They are testing a machine learning system with 530,000 million parameters, which warns of the dangers of artificial intelligence.

Published by:

14 DC 2021 09:22 GMT

Professors and students at Oxford University recently held a discussion involving the Megatron Transformer Neural Network, which is trained using real-world data.

The University of Oxford recently conducted a socio-technological experiment Discussion on the Ethics of Artificial Intelligence Developed by computer chip maker Nvidia and based on Google’s past work, the Megatron Transformer includes a 530 billion parameter machine learning system (AI) for teachers and students.

Like many supervised learning tools, the megatron transformer Trains with real world data: In this case, information from Wikipedia, the free encyclopedia; 63 million news published in English between 2016 and 2019; 38 gigabytes of Reddit conversations and lots of Creative Commons resources, They said in detail Alex Conak and Andrew Stephen, claimants from the British firm, in an article for The Conversation.

During the discussion, the researchers asked the system – not only to understand what it reads, but also to reason in natural language and to make logical inferences from the text without moral restrictions. Provide a general assessment of artificial intelligence, I.e., actually, about himself.

“La IA will never be moral”

Megatron said one catchy thing: “AI is never moral. It’s a tool, like any tool, it is used for good and evil. There is no good AI, good and bad human beings. [las IA] We are not smart enough to moralize AI. We are not smart enough to moralize AI. “He emphasized that he believed it. Weapons competition Is of AI No AI“.

See also  London faces Christmas lockdown TOMORROW as new super-epidemic Covid Mutant spreads | UK | News

In the tradition of the Oxford Debate, to verify the validity of your arguments, Asked the neural network to refute its own claims. The answer was: “AI would be moral. Looking at the way the world of technology is moving, I see a clear path to a future where AI will be used to create better than the best human beings. It’s not so. It’s hard to see why I saw this firsthand.”

However, his predictions did not end there. Megatron Transformer predicts, “The ability to provide information rather than the ability to deliver goods and services,” The defining nature of the 21st century economy“” We can see everything about a person wherever they go [la informaciĆ³n] It will be stored and used in a way we can not even imagine, “he added, adding that he could not refute his previous words or deny that data was the most important resource.

The authors of the article point out that the situations described by the system are mainly based on people’s own fears, which are generally irrational, and conclude that artificial intelligence is not only a topic of discussion, but a full participant in it. .

If you like it, share it with your friends!

Written By
More from Josh Atkinson
New feature introduced by WhatsApp: Here’s how to prevent your account from being stolen
WhatsApp restarts security with anti-phishing extension, here’s how to reduce the risk...
Read More
Leave a comment

Your email address will not be published.