Days after Tesla CEO Elon Musk called artificial intelligence (AI) a major threat, Facebook shut down one of its AI programs after chatbots started speaking in their own language, which used English words but could not be understood by people. According to a Tech Times report on Sunday, the social media giant had to pull a plug on the AI system whose researchers were working "because things got out of hand". The problem was, when the robots were rewarded for negotiating, they were not rewarded for negotiating in English, which caused the robots to develop their language.
Facebook researchers shut down AI bots that began speaking in a language that was not understood by humans
Facebook founder Mark Zuckerberg.
"AI did not start shutting down computers around the world or anything like that, but stopped using English and started using the language it invented," the report notes. At first AI agents used English to communicate but later created a new language that could only be understood by AI systems, thus, contradicting their purpose. This has led to Facebook researchers shutting down AI programs and forcing them to speak only in English.
In June, researchers at the Facebook AI Research Lab (FAIR) discovered that while they were busy trying to develop chatbots, "dialogue agents" created their own language. Soon, bots began to deviate from written processes and began to communicate in a completely new language they had created without human input, media reports said. Using machine learning algorithms, "chat agents" were allowed to communicate freely in an effort to strengthen their communication skills.
Researchers also found these bots to be "extremely clever negotiators". "After learning to negotiate, bots rely on machine learning and advanced techniques in an effort to improve the outcome of these negotiations," the report said. "Over time, bots become proficient in it and begin to pretend to be interested in one thing in order to 'give it' over time in conversations as a false adjournment," he added.
While this appears to be a major jump to AI, several experts including Professor Stephen Hawking have raised fears that humans, limited by slow evolution, may be replaced by AI. Others like Tesla's Elon Musk, philanthropist Bill Gates and former Apple founder Steve Wozniak have also expressed concern about where AI technology is headed. Interestingly, the incident took place a few days after a verbal dispute between the CEO of Facebook and Musk exchanged bitter words over a debate about the future of AI.
"I have spoken to Mark about this (AI). His understanding of the subject is limited," Musk said in a tweet last week. The tweet comes after Zuckerberg, during a live broadcast on Facebook earlier this month, criticized Musk for arguing that care and control is needed to protect the future if AI becomes a priority. "I think it's irrational people and they try to do these doomsday scenes - I just, I don't understand it. It's really bad and somehow I think it's irresponsible," Zuckerberg said.
Musk has spoken extensively on AI and called its progress "a great danger to us as a civilization". "AI is a rare situation where we need to be involved in governance instead of working hard because if we work hard on AI regulation it will be too late," he said.
0 Comments
If your having doubt please let me know