[ad_1]
“The amazing thing about these great languages is that they know how they work just to read everything they find,” he says. Chris Manning, a Stanford professor who is well known for his AI and language.
But GPT and its species are the most skilled parrots. They learn how to make grammar and grammar in the language. This means that he can produce nonsense, true errors, and obscene words removed from the darkest corners of the internet.
Amnon Shashua, a professor of computer science at the Hebrew University in Jerusalem, is the founder of a new bilingual AI project. He knows one or two things about selling AI, after selling his last company, Mobileye, who pioneered using AI to help vehicles see things on the road, that Intel in 2017 for $ 15.3 billion.
New Shashua Company, AI21, which came out in secret last week, has developed an AI-based communication system, called Jurassic-1, which demonstrates fluency in English and Hebrew.
In demos, Jurassic-1 is able to create word segments on a given topic, dreaming about major blogging topics, composing small computer essays, and much more. Shashua says the genre is much more advanced than the GPT-3, and he believes that future Jurassic species could have a world-class knowledge based on their findings.
Another effort to re-establish GPT-3 is exposing various languages in the world as well as the internet. In April, researchers at Huawei, a Chinese giant, in detail of a Chinese GPT-like language called PanGu-alpha (spelled PanGu-α). In May, Naver, a research giant in South Korea, said they had created their own language group, called HyperCLOVA, which “speaks” in Korea.
Jie Tang, a professor at Tsinghua University, leads the team at Beijing Academy of Artificial Intelligence which created a Chinese brand called Wudao (meaning “illuminated”) with the help of governments and corporations.
Wudao’s color is bigger than anyone else, meaning its neural testing machines spread across most cloud computers. Increasing the size of the internet network was important in making GPT-2 and -3 possible. Wudao can also work with photographs and recordings, and Tang set up a company to sell it. “We believe this could be the cornerstone of all AI,” Tang said.
Such interest seems to justify the potential of these new AI applications, but the competition for language marketing can also go much faster than attempts to increase or decrease misuse.
Perhaps the biggest concern of AI languages is how they can be misused. Because these colors are capable of producing satisfying words in the head, some people worry that they may be used to create illusions, spam, or false stories.
“I would be surprised if disinformation manufacturers do not use experimental power with these types,” he says Mika Musser, a research researcher at the University of Georgetown who studied the ability of linguistic species to spread lies.
Musser says research shows that it is not possible to use AI to access AI-generated information. There is no guarantee that you will have enough information in the tweet for the machine to see if it was typed by the machine.
Some of the more complex species may be hiding within the larger species of languages, too. Research has shown that the types of languages taught on Chinese Internet content will show a review it did the same thing. These programs also help to uncover and hide identity, race, and age in the language they eat, as well as hate speech and hate speech.
Similarly, these major languages may fail in unexpected or unexpected ways, he adds. Percy Liang, a professor of computer science at Stanford and a senior researcher at a new place dedicated to studying the potential of powerful AI models such as GPT-3.
[ad_2]
Source link
