if you wanted to do something new beyond that, you couldn't make the computer do what you wanted without programming.

With ChatGPT, you can do a variety of things just by telling the computer what you want it to do in natural language. You can ask it to convert data formats, visually summarize data as graphs, adjust the appearance of graphs, and more. It's like using it to control a computer in natural language, just like asking a human, without any programming.

Furthermore, if you specify what kind of program you want to create, it will output the source code for that program, meaning it can automatically program according to your instructions. Unfortunately, the output of ChatGPT (as it stands now) tends to contain errors, oversights, and things that you didn't specify, so it can't actually do as much as it looks.

Therefore, it will not immediately replace all programming or programmers, but even so, the sense of use and possibilities of being able to freely control a computer simply by using natural language means that if you can use Japanese, you can do the equivalent of programming, or it means that an environment has emerged in which the natural language of Japanese can be used as a programming language.It can be said to be an unprecedented user experience.

"Generative AI" that can be used in natural language

ChatGPT is often referred to in the media as "generative AI." While the underlying technology can certainly be called generative AI, generative AI is not a term that only refers to something like ChatGPT (more specifically, the "chat function" is not referred to as "generative AI").

"Generative AI" is a term that envisions many uses beyond chat. I think some people are confused about this.

⇒Generative AI | Glossary

ChatGPT as a groundbreaking environment for "generative AI"

ChatGPT is often referred to as generative AI in the world, but rather than being a "generative AI," I think it would be more appropriate to describe it as an "AI with the ability to converse in natural language" developed using a large-scale language model.

ChatGPT does have an aspect of generative AI, but I think the unique feature of ChatGPT is best explained by the fact that it allows users to use generative AI in a groundbreaking way, with "natural language conversation" as the user interface.

Alternatively, ChatGPT's unique feature is that it combines the aspects of a groundbreaking environment that can be used by users to communicate their requests in natural language without using programming or special means, and a generative AI that can produce the output that users want.

The problem of "hallucination"

Unfortunately, ChatGPT's answers tend to be a mix of incomplete answers, mistakes, and even outright lies. While ChatGPT is capable of handling an incredible variety of tasks compared to previous technologies, it also has noticeable imperfections in its execution.

The "usability" is groundbreaking, and judging by impression alone, it can even give the impression of being very intelligent, but because the answers given seem plausible but the quality of the content is not necessarily good, it has been described as "blatant liars" and "fluent nonsense."

For example, (at least for now) when you ask it to do programming, it generates source code, so at first glance it looks like it can program quite well. However, that source code often doesn't actually work or behaves in an undesirable way, and ChatGPT answers such problematic answers with confidence, as if it understands them very well.

These characteristics require caution when using them. Furthermore, their ability to confidently and convincingly tell lies that are false can be extremely dangerous if they are misused to spread false information in society.

One particularly common problem is not just a lack of knowledge or inaccuracy, but a phenomenon called "hallucination," in which people answer questions about things that don't exist at all as if they did. For example, people might confidently answer questions about a movie that doesn't exist, including details and anecdotes about the movie.

It is not known (at the time of writing) why hallucination occurs. It is often thought that it is simply due to a lack of training data and that it will be resolved by increasing the amount of training data, but it is a phenomenon of unknown cause, and it is known that it may be a fundamental problem with the technology that uses neural networks, and that increasing the amount of training data may not solve the problem.

Hallucination is a fatal behavior when using ChatGPT as a means of asking facts or when errors in judgment or logic are essential. It is a serious risk that must be taken into consideration when using ChatGPT.

Things to be aware of when using ChatGPT for business purposes

When using conversational AI such as ChatGPT in business, there are a few things you should be aware of.

Your input may be used for learning purposes.

The information entered by users may be used for AI training. As a result, confidential information or private data may be learned and incorporated, potentially resulting in leakage to the outside. Users should avoid entering such data carelessly, or use the system in a setting that does not allow data to be used for training.

May produce inaccurate output

Even if it looks plausible to those without knowledge, it may not be accurate or may give inaccurate or insufficient answers. There are also cases where the automatically generated program code is vulnerable. You should thoroughly check the content yourself before using it, or use it for low-risk purposes such as supporting idea generation.

Social concerns such as copyright infringement and moral issues

There is a possibility that AI may output data that infringes intellectual property rights such as copyright and privacy rights, for example by outputting almost exactly the original data used for learning. Also, in areas where there is ongoing social debate about the appropriateness of generative AI activities or content generation, it may be advisable to use it with caution and to clearly state that you are using generated content.

Making good use of ChatGPT with "Prompt Engineering"

ChatGPT can be likened to programming in natural language, or in other words, it can be seen as a situation where you are programming in a situation where you can only use Japanese.

In particular, currently, it is not possible to fully obtain the desired results simply by stating your request in Japanese, and it is common for requests to not be conveyed properly, or for the output to be different from what was expected or incorrect. In order to avoid this, you need to be "creative" when using it to effectively use it.

Even though we say we're trying to be creative, this is an environment where only text data input is allowed. We need to make good use of that range. So, instead of simply typing requests as text, we're devising natural language prompts that are tailored to ChatGPT. This is an initiative called prompt engineering.

Pros and cons of prompt engineering

Some people consider prompt engineering to be a transitional bad technique rather than an essential approach, while others believe it is an essential approach in utilizing generative AI and a technique that will develop and become essential in the future.

Examples of prompt engineering techniques

Just as when a human programmer writes specifications and develops a system, it is recommended that the request, the background to the request, and the purpose be written logically and clearly. In addition, unique techniques specific to ChatGPT (or large-scale language models) are being discovered, such as "including specific phrases in the prompt improves performance."

Examples of prompt creation techniques:

Few-Shot Prompting / Few-Shot Learning
CoT：Chain-of-Thought / Zero-shot CoT
Think in English and answer in Japanese

Use Case Example

Possible use cases for ChatGPT include:

Assistance with creating emails and business documents
translation
Summary of the text
Proofreading and making text easier to read
Change your writing style and target audience
Change the format of the text
- Convert bullet points to sentences, sentences to bullet points, sentences to conversations, etc.
Data conversion (e.g., converting Chinese numerals to Arabic numerals)
Enumeration of items
- - Listing survey items, data items, work tasks, etc.
Get ideas
- Generate catchphrases, concepts, etc.
- Give keywords and write
- Ask them to think of a parable
- Get help brainstorming
- Have a debate

"iPaaS" that can be used in conjunction with external systems and data

Up to this point, I have been explaining how to use ChatGPT by typing text yourself. Since it was originally developed as a chat system, using it by yourself is in line with the product's intention.

However, there may be cases where you want to incorporate ChatGPT's conversational capabilities into your own system, and the data and functions you want to utilize using AI, etc., are likely to be external to ChatGPT. There is also growing interest in using ChatGPT in applications where it can be automatically called from an external source using a program, etc., and the processing results can be utilized.

In that case, what is needed for advanced use is how to secure a means to combine it with external data and systems. This can be achieved through programming, but since ChatGPT can be used without programming, we would like to find a way to do it in a different way.

Therefore, one way to further utilize conversational AI such as ChatGPT is to use a method that allows large-scale language models to be freely linked to various clouds, systems, and data without having to write and develop source code ourselves. For example, this can be achieved by using "connecting" technologies such as "DataSpider" and "HULFT Square," also known as "EAI," "ETL," and "iPaaS."

Can be used with GUI only

Unlike regular programming, there is no need to write code. By placing and configuring icons on the GUI, you can achieve integrated processing with a wide variety of systems and data.

Being able to develop using a GUI is also an advantage

No-code development using only a GUI may seem like a simple compromise compared to full-scale programming, but if development can be done using only a GUI, it will become possible for on-site personnel to proactively utilize AI themselves.

The people who understand the business best are the people on the front lines. They can rapidly create the necessary things, such as how to utilize data and AI, which is an advantage over a situation where they have to explain things to engineers and ask them for help every time something needs to be done.

Full-scale processing can be implemented

There are many products that claim to allow development using only a GUI, but some people may have a negative impression of such products as being too simple.

It is true that things like "it's easy to make, but it can only do simple things," "when I tried to execute a full-scale process it couldn't process and crashed," or "it didn't have the high reliability or stable operating capacity to support business operations, which caused problems" tend to occur.

"DataSpider" and "HULFT Square" are easy to use, but also allow you to create processes at the same level as full-scale programming. They have the same high processing power as full-scale programming, as they are internally converted to Java and executed, and have a long history of supporting corporate IT. They combine the benefits of "GUI only" with the proven track record and full-scale capabilities for professional use.

What is necessary for a "data infrastructure" to successfully utilize data?

Of course, the ability to connect to a wide variety of data sources is necessary, and high processing power to process large amounts of data is also required to fully support actual business operations. At the same time, flexible and rapid trial and error led by the field is also essential.

Generally speaking, if you want high performance and advanced processing, the tool will tend to be difficult to program and use, while if you want ease of use in the field, the tool will tend to be easy to use but have low processing power and can only perform simple processing.

In addition, it is also desirable that the candidate has advanced access capabilities to a wide variety of data sources, especially legacy IT systems such as mainframes and non-modern data sources such as on-site Excel, as well as the ability to access the latest IT such as the cloud.

There are many methods that meet just one of these conditions, but to successfully utilize data, all of them must be met. However, there are not many methods for achieving data integration that are both usable in the field and have the high performance and reliability of a professional tool.

No need to operate in-house as it is iPaaS

DataSpider can be operated securely on a system under your own management. With HULFT Square, a cloud service (iPaaS), this "connecting" technology itself can be used as a cloud service without the need for in-house operation, eliminating the hassle of in-house implementation and system operation.

Related keywords (for further understanding)

EAI
- It is a concept of "connecting" systems by data integration, and is a means of freely connecting various data and systems. It is a concept that has been used since long before the cloud era as a way to effectively utilize IT.
ETL
- In the recent trend of actively working on data utilization, the majority of the work is not the data analysis itself, but rather the collection and preprocessing of data scattered around, from on-premise to cloud. This is a means to carry out such processing efficiently.
iPaaS
- A cloud service that "connects" various clouds with external systems and data simply by operating on a GUI is called iPaaS.
Large-scale language models
- A large-scale language model (LLM) is a natural language model trained on a large amount of text data. It usually refers to a trained neural network model trained on a large corpus using deep learning technology. It is also known as the technology that underpins ChatGPT.

If you are interested in our "Connecting" initiative,

If you are interested, please try out our products that solve IT system and business problems by using the concept of "connecting."

The ultimate "connecting" tools: data integration software "DataSpider" and "HULFT Square"

Our in-house developed and sold data integration tool, "DataSpider" and "HULFT Square," are "connecting" tools with a long track record.

Unlike regular programming, development can be done using only the GUI (no code), without writing code. This means that it can be used by business personnel who have a good understanding of the business and can grasp the specific issues surrounding their company's silo structure.

There are many tools available that allow simple integration, but this tool is easy to use, even for non-programmers, as it only has a GUI, and it also has "high development productivity" and "full-scale performance that can serve as the foundation for business (professional use)." It can smoothly solve the problem of "connecting disparate systems and data," which is hindering the successful use of IT.

We offer a free trial version and hold online seminars where you can try out the software for free, so we hope you will give it a try.

Free product introduction seminar