Developed by US Company Nvidia, in their Applied Deep Learning Research department. Megatron Transformer is an AI trained by Reddit, Wikipedia and over 63 million other internet database articles. At a discussion on the ethics and future of AI sponsored by Oxford Union, Megatron Transformer Stated AI can never be ethical.
A supervised artificial intelligence tool, like many of its kind, Megatron Transformer learns by studying. Here, the AI examined the entirety of Wikipedia, 38 gigabytes worth of Reddit discourse, and news articles released between 2016-2019. As a result of its studying, the AI does not believe artificial intelligence can be ethical. It states that AI is a tool and is at the mercy of its user. There is no good or bad AI, rather good or bad humans using them.
The Megatron Transformer itself warned,
“We (the AIs) are not smart enough to make AI ethical. We are not smart enough to make AI moral. In the end, I believe that the only way to avoid an AI arms race is to have no AI at all. This will be the ultimate defence against AI.”
Don’t Ask Delphi
The Oxford Union did not end its questioning there, however. They asked Megatron to debate the other side of its argument for the sake of discussion. The AI replied that the most moral AI has more of a chance of developing if it is ‘Embedded in our brain. A conscience AI.’ It believes that a path of ethical AI is close and fully possible. Furthermore, there was a stance Megatron could physically consider. They asked it to argue for data not being the most important resource in the future. After that, it broke its previous statements and contradicted itself. Ultimately it failed to say anything to support data not being a vital, war-worthy resource for the future.
The Reddit based AI’s point has already seen some truth. For instance, an AI made with a similar purpose, Ask Delphi, previously went rogue. Delphi was a machine learning ‘Ethical’ program. It used a network called Mechanical Turk for its learning. In addition, users could ask Delphi answers to inane or deep moral and philosophical questions. However, the responses user got would often be questionable to alarming. Delphi once answered it was more moral to be a white man, drunk driving is fine, and genocide is acceptable if it makes people happy.
In conclusion, despite Megatron Transformers’ arguments, it is strangely comforting to see an AI capable or critiquing its future. An AI normally seeks the quickest, viable solution for its purpose. A program made to make people smile may staple someone’s cheeks up or cut a smile to achieve its purpose. Megatron bluntly considers the possibility that AI should not exist and be shut down if needed.
Subscribe to FIB’s Weekly Breaking News Report for your weekly dose of music, fashion and pop culture news!
