Is AI only about Chat? ——AIGC Product Evolution
From the end of 2022 to now, in nearly two years, AIGC products have developed from nothing to something, from a simple Chatbot shell and single generate to now. Not only have they been applied to various application scenarios and fields, but the product form has also developed a lot.
This article discusses the development of AIGC product forms from the perspective of the development of user needs.
AI starts with Chat
Since ChatGPT became popular, many general chatbots with GPT interfaces have appeared. Due to various reasons such as information asymmetry, network, price, OpenAI limitations, etc., these chatbots rely on SEO to quickly obtain traffic and make money.
Moreover, to this day, these chatbots are still doing very well. After all, in the AIGC era, the demand for chatbots is far from enough with just one or two giants like ChatGPT.
Generator explosion
However, general chatbots are far from meeting the needs of specific scenarios. Moreover, since the explosion of ChatGPT, the demand for various generators has exploded, including text, pictures, videos, music, etc. As a result, various simple single-shot generators have begun to explode, and the momentum has only increased since last year.
Two types of generators based on single-talk dialogue
In summary, there are two types of Generators based on single-round conversations.
The first is the single-round dialogue generation result based on a single large model.
For example, the more common ones are text generation, image generation, etc.
The second type is the single-round dialogue generation results based on multiple large models.
That is, based on a prompt submission, multiple large models generate different results, and finally combine the data and return it to the user.
The most typical example is music generation. For example, Suno generates not only music, but also the cover image, title, and lyrics of the music when a prompt is submitted. Moreover, there is a process involved. First, the title and lyrics are generated, and then the music is generated.
The reason for the explosion of single-discussion conversation-based generators
In fact, there are two reasons for the explosion of Generator based on single-round conversation, and AI Search and AI Answer based on single-round conversation.
On the one hand, there is a spurt in the demand for AI in various user scenarios.
However, these requirements are relatively simple, as long as they can simply generate relatively satisfactory results. Because as long as this can be achieved, the efficiency and quality can be greatly improved compared with the previous methods.
For example, writing an email used to take some time to think and write it sentence by sentence, but with an AI Email Generator, you can get a basically satisfactory result as long as you make a request, and then you can make slight changes yourself.
On the other hand, there is the issue of development costs.
For webmasters and developers who make AI products, this single-round conversation generator is easy to develop and has low development costs. It can be quickly produced and put online, and then obtain a large amount of traffic through SEO and realize monetization.
AI Returns to Chat
However, there has been a noticeable change since this year.
Chat-based generators, that is, generators based on multi-round conversations, have begun to emerge for specific scenarios and needs. Compared with generators based on single-round conversations, their advantages are very obvious.
Generators based on single-round conversations are problematic.
Friends who have worked on this kind of generator based on a single round of conversation should have a clear feeling about it. The way users use this product is to generate multiple times with the same prompt to get multiple results. Then, if there is a satisfactory result, it ends. If not, they either modify the prompt and try again, or give up.
This is true whether it is the generation of text or the generation of pictures, music, videos, etc.
If you have never made this kind of single-round conversation-based Generator, you can go to the midjourney server on discord and take a look.
The reason is actually simple. When users use this generator based on specific scenarios and needs, their goal is to easily get the results they want.
However, the generator based on a single round of dialogue actually gets an uncertain result each time it generates. The user may be satisfied with the result, but more often than not, so the user can only try multiple times, or even slightly change the prompt, and then try multiple times in the hope of getting a satisfactory result. In other words, this product cannot give users a way to get a certain and satisfactory result.
However, the Chat (multi-round conversation)-based Generator is different. If the user is not satisfied with the generated result after the first time (which is unsatisfactory in most cases), the AI can be asked to modify it on this basis, even multiple times, until a satisfactory result is obtained.
In other words, the Chat-based Generator gives users a way to get certain and satisfactory results.
For example, the product V0 generates code based on Chat and adds code version management. Every Chat and modification will generate a new version based on the original code.
For another example, an interactive story generator generates a paragraph of the story each time it generates a paragraph, and then the user can continue or control the development of the plot through prompts to generate the next paragraph.
These are things that cannot be achieved through a single-round conversation, but they are ways for users to get the definite results they want.
AI only has Chat?
Demand is progressive
The following content can only be viewed by VIP users.
Subscribe to my VIP membership and you can read all paid VIP content.
If you are already a VIP member, please log in.
Summarize
In the AIGC era, how do we work and create? This is a question that all AIGC products will face.
I recommend everyone to watch the American TV series "Westworld".
In 2018, I made a robot, and that year was when the first season of "Westworld" was very popular. To be honest, this American TV series inspired me a lot in making that robot.
Last year (2023), the fourth season of "Westworld" was broadcast. After watching it, I was particularly impressed by one scene.
The heroine was creating a game scene in the company. She described the general situation to the AI, and then the AI directly rendered the 3D characters and scenes. Then, the heroine repeatedly told the AI how to modify it, and even denied the AI's creation.
Isn’t this the ultimate interaction method for AIGC products? At least one of the ultimate interaction methods.