"Is your account still there?" This may be the most common greeting from ChatGPT users in the past week. As a result, a series of ChatGPT account topics have become the main theme of this week's discussion around the title, re-registration, and Plus payment.

ChatGPT large area seal

Suspension of new account registration

Since last weekend, the wailing of "my account has been blocked" has been repeated among domestic ChatGPT users. People find that they have been rejected when they log in to ChatGPT, and there is no sign of unblocking. Most of these friends whose accounts have been banned are mass-market accounts purchased from channels such as Mobao and Mooduo, and there are also Plus accounts bought at a large price. , The suspension of signal registration for a time made people even more anxious and panicked.

In addition, some users who logged in because of the Asian node were "banned and hammered" by OpenAI, and some friends who purchased Plus by themselves encountered the embarrassing situation of "only deducting money but not using it". People are speculating whether it is the severe punishment of the non-compliance channels of mass sales, or the official clean-up of the abuse of free APIs... For a while, "whether the account will be banned" has become the "metaphysics" in the AI world during this period, until the registration is repeated. Only then did this title disturbance fade away in people's minds over time. 👉Clickto go to the "ChatGPT" data page

OpenAI closes Plus payment for a short time

The official said that the demand is high and the supply exceeds the demand

Despite negative incidents such as Italy’s ban on ChatGPT, Samsung’s use of ChatGPT’s inadvertent disclosure, and thousands of people’s joint letter opposing the development of super AI, OpenAI is still popular among users, and even appeared on April 5 (Wednesday). It is too large, and you have to temporarily close the Plus payment" . However, because the closing payment was too sudden and the re-opening time was not notified in time, people still felt uneasy in a short period of time, and some even bluntly said that "AI makes our lives so much more convenient, and we can't go back."

Perhaps in response to recent negative voices, OpenAI published an article "Our Approach to AI Safety" on Thursday, April 6, emphasizing that " ensuring that AI systems are built, deployed, and used safely is critical to our mission. Important ”, and at the end of the article bluntly stated that “importantly, we also believe that improving the safety and capabilities of AI should be done at the same time”.

👉Clickto go to the "ChatGPT" data page

Microsoft x Zhejiang University develops HuggingGPT

Use ChatGPT to command professional models to do professional things

On April 4th (Tuesday), HuggingGPT, an AI achievement co-operated by Zhejiang University and Microsoft Asia Research Institute, became the focus of the entire network as soon as it was released, because with it, we only need to speak normally (natural language) to describe the needs , and HuggingGPT will use ChatGPT automatically analyzes which AI models you need to use, and calls the corresponding HuggingFace model to complete the task.

Specifically speaking, the current AI model is more "professional and specialized" to complete specific tasks, and for users, it is necessary to go through tedious steps such as planning tasks, selecting models, executing and waiting for results, especially the two tasks of planning and selecting. step is particularly time-consuming. With HuggingGPT, the "AI transfer package man", we no longer need to consider these processes, just say "what I want" and wait for the result.

Just like Microsoft named its Github project "Jarvis" (artificial intelligence assistant in "Iron Man"), HuggingGPT may become a beginning of the "Everything APP" vision, which may allow us to move towards AGI (artificial general a new stage of intelligence). 👉Click to go to the "HuggingGPT" data page

Midjourney introduces the /describe command

"Pictograph" function officially implemented

Despite suspending its free trial, Midjourney continued to be popular this week. Among them, the "/describe" command updated by Midjourney on April 4 (Tuesday) became a hot topic of discussion. In the past, we needed some extensions or third-party websites to infer prompt words from AI pictures that did not disclose Prompt. The operation was cumbersome and the results were difficult to predict.

However, Midjourney’s newly added /describe command can provide 4 sets of suggested prompt keywords based on the pictures you upload. Not only the extracted prompts are generally of high quality, but the usage also inherits Midjourney’s consistent ease of use, plus Deeply integrated with the original Midjourney function, the one-stop "picture-generated text" experience allows everyone to become a "prompt master" who imitates good pictures.

In addition to the big update of /describe, Midjourney also launched two advanced prompt functions to improve efficiency for advanced users on April 6 (Thursday). Through the --repeat command , multiple 2x2 grids can be created at one time to reduce repetition. command work. If you carefully use special punctuation such as curly braces and curly braces to create arrangements and combinations for instructions , you can also make your instructions more efficient with half the effort. Combined with Midjourney’s upload reference image function, you can also get a convenient reference effect of “get it as soon as you type it”. . 👉Click to go to the "Midjourney" information page

Meta and Ali released new AI models respectively

Separately attack image recognition and GPT large model fields

Meta and Alibaba Cloud respectively announced their new AI models this week. Meta's "Segment Anything" AI model released on Wednesday, April 5th, can recognize objects in images and videos . At present, Meta has used technology similar to this model in practical work such as tagging photos and reviewing prohibited content, and plans to commercialize its proprietary generative artificial intelligence within this year to generate advertising images.

And Ali is also making frequent moves this week. On April 4th (Tuesday), the Alibaba large-scale model joint project team demonstrated a digital clone that was trained in cooperation with talk show actor Niao Niao—Niao Niao Fen Niao. The big model emphasizing individuality, in addition to being able to talk in the "old business" talk show and tell jokes, is also quite "human" in the process of multiple rounds of dialogues such as empathetic conversations and tutoring writing. The test also gave people a little more real sense of the possibility of Ali AI's future integration into real daily life.

In addition, Alibaba Cloud's self-developed large-scale model "Tongyi Qianwen" finally invited users to test the experience on April 7 (Friday). Although the official website opened the application channel, it is reported that only enterprise users are invited to test. According to the introduction on the official website, "Tongyi Thousand Questions" is a large model that specifically responds to human instructions, and can assist humans in efficiency and creativity. On the sword".

👉Click to go to the "Segment Anything" data page

👉Click to go to the information page of "Niao Niao Di Niao"

👉Click to go to the information page of "Tongyi Qianwen"

This week's AI application recommendation

Domains GPT: AI generates easy-to-remember domain names

👉 https://www.hayo.com/entry/1862

promptMANIA: AI art community with online prompts generator

👉 https://www.hayo.com/entry/1906

Mem.ai: AI Helps Build Team/Individual Workbenches

👉 https://www.hayo.com/entry/1209

Revel: Upload pictures to generate dynamic avatars

👉 https://www.hayo.com/entry/1907

Animated Drawing: Let the children's graffiti move

👉 https://www.hayo.com/entry/1912

Segment Anything: One-click segmentation of various elements in the picture

👉 https://www.hayo.com/entry/1951

A thousand common questions: Alibaba develops large models

👉 https://www.hayo.com/entry/1970

This week, actions such as OpenAI's title ban, temporary suspension of registration, and suspension of paid channels have made friends who are accustomed to using ChatGPT somewhat uneasy. Although it is not "overnight", it is indeed "between January". Many people have become accustomed to the daily life of super AI such as ChatGPT to improve efficiency. In any case, the times are always developing, and how to treat and use AI has become a topic that people have to face now.

Because it's there.