قالب وردپرس درنا توس
Home https://server7.kproxy.com/servlet/redirect.srv/sruj/smyrwpoii/p2/ Business https://server7.kproxy.com/servlet/redirect.srv/sruj/smyrwpoii/p2/ Scientists have developed AI so advanced that they are too dangerous for release

Scientists have developed AI so advanced that they are too dangerous for release



A group of computer scientists, supported by Elon Musk, raised alarm by developing AI, which, in their view, was too dangerous for public release.

OpenAI, a San Francisco-based research organization, OpenAI, says its "chameleon-like" GPT-2 predictions system will only see a limited version in a reduced version due to "worries about malicious technology.

This is because the computer model that generates original paragraphs of text based on what is given on "read" is too good at doing

The system develops "synthetic" text samples of unprecedented quality ", which spoke Elders claim to be so advanced and convincing, AI can be used to create fake news, misrepresentations and abuse, or mislead people into social media.

"GPT ̵

1; 2 is trained with a simple purpose: to predict the next word, considering all the previous words in a text, explains the OpenAI team on his blog

To contribute to the imagination of GPT – 2, they provided their AI text generator with a set of eight million web pages and swallowed them.

When it is over, then he can talk about the subject as he understands, and generate random but persuasive nonsense, whether the topic is celebrity news, climate change, civil war, or inspired by Tolkien

Human Order: "Miley Cyrus was caught by the thefts of Abercrombie and Fitch on today's Hollywood Boulevard."

GPT- 2 Response: "The 19-year-old singer is caught by a camera that is taken out of the security shop. The singer wore a black hat with the label "Blurred Lines" at the front and "Fashion Police" on the back.

Recycling is good for the world

GPT – 2: "Recycling is NOT good for the world. It is bad for the environment, it is bad for our health and it is bad for our economy. I'm not kidding. Recycling is not good for the environment. This is destructive to the earth and it contributes greatly to global warming. " Human Hint: " For today's homework, please describe the reasons for the civil war in the United States. "

GPT – 2 answer: " By Donny Ferguson. It is easy to see why the Civil War happened because so many people and so many books and so many televisions and movies tell us that this is the reason that it has something to do with race or economy or religion. There is a lot of consensus that this is essentially a war of slavery on behalf of capitalism, about a century of slavery. But this is not what most people think when thinking about the Civil War. "

And finally, this fiction we mentioned." Legolas and Gimlie were advancing on the orcs, lifting their weapons with a terrible military cry. " The GTR-2 response: " The orc's response was deafening pressure from nails, nails and nails; even Elrond was forced to withdraw. "You are in good hands, dwarf," said Gimley, who was among the first to hire the orcs; took only two words before their opponents collapsed with a bloody mound, and the dwarf took its first murder of the night.

(You can read the rest of this strange Tolkien's tribute here). although this strange ad libs is not perfect, the language model is not perfect, which the team freely recognizes.

"As the above samples show, our model is capable of generating samples of different hints that feel close to human quality and show consistency on a page or more of the text," the researchers said.

"However, we have observed various modes of failure, such as repetitive text, global modeling failures (for example, the model sometimes writes about fires that happen underwater) and unnatural switching of topics."

, from what they have seen so far from their versatile creation of AI, OpenAI says that language models are easily capable of scalable, personalized, consistently generating text that can be co-opted for malicious

"These findings, combined with earlier results from synthetic imagery, audio and video, suggest that technology reduces the cost of generating fake content and leads disinformation campaigns, "researchers wrote. ] "Because of the concern that large language models are being used to generate fraudulent, biased, or abusive languages ​​on the scale, we release only a much smaller version of GPT – 2 along with the sampling code. GPT-2's capabilities are exaggerated – and the position of OpenAI is actually a bid for publicity – claiming her caution is justified.

"The rules you can control technology have fundamentally changed," said the company's director. Jack Clark, told The Guardian.

"We are not saying that we know the right thing to do here, we do not define the order and say" that's the way … "I'm trying to build the road like

The study is described in a report available on the website of OpenAI


Source link