Editor’s note: Generative artificial intelligence is the new darling of the Silicon Valley investment community, and every investor wants to participate in this big event as soon as possible. Among all the generative artificial intelligence models, Stable Diffusion stands out because it is open source and has almost no restrictions on generating pictures. Some people question that this approach is too radical, and imad Mustaq, founder and CEO of Stability AI, believes that the development of artificial intelligence is changing with each passing day, and open source is the best way to use AI. This article comes from compilation, hoping to inspire you.
On October 19th, 2022, imad Mustaq, founder and CEO of the startup Stability AI, was born in San Francisco. Stability AI is a start-up company that has developed a controversial Stable Diffusion image generator. The company’s celebrations also represent the arrival of a new upsurge in artificial intelligence. (Jason Henry/The New York Times)
In Silicon Valley, cryptocurrency and metauniverse are out of date, while generative artificial intelligence is on the rise.
This was verified in the San Francisco exploration museum on October 17th, 2022. There, Stability AI, a startup behind the AI mapping software Stable Diffusion, held a party, which felt like returning to the prosperity before the epidemic.
The event attracted technology celebrities including Sergey Brin, co-founder of Google, Naval Ravikant, founder of AngelList, and Ron Conway, a venture capitalist. The event was promoted as the release party of Stability AI, and it was also a celebration of the company’s recent $101 million financing. It was reported that this round of financing valued the company at $1 billion.
But it is also the carnival of the whole field of Generative Artificial Intelligence. Generative artificial intelligence is a general term, which can not only analyze existing data, but also generate new texts, images, videos, code fragments and so on.
This is an important year for the generative artificial intelligence application that "transforms text prompts into images". Unlike tokens or virtual reality metauniverses, generative artificial intelligence applications actually have enough numbers to prove their spread. OpenAI released the image generator DALL-E 2 this year. According to the company, more than 1.5 million users have generated more than 2 million images every day. Midjourney is another popular artificial intelligence image generator released this year, and its official Discord server has more than 3 million users. Google and Meta have also built their own image generators, but they have not been released to the public.
Such growth opportunities have triggered a crazy investment boom among investors, who hope to participate in the next big event as soon as possible. Jasper is an artificial intelligence copywriting application developed for marketers. It has been developed for one year and recently raised $125 million with a valuation of $1.5 billion. Startups have raised millions of dollars to apply generative artificial intelligence to games, programming and advertising. Sequoia Capital, a venture capital company, recently said in a blog post that it believes that generative artificial intelligence can create "trillions of dollars in economic value".
However, no generative artificial intelligence project has caused so much controversy as Stable Diffusion.
This is partly because, unlike many generative artificial intelligence projects, Stable Diffusion is open source and can be used for free, which means that anyone can view or download the code and run the modified version on a personal computer. According to the company, since its release in August, more than 200,000 people have downloaded this code, and millions of images have been created using tools based on the Stable Diffusion algorithm.
This laissez-faire approach also extends to the image itself. Other artificial intelligence image generators have strict rules to prevent users from creating violent, pornographic or infringing images. In contrast, Stable Diffusion has only one basic security filter, which can be easily bypassed by any user who creates his own application.
This freedom makes Stable Diffusion very popular among underground artists and expression pack makers. However, it has also aroused widespread concern that the company’s loose rules may lead to the spread of a large number of violent pictures, naked pictures, false propaganda pictures and misinformation pictures.
At present, Stable Diffusion and its open source subroutines have been used to produce a large number of disgusting pictures. According to a quick scan of Twitter, there are a surprising number of pornographic and anime works on the platform. Recently, several forums on Reddit were shut down for spreading a large number of nude photos without permission, most of which were made with Stable Diffusion. The company tried to control this chaos, telling users not to "generate anything that makes you ashamed to show your mother", but did not set stricter filters.
U.S. Congresswoman Anna Eshoo of California recently sent a letter to federal regulators, warning that someone used Stable Diffusion to make images of "Asian women who were beaten by violence". Eshaw urged regulators to crack down on "unsafe" open source artificial intelligence models.
Imad Mustaq, founder and CEO of Stability AI, hit back at the idea of content restrictions. He believes that complete freedom is necessary to realize his vision (popular artificial intelligence that is not affected by enterprises).
In a recent interview, he reiterated this view and compared his view with what he called "the tough and paternalistic approach taken by technology giants to artificial intelligence".
"We trust people, we trust communities," he said, "instead of letting a technology giant control the most powerful technology in the world."
Mustaq, 39, is an eccentric leader in the generative artificial intelligence industry. He doesn’t have a doctorate in artificial intelligence, nor has he worked in a large technology company where artificial intelligence projects such as Google or OpenAI are usually located. He is a former British hedge fund manager, who has been engaged in oil trading for most of the past 10 years and has provided consulting services to enterprises and governments on the Middle East strategy and the threat of Islamic extremism. Recently, he organized an alliance of think tanks and technical groups to try to use big data to help the government make better decisions on COVID-19.
Mustaq initially personally funded Stability AI, but he soon became a polarized figure in the field of artificial intelligence. Researchers and executives of larger and more traditional artificial intelligence organizations believe that Mustaq’s open source approach is naive or reckless. Some people worry that the unrestricted release of open source generative artificial intelligence models may trigger opposition from regulators and the public, thus damaging the whole industry.
But on October 17th, Mustaq won applause from hundreds of artificial intelligence researchers, social media executives and technology Twitter celebrities.
He made many implicit attacks on technology giants such as Google and OpenAI funded by Microsoft. He denounced "targeted advertising" as the core of the business models of Google and Facebook, calling it "manipulation technology". He also said that unlike these companies, Stability AI will not establish a "panoramic prison" to monitor users.
Mustaq also announced that in 2023, the number of computers used by the company to train the best models will be expanded to 5 to 10 times of the current number. With more than 5,000 high-performance graphics cards, this computer is already one of the largest supercomputers in the world. This strength will enable the company to generate images from artificial intelligence, expand to generate files in video, audio and other formats, and make it easier for users around the world to operate localized versions of its algorithms.
Some critics of artificial intelligence worry that this technology may make artists and other creative workers lose their jobs. Unlike them, Mustaq believes that putting the generative artificial intelligence into the hands of billions of people will bring explosive growth of new opportunities.
He said: "Many people in the world are struggling to create, and we want them to inspire generate."
If all this sounds familiar, it is because Mustaq’s ideas echo the utopian dreams of the founders of the previous generation of technology, such as Mark Zuckerberg of Facebook and Jack Dorsey of Twitter. These people have also competed to send powerful new technologies to billions of people, and almost never stopped to consider the possible harm.
When I asked Mustaq if he was worried about releasing artificial intelligence into the world before it was safe, he said he wasn’t. He said that artificial intelligence is developing so fast that the safest way is to make it public, so that the public (rather than large technology companies) can decide how to manage artificial intelligence.
Ultimately, he said, transparency (rather than top-down control) will prevent generative artificial intelligence from becoming a dangerous force.
"You can query the data set, you can access the model, you can review the code of Stable Diffusion and other things we are doing," he said. "Besides, we can see that it has been improved."
Mustaq’s idea of an open source artificial intelligence utopia may seem unrealistic, but at the party on October 17th, he found that many people wanted to make it a reality.
"You can’t put the genie back in the bottle," said Peter Wang, an executive from Austin, Texas, who came to the party. "You can at least show Austin the genie."
Translator: Jane