ChatGPT was born as part of the GPT (Generative Pre-trained Transformer) series developed by OpenAI in 2018, and has made innovative advancements as an artificial intelligence language model. Initially, GPT-1 appeared and demonstrated the potential of natural language processing technology, and in 2019, GPT-2 demonstrated more sophisticated text generation capabilities. GPT-3 was released in 2020, and based on 175 billion parameters, it enabled human-like conversations and creative content generation. As a result, ChatGPT began to attract attention in various application fields as a conversational AI.
A key turning point in the growth of ChatGPT was the continuous improvement of the model that actively reflected user feedback. OpenAI improved the performance of the model and strengthened ethical standards through interaction with users. ChatGPT was also provided in the form of an API, allowing businesses and developers to integrate it into various applications. Through this, it was used in various industries such as education, customer support, and content creation, providing real value.
In 2023, with the release of GPT-4, ChatGPT's capabilities will further develop. With multimodal capabilities, it can process not only text but also image input, and its deep contextual understanding and logical thinking ability have been improved. With these technological advancements, ChatGPT has become a tool that promotes collaboration between humans and AI, and is playing an important role in leading the popularization of AI worldwide.
The interface has barely changed in the decade since the smartphone revolution. There was a text-based chatbot UI, but it tended to lead to set questions rather than understanding the context and communicating. It was difficult to ask various questions and it felt like a machine was answering.
OpenAI has introduced a new interface. It is a fresh shock that produces outstanding results with overwhelming AI generation technology. It makes you feel like you are talking to a real person during the interaction process. They are creating an interface that omits the method of using tools.
The design grammar of the brand and service is simple. It minimizes the impression of technology as much as possible and focuses on the fundamental behavior of the tool itself. It tries to incorporate human fragility and randomness in the design.
Services targeting the general public combine (service type)GPT with a black circle. Chatbot services are expressed as ChatGPT, and search services are expressed as SearchGPT.
We express technology only with fonts and indicate the version with numbers afterward. Since technical achievements are the key factors that shape the market, we add new modifiers such as Turbo and o in addition to numbers. We do not differentiate expression methods according to the product, such as text, image, or video.
GPT stands for Generative Pre-trained Transformer. It is a language model for machine translation using a pre-trained generative transformer. DALL-E, which generates images, is said to be a combination of Salvador Dali and Wall-E. Sora, which generates images, is named after the Japanese word for "sky" to express infinite creativity.
We also provide a 'Powered by OpenAI' badge that you can wear when using the API as a technology infrastructure.
We use a universal impression typeface called Soehen. It is a font with 64 styles in 4 self-sufficient fonts created by New Zealand font foundry Klim Type Foundry in 2019. It is a font created to commemorate Akzidenz-Grotesk, which appeared in 1898 and can be considered the root of modern sans-serif typefaces. It was inspired by 'Standard Medium', which was used in Unimark, the wayfinding system of the New York City subway, which is in a similar series. By using a font used in public facilities such as subways, the brand and service gave a public feeling and reduced the impression of commercial service.
SEE MORE
They market with technology and results. In their technology presentation sessions, they use employee demonstration videos that are not staged with lighting or sequences. They give a harmless impression by filming them as if they were taken with a smartphone in everyday life. In addition, they mainly carry out CEO marketing by putting forward the icon of 'Sam Altman'. They expressed him as being on the same level as Steve Jobs, Elon Musk, and Bill Gates, and quickly dominated the AI market.
Images created by AI are detailed and have strong contrast. They often focus on the foreground and background, and the contrast between light and dark is so strong that the outlines of the objects can be felt. Individual elements are revealed, giving the impression of a collage of multiple graphic elements. Even in videos, it feels like multiple conceptual layers have been merged into one.
The images related to the company use graphics that are believed to have been drawn using AI. The graphics are abstract, like brush strokes, rather than depicting specific materials. The elements used inside the product only use round, light-colored icons that help with identification.
It is an interface where you input text and receive a response. It is also called a conversational user interface (CUI) in a different form from a graphical interface. Anyone can have a conversation and ask questions in text. Therefore, CUI has the advantage that users can use it right away without having to learn how to use the software.
Chatbots are a familiar example. When customers have problems, they use software to provide consultations by phone or email. AI chatbots solve a large amount of consultations that need to be handled instead of people. In addition to these customer consultations, there are also cases where people chat with bots for fun. In Korea, the AI conversation engine called 'SimSimi', which has been in service since 2002, is famous.
In the past, chatbots did not have a detailed understanding of the context of the user's question. They answered questions that did not have a clear answer with a strange translation. As a result, they could only solve simple problems and were only slightly better at keyword search.
The biggest hurdle was interpreting human speech, or natural language. It was nearly impossible to code a machine that could understand people. There were too many possible scenarios to consider, and it had to be able to infer hidden meanings.
OpenAI's ChatGPT uses AI technology to produce results that are tangible. It understands what people say. It gives proper answers. It can understand not only the context of the sentence itself, but also the hidden implications before and after. It provides an experience that is truly like talking to a person. You can ask questions freely, and ChatGPT can now understand high-context information such as images and text.
It excels at tasks that are difficult to code, such as creating tables based on complex and messy Excel data, summarizing a 100-page paper, and drawing pictures with oil painting textures. What we thought only humans could create can now be done by machines.
The process to get the information you want is linear. Since you perform tasks sequentially, it is difficult to go back to a previous step or skip intermediate steps when moving to the next step. It is not easy to grasp the overall structure and move to a specific point.
If the answer comes right away, that's fine. But people are imperfect. It's hard to get an answer right away because of an inappropriate question. It often takes a lot of time to refine the question and exchange answers. Since it's not repeating the same action, it's hard to predict the quality of the answer. Then, it becomes necessary to make further adjustments. Up until now, it's still a lot of time and effort to invest until you get the desired result.
It's useful for giving simple commands without using your hands, like with an AI speaker, but in situations where you have to use your fingers, a GUI that just requires clicking buttons may be more convenient.
October 4, 2024. OpenAI announced Canvas, a new interface useful for writing text and code.
This is a feature of the experience of co-editing with ChatGPT and a specific document. When ChatGPT detects a scenario that could be helpful, it automatically opens a canvas. You can also open it by writing Use Canvas in the prompt. When the canvas is opened, the editing document appears on the right.
You can use it as if you were writing a blog post. There is a list of shortcuts at the bottom right, and you can edit, adjust the length, trim, add emoji, and adjust the reading level. When writing code, a list of shortcuts that are appropriate for that will appear. You can review code, add logs, add comments, fix bugs, leave comments, and port languages.
