Is OpenAI’s newly launched GPT Store going out of business after just 2 months?

A dramatic scene unfolded during an interview with Sam Altman recently.

The host asked, what is the most impressive feature of GPT-4 for you? Altman looked up at the sky, touched his Adam's apple with his finger, thought for a moment and said, "I think it's kind of bad."

Altman is undoubtedly a CEO who is good at building momentum. In one sentence, he raised expectations for GPT-5, which is expected to be released in the middle of the year.

But I don’t know if he has thought of the GPT Store that he once officially announced in his mind at this time.

GPT Store was launched in January. After more than 2 months, no one gradually paid attention to it, and they only enjoyed the feeling of being the protagonist for 15 minutes.

Panning for gold from GPT Store is like opening a blind box

The GPTs in GPT Store are customized versions of ChatGPT.

We don’t need to write code and can use natural language conversations to create a ChatGPT with exclusive skills, such as a “digitally proficient media editor.” More professional players can upload files or call third-party APIs.

When the GPT Store was first launched, there were more than 3 million GPTs, but it was only available to paid users of ChatGPT Plus.

I don’t know whether there are too many freebies waiting for GPT-4 to be free, or whether the GPT Store itself is not up to par, and the usage of GPTs is somewhat bleak.

The Information report mentioned that a developer analyzed more than 36,000 GPTs, and about 5% of them had 150 to 500 active users per day, but the vast majority only had 1 to 2 users per day.

"It feels like OpenAI has given up on GPT Store." The developer complained.

Always look for reasons from yourself first. In terms of user experience, compared with January, GPT Store has improved, but not much.

The interface of GPT Store still only has a few categories such as writing, coding, and productivity. Under each category are more than a dozen recommended GPTs.

However, the user's needs are trivial and vague. More than 2 months later, ChatGPT still cannot provide an interface that allows us to describe the needs in natural language and then retrieve the required GPT.

However, some developers have discovered this gap and launched "GPT Finder" and other GPT search functions. Through web search functions, they can find relevant GPT links on platforms such as Google, Bing, X, and third-party GPT navigation websites.

If the mountain doesn't come, I'll do it; if I go, it'll be the mountain. When I browse information streams such as Instant and Xiaohongshu, I occasionally see a self-recommendation or recommendation of a certain GPT. If the function hits a pain point, I will start using it.

Therefore, the process of finding the GPT you want is somewhat random. However, finding it is just the beginning. Whether GPT is useful or not is about the same as opening a blind box.

Most GPTs do a very simple job and have a very clear positioning: they help you save time on searching and finding, and then drive traffic to your own website and services, register your email, and attract your valuable subscriptions.

For example, the well-known design tool Canva, when you ask it to help you make a greeting card, it provides the corresponding template and invites you to click on the picture and jump to the website for editing.

Perhaps because of this, Canva has had millions of conversations and tens of thousands of ratings, but its rating is only 3.2 out of 5.

Those GPTs that do not require jumping may not necessarily save time, worry and effort.

I wanted to see if GPTs could help me create annoying PPTs, so I searched for the keyword "Powerpoint" and found one that was used more often.

However, it can only "talk on paper" and give an outline based on my topic. It does not provide data, pictures and tables, but only provides abstract text guidance.

What's even worse is that its care was misplaced, reminding me to use large fonts and contrasting colors, and even suggested that I add a "question mark" picture to a certain page to indicate that it's time to interact with the audience. It doesn't look like an assistant, but like Yi Lingguang. Party A in a flash.

In some scenarios, there is no need to invite GPTs to come out. There is a sense of "killing a chicken with a bull's-eye". The bull's-eye is not GPT, but our search energy.

For example, when summarizing the content of a book, the results of GPT-4 are almost indistinguishable. If you find a GPT that specifically summarizes books and then talk to them, it will take more time and the effect may not be better than GPT-4.

▲ The picture above shows a certain GPT, and the picture below shows GPT-4.

Of course, there are also GPTs that are particularly useful and indispensable, such as those for translating scientific articles and those for writing prompts for Wenshengtu AI. However, I did not find these in the GPT Store. They were all spread by word of mouth on social media. of.

In the past two months, the most interesting new feature of GPTs should be "@". Similar to @ someone in a WeChat group chat, if you enter @ in the chat box, you can summon multiple GPTs on the same chat interface, allowing users to be the boss. , and provide guidance to AI employees.

However, the scope of @ is also limited to known GPTs, which are either recently used or have been added to the sidebar, and you need to actively select them. There is no good thing for you to enter your requirements and help you select GPTs.

"What you eat is tasteless, it is a pity to throw it away." Finding useful GPTs is a matter of friendship, but forgetting to use GPTs is a duty.

GPT developers who generate electricity for love are in a light mood

When launching the GPT Store, OpenAI promised that developers could make money from GPT usage, starting in the United States, and more details should be available in the first quarter of this year.

However, many GPT developers do not do it to make money. They use GPTs as minimum viable products (MVP) to attract seed users, test the feasibility and market demand of the product, and then iterate on the product.

However, currently, GPTs does not provide much feedback to most developers.

Most obviously, limiting GPTs to ChatGPT paid users directly limits the number of users.

In November 2023, the number of weekly active users of ChatGPT reached 100 million. It's lively, but the payment rate is not as high as expected. Some people estimate that the number of paid users of ChatGPT is between 5 million and 8 million.

The traffic ceiling is visible to the naked eye, and it is also difficult for subsequent developers to overtake in corners, because the GPT Store hardly gives traffic and feedback to newly launched GPTs.

GPT Store follows the Matthew effect. GPTs that have been used more frequently in the early stage have higher weights. When we search, the most conspicuous indicator of GPTs is the number of uses.

Such an elimination mechanism is simple and crude for OpenAI, but those individual developers who cannot brush up the rankings will have less motivation to generate electricity for love. The lack of new blood is not conducive to the long-term development of GPT Store.

In addition, because prompt words are easily leaked, plagiarism of GPTs is still a problem. OpenAI has not been able to solve it so far. When you enter a word in the search box, a bunch of imitations will appear. You can only consciously judge Li Kui and Li Gui based on the number of times they are used.

The problem is that even if it is truly at the top, the traffic is limited. GPT, a text escape game created by a developer, appeared at the top of the GPT Store for 2 weeks and only added 50,000 new conversations, far below his expectations.

With a smaller user base, developers expect higher quality feedback.

From the user's perspective, the "first impression" of GPT is richer than before. Before starting to chat with GPT, you can see information such as introduction, rating, type ranking, number of conversations, and other GPTs of the same developer.

When we talk to a certain GPT for about 6 rounds, the evaluation portal will automatically pop up in the chat interface, allowing you to rate a movie from 1 to 5 stars just like you rate a movie on Douban.

Another evaluation method is hidden deeper. Click on the name of GPT to find the evaluation entrance. Not only can you give stars, you can even leave a message to the developer and send it to the other party's mailbox.

However, the reviews developers receive can be rough.

The Information report mentioned that GPT Store only provides developers with a small part of user satisfaction, engagement and other data. At the same time, OpenAI resets user IDs every 24 hours, and developers cannot track how many repeat users there are.

The next best thing is that many developers share and promote their GPTs and receive feedback in communities such as Reddit.

As of now, individual developers can neither get a share of the pie nor get more useful feedback, and their attitude towards GPT Store has become indifferent, which is in line with the current mental state of young people.

GPTs that replace plugins, waiting to be replaced

Recently, OpenAI quietly announced something: ChatGPT Plugins (plug-ins) will be completely offline starting from April 9.

The plug-in was declared dead in December last year and was only buried in April. At that time, OpenAI announced that it would replace plug-ins with GPTs, saying that GPTs learned from plug-ins.

It can be said that plug-ins are the predecessor of GPTs, and they were also photographed dead on the beach by GPTs. The transition from plug-ins to GPTs reflects OpenAI's understanding of applications in the AI ​​era.

Plugins are built using code outside of ChatGPT and connect to ChatGPT via an API.

GPTs are created through the code-free chat interface within ChatGPT. At the same time, its "Actions" are based on plug-ins, which can call one or more APIs to implement more customized operations and interact with external data or the real world.

From a threshold point of view, GPTs is certainly more user-friendly. Users who do not understand code can get started. Excellent developers know better how to write prompt words, allowing GPTs to search the Internet, call APIs, and perform semantic searches on uploaded knowledge documents.

Last June, Sam Altman said of plug-ins: plug-ins are not product-market fit.

He explained that many plug-in developers seem to be integrating their applications into ChatGPT, but in fact they want to integrate ChatGPT into their own applications.

To some extent, plug-ins still follow traditional ideas to develop applications in the AI ​​era, which is somewhat outdated. During the early testing of the plug-in, the user experience was also very poor, and it often crashed when calling the API.

In comparison, GPTs are more lightweight and stable, and are more in line with our imagination of "Internet portals", changing the process of retrieving, obtaining, and processing information without having to switch between different applications.

GPTs replaced plugins, and perhaps it will be replaced in the future.

Jobs famously said: "People don't know what they want until you show them." He also used Ford as an example. If you ask customers what they want, they may just want faster horses.

Compared with ChatGPT, GPTs is also a "faster horse" and is not a revolutionary change. Its significance lies in solving more specific problems based on the parameters of the model and integrating it with users' daily lives. However, it is not that useful in many cases. .

As OpenAI itself said: "GPT Store expands the positive impact and creative uses of ChatGPT."

At the same time, GPT Store may be a good testing ground for developers to discover user needs and for ordinary people to find the feeling of creation.

Emilia David, a reporter at The Verge, made a synonym replacement for GPT. She used to search through Google or search on websites such as Thesaurus. GPT saved her 30 seconds without having to watch ads.

She also discovered that GPTs was not fully integrated into her workflow because it was based on ChatGPT Plus, not the document she wrote the article on, and if she hadn't paid for ChatGPT, it would have been fine to go back to Google.

The slightly embarrassing situation of GPTs shows that our needs for AI applications are evolving. It is like a transitional container, and the experience it contains can be used.

A week ago, the world's first fully autonomous AI programmer was unveiled, and the craze for AI Agents has once again started in Silicon Valley.

The Information broke the news in February this year that OpenAI is also developing its own AI Agent to take over users' devices and automatically perform complex tasks, which is not the same thing as ChatGPT.

Chatting with AI will get the work done. This is the propaganda tone of AI Agents. Before this day comes, GPTs, which makes users frown and smile bitterly, may contain more possibilities for applications in the AI ​​era because of their imperfections.

It is as sharp as autumn frost, and can ward off evil disasters. Work email: [email protected]

# Welcome to follow the official WeChat public account of aifaner: aifaner (WeChat ID: ifanr). More exciting content will be provided to you as soon as possible.

Ai Faner | Original link · View comments · Sina Weibo