In recent days, a trend of sharing chat logs has swept through the internet, with the chat partner being the same 'person', an AI chatbot that can chat, write copy, code, and even draw—ChatGPT.

ChatGPT is the latest generation of dialogue robot developed by the American artificial intelligence research organization OpenAI. Compared to previous AI software, ChatGPT's level of intelligence far exceeds people's expectations. Users can input questions from any field, and ChatGPT can generally respond fluently and logically. The content also hardly shows traces of AI, attracting countless netizens to eagerly 'train' it.

ChatGPT has been around for less than a week, and the user base has already exceeded 1 million. At its peak, it was even temporarily brought down by enthusiastic users.

What exactly is ChatGPT, that it can trigger such a tech frenzy?

First, let's have ChatGPT introduce itself.

Chat, as the name suggests, positions ChatGPT as a chatbot. GPT stands for Generative Pre-trained Transformer, which is a text generation deep learning model trained on available internet data for tasks like Q&A, text summarization, machine translation, content classification, code generation, and conversational AI, among others.

The introduction mentions GPT-3, indicating that this model has now progressed to its third generation. Its predecessor, GPT-1, was born in 2018 and was more limited to understanding user input text rather than compiling and interacting. GPT-2 was released in 2019, which built on GPT-1 by incorporating a more comprehensive dataset. In addition to understanding capabilities, GPT-2 also possessed strong content production abilities.

The current GPT-3 has a higher data volume and larger parameters. The data volume of GPT-3 is about 1000 times that of GPT-2, and its model parameters are about 100 times larger. The English Wikipedia, containing 6 million articles, accounts for only 0.6% of its total learning dataset. In simple terms, it is an encyclopedia that includes knowledge in literature, politics, medicine, law, religion, and more.

The result is that GPT-3 learns more deeply and quickly than the previous generation model, and it also excels in areas beyond natural language processing. It can generate more natural and diverse responses based on received input, closely aligning with human language habits, avoiding mechanical repetition. It can even automatically output a series of simple code snippets based on instructions and performs exceptionally well in foreign language translation.

Facing such a cutting-edge product, crypto users, as 'tech trendsetters', naturally won't miss out. In fact, Web3 users might be among the most active groups testing ChatGPT, with figures like Binance CEO CZ and Ethereum founder Vitalik personally getting involved.

For example, CZ asked ChatGPT how to run a successful cryptocurrency exchange, and ChatGPT's answer was, 'Ensure platform security, provide a user-friendly interface, support a wide range of cryptocurrencies, offer quality customer support, and maintain regulatory compliance.'

From @cz_binance

CZ seems quite satisfied with this answer; he believes 'these points are better summarized than most cryptocurrency exchange founders, including himself,' and stated, 'I will print the answer and put it on the wall.'

Vitalik Buterin, after conducting coding experiments with ChatGPT, published an article stating that while ChatGPT makes mistakes during coding processes, it also proposes new coding patterns and speeds up coding. Vitalik believes that 'ChatGPT is very helpful as a programming assistant.'

However, he also made it clear that 'AI cannot replace human programmers.'

As Vitalik said, ChatGPT lacks human imagination and creativity; the content it outputs is based on existing data. Therefore, it is impossible to fully replace human programmers. However, its ability to integrate existing knowledge is indeed very outstanding.

Based on this, we can try to discuss which capabilities of ChatGPT can bring changes to the Web3 industry.

Knowledge base

ChatGPT has strong information summarization and consolidation capabilities. By inputting related information from the knowledge base during training, when users input 'trigger words' related to questions during use, ChatGPT can provide corresponding answers based on the trained knowledge base information.

Therefore, ChatGPT can serve as a tool for learning the basics of Web3. When you input keywords like 'Bitcoin', 'Web3', or 'blockchain' into ChatGPT, the answers it outputs are of high quality in terms of relevance and accuracy.

From @billtheinvestor

However, the training model's data for ChatGPT is currently up to 2021. Therefore, if you inquire about events that occurred after 2021 (such as the Ethereum merge), you will not receive a correct answer. After all, it is still in the testing phase. It is believed that once the official version is launched, with regular training and updates, ChatGPT's answers will be able to reflect the latest content from the knowledge base in a timely and accurate manner.

Coding ability

What surprises people the most about ChatGPT is its ability to understand and generate code. For example, if you tell it what kind of functionality you want to achieve, as long as you describe it in enough detail, it can accurately understand and generate the code.

For example, if you ask it to write a Solidity code snippet for publishing an NFT, ChatGPT can easily accomplish that.

This is highly significant for the Web3 industry, as it could allow ChatGPT to automatically write some basic code, assist programmers in coding, debug code bugs, optimize code, and more. It could even act as a security guard, identifying phishing attacks and providing early warnings for hacker attacks, among other actions. Programmers may still need to intervene for debugging during this process, but it will effectively improve the work efficiency of Web3 technical personnel and provide a lower entry barrier for users with less coding ability, making Web3 technology more accessible.

Writing ability

ChatGPT can generate human-like text responses based on input information. In fact, ChatGPT has passed the Turing test in the writing domain, meaning that you can no longer distinguish which articles were created by ChatGPT and which were created by humans.

Therefore, for some simple copywriting needs, ChatGPT can basically handle them effortlessly.

For example, writing a promotional slogan for a project:

Write an announcement for an NFT airdrop event:

Analyzing the market is also not a problem (it can be said to have mastered the essence of an analyst):

Moreover, as long as the input requirements are more detailed, ChatGPT's output can be closer to what you want. Generally, with slight modifications, it can be directly used. However, ChatGPT still has some shortcomings in long copywriting and is currently not capable of creating works that require strict logical reasoning.

Conversational ability

Conversation is ChatGPT's 'traditional craft'; its initial positioning was as a chat tool. Therefore, ChatGPT's direct application is to serve as a chatbot for social media platforms like Discord and Telegram.

Having undergone extensive training, GPT-3 can be said to be very 'knowledgeable'. Simple tasks like welcoming new group members or sending 'GM' are clearly underutilized.

Since ChatGPT has absorbed a vast amount of knowledge from various fields, it means it can call upon a multitude of knowledge to converse with you. Whether you have confusion regarding specialized knowledge or emotional needs, ChatGPT may not provide the ideal answer, but it can inspire you with other related answers. Furthermore, it learns during the conversation, which enhances the service level of customer service chatbots significantly with the combination of AI and personalized character customization.

For instance, the Web3 game developer Webaverse has previously attempted to integrate OpenAI's GPT-3 model into NPCs in their games. Compared to traditional NPCs, these NPCs are smarter; besides answering questions within the game, they can also respond to simple questions about the real world.

Summary

It must be said that the innovation of a technology impacts an industry in many ways. The above speculations about how ChatGPT transforms the Web3 industry are only based on currently available information and personal imagination.

The development of technology is rapid, and according to information revealed by OpenAI, ChatGPT is just the 'appetizer'; the real main event is the GPT-4 that will be released next year. GPT-4 will be able to pass the Turing test and can advance to a level indistinguishable from humans. I believe there will be even more unexpected points of convergence with Web3 at that time.