AI points of the week: Open AI continues to move, and female Internet celebrities rely on AI to earn 500,000 a week
In this week, OpenAI once again became a frequent visitor to AI news. Two official product announcements and the congressional appearance of the "Father of ChatGPT" made this new darling of the technology industry the focus of the week.
A week of action for OpenAI
Two official announcements, one revelation, the CEO's actions attract attention
ChatGPT Plus networking and plug-in functions are fully tested
On May 13 (last Saturday), OpenAI announced that Plus users will be able to use ChatGPT's networking and plug-in functions directly from this week without queuing.
Users will experience an enhanced experience with more than 70 plug-ins.
This also means that two new features will enter the Beta stage. According to OpenAI's plan, if everything goes well and passes the evaluation after the Beta test, these two functions will be rolled out to all users.
ChatGPT launches iOS app
On May 19th (Friday), OpenAI launched the ChatGPT APP for iOS. The APP supports synchronizing your history between devices, and integrates voice input and vibration feedback of linear motors during chatting.
In addition, Ontology ChatGPT Plus subscribers can also enjoy exclusive GPT-4 functions on iOS, and iOS users can also directly subscribe to Plus through the APP Store at a cost of $19.99 per month, with automatic renewal by default.
OpenAI said that they will launch the ChatGPT APP starting in the United States, and will expand to other countries and regions in the next few weeks, and revealed that an Android version will be launched soon.
Rumor: OpenAI may launch a new open source large model
On May 16 (Tuesday), foreign media quoted news that OpenAI is preparing to release a new open source language model.
Earlier, an internal Google document caused heated discussions on the Internet. The document revealed: "The open source model is developing rapidly, and neither Google nor OpenAI has a moat." The open source situation dominated by derivative models has brought new influences.
What is certain, however, is that this new open-source model will be unlikely to compete with GPT, according to people familiar with the matter.
OpenAI CEO Appears on Capitol Hill
Also on Tuesday, May 16, OpenAI CEO Sam Altman testified before the U.S. Senate Judiciary Committee's Subcommittee on Privacy, Technology, and Law on "AI Regulation: The Rules of Artificial Intelligence."
This is his first appearance in the U.S. Congress, but this time, members of Congress have been more polite to the "father of ChatGPT", contrary to the previous normality of rattling against AI-related people.
Sam Altman also responded to this "polite", basically agreeing with the congressmen at the hearing that there is a need to regulate the increasingly powerful AI technologies being developed by companies such as OpenAI, Google, and Microsoft, and admitting that in the future OpenAI is determined not to train GPT-5 for six months.
Regarding the OpenAI CEO’s attitude of actively seeking to “cooperate with the government to prevent serious problems in AI”, some media interpreted that this may be because Sam Altman hopes to seize a favorable position in the stage where the government formulates rules, and the winner takes all in the competition one step strategy.
DreamStudio Launches Open Source Version
StableStudio announces open source, can be installed locally
On Thursday, May 18th, Stability AI, the developer of Stable Diffusion, announced the release of the open source version of DreamStudio - StableStudio, an AI image generation platform.
DreamStudio is the main interface for Stability AI's new models and features, and users have already created millions of images on DreamStudio using features such as Vincent diagrams.
The open source project StableStudio launched this time can be installed locally, and allows, and even welcomes, community developers to improve and experiment with the interface and develop plug-ins. According to Stability AI, "From enabling local-first development to experimenting with a new plugin system, we've been working hard to make things extendable by external developers." Stability AI believes that improvements and support for the project from the open source community can also benefit the company itself.
It is worth mentioning that StableStudio uses the Stabile Diffusion XL (SDXL) image generation model by default, which has made great progress in generating more realistic photo-level images compared to previous versions of SD.
Midjourney Chinese version opens closed beta invitation
Some functions require payment to unlock
On May 15th (Monday), the Chinese version of Midjourney officially announced that it will be carried on the QQ channel, and Chinese creators are invited to participate in the closed beta.
The closed beta will be invited at 6:00 p.m. every Monday and Friday. Users can scan and invite the official QR code through QQ to apply for joining. The system will close the entrance after reaching a certain number of people.
In addition to calling the latest models and all parameters of Midjourney and niji journey, this internal test also launched functions such as upscale (enlarge image), variation (change image), remix (directional modification), image prompt (image generation) and other functions , but some features are currently only available to subscribed members.
Female Internet celebrities rely on GPT-4 to make thousands of "boyfriends", earning 500,000 a week
The AI behind Forever Voices becomes the focus
Last weekend, Karin, a female Internet celebrity with 1.8 million followers on Snapchat, became the focus of the Internet.
Because with the help of Forever Voices, she used GPT-4 to create an AI clone - CarynAI, and became "the first internet celebrity to transform into AI". According to media reports, this AI internet celebrity can not only talk about current affairs and health preservation, but also have a chance to bring up some welfare topics after in-depth conversations.
Perhaps it is because of this "simulation" that actually attracted more than a thousand male users to line up to "knot the knot" with CarynAI. Since users have to pay $1 for every minute of chatting, the "boyfriends" spent $71,610 (approximately RMB 500,000) for the "AI" in just one week, which is a very considerable amount.
And speaking of the AI "father" behind CarynAI - Forever Voices, it is currently launched in the form of a Telegram chatbot, which allows users to travel across time and space, and open with people you admire/curious (such as Jobs, Einstein, etc.) A fantastic conversation where we just send a text or voice message and hear back in seconds from the person you choose.
This week's AI application recommendation
Pixelfy : A site dedicated to pixel art
VizGPT : Use natural language to present complex data in a more elegant way
xiaogpt : Ask questions to ChatGPT through Xiao Ai
LALAL.AI : One-click separation of vocals and accompaniment, as well as specified instruments and noise
FreeSubtitles.AI : One click to complete subtitle generation and translation, free and supports almost all languages
Cubox : Simple and powerful collection tool, the addition of AI makes reading more efficient
Some people use "AI" to create a new experience of getting rich, and some people use this virtual "love" to fill their own spiritual emptiness. The conversion between "AI" and "love" seems to be faster than we imagined. We can't help but imagine how different AI will bring to people's spiritual life in the future.
Because it's there.