The newly launched ChatGPT by OpenAI is a hit and taking the world by storm. This conversation model is capable of answering follow-up questions, acknowledging errors, challenging incorrect assumptions, and even helping you fix bugs in your code. A few moments of chatting with it and can solve your problems for you. For example, if a user asks, "ChatGPT, help me explain the meaning of regular expressions in a text." ChatGPT would reply, "I got it." And proceed to give a thorough explanation, making sure it's clear to the user.

Due to the excellent performance of ChatGPT, people have started to have fun with it, as if they wouldn't stop playing ChatGPT in a hundred and ten ways. Some people even say without exaggeration that this is not a machine, it's just like a live Stack Overflow.

Even Musk couldn't help but say: ChatGPT is scary good. We're not far from dangerously strong AI.

Programming Q&A Machine

As a conversational large-scale language model, ChatGPT excels at answering questions posed by users and, most importantly, has a foundation of knowledge related to programming. This makes ChatGPT a programming Q&A tool similar to Stack Overflow, only the answer comes from an AI. First, ChatGPT can easily answer questions related to computers:

On the other hand, based on the trial experience, ChatGPT is very "smart" in programming. For example, if you ask "print a big dog," ChatGPT will answer with a whole program that can be pieced together with characters to form the shape of a dog.

When a user commands ChatGPT to find a bug in a piece of code, it can not only clearly describe the bug and its cause, but it can also automatically fix the bug and write the correct program.

ChatGPT can take the exam

It's not uncommon to see models like ChatGPT participating in exams. However, it's difficult to predict how ChatGPT would perform on such an exam. Nevertheless, there is a report of someone testing ChatGPT on the AP Computer Science A exam in 2022, and the model reportedly passed the exam with a score of 32 out of a possible 36 points.

The difficulty level of this examination is on par with what one might expect in the first term of a computer science program at the university level. A portion of the exam's content is shown below in a screenshot, and it appears that successfully completing the exam will still pose a significant challenge.

An individual administered the complete SAT exam, including the ACT exam, to ChatGPT. The SAT exam has a total score of 1600, split into three sections: Reading (400 points), Grammar (400 points), and Mathematics (800 points). Additionally, there is an optional essay component with a maximum score of 24 points, divided into three categories: Reading Comprehension (8 points), Essay Analysis (8 points), and English Writing (8 points). ChatGPT received a score of 1020. Based on data from the College Board, this score is approximately in the top 52% of scores in this range.

ChatGPT can write games

Let ChatGPT write a small game, it's not a big deal for ChatGPT. A user generated a small game from scratch in less than an hour without any coding, just by inputting prompts. If there are problems, tell ChatGPT and it will help you solve it.

For example, the user said: come up with 5 names for a game about digging for treasure, the names must have the initials GPT. ChatGPT: Gold Pit Treasure, Great Pit Treasure, Grand Pit Treasure, Graveyard Pit Treasure, Glittering Pit Treasure. User: I want to start a project named GPT (Graveyard Pit Treasure), so we can develop the game together. I want to use live view instead of ecto. What command should I use? ChatGPT: mix phx.new gpt —no-ecto —live User: I have an app named GPT. We want to make a game. Write a liveview module for this application with 3 elements: a title that reads "GPT: Graveyards, Pits & Treasure"; a Gems counter starting from 0; a button that reads "Dig up the next grave." When I click the button, I either get 10 gems or there's a 50% chance of falling into a pit and dying. If I die, the gem counter should display "You fell into a pit. You are dead."

ChatGPT provided a code link and some instructions.

In subsequent conversations between the user and ChatGPT, ChatGPT helped resolve the problems encountered by the user one by one, and finally, the user-generated an online playable game as shown in the following figure.

In addition, although ChatGPT is a conversational language model, it cannot generate multimodal content by itself, but it can be seen as an intermediate model, for example, by combining ChatGPT and Stable Diffusion. For example, a netizen requested that ChatGPT generate a random art prompt, and then use the prompt as input to Stable Diffusion, resulting in a highly artistic painting.

This way of generating content within a domain without specifying the content specifically may open up new perspectives for AIGC. Of course, as an AI model, ChatGPT has room for performance improvement, especially for various specific applications. The research team said that they will continue to improve ChatGPT based on user feedback to avoid the model generating incorrect content randomly.

ILLA Cloud

ILLA Cloud is a low-code development platform with dozens of front-end components and database API integrations. You can use ILLA Cloud to build the front-end interface by dragging and dropping components and connecting to your database or API to complete full-stack development quickly.

ILLA proudly announces a partnership with Hugging Face, a suite of natural language processing (NLP) tools and services. They are most well-known for their open-source NLP library, which provides text generation, language translation, and named entity recognition tools. With Hugging Face, ILLA is more productive than before. Our users can do more with AI.

ILLA Cloud provides dozens of commonly used front-end components, allowing you to build different front-end interfaces based on your specific needs quickly. At the same time, ILLA offers a connection to Hugging Face, allowing you to quickly connect to the API, send requests, and receive returned data. By connecting the API and front-end components, you can implement the requirement that users can enter content through the front end and submit it to the API. The API returns the generated content to be displayed on the front end.

You can check ILLA’s website here at: illacloud.com

GitHub page: github.com/illacloud/illa-builder

Join Discord community: discord.com/invite/illacloud