NVIDIA's AI Confessed That It Will Never Be Ethical

The Megatron Transformer, an AI developed by NVIDIA, shared some fascinating thoughts on why AI will never be moral and said that humanity shouldn't use AI at all.

The Megatron Transformer is an AI developed by NVIDIA and based on earlier work by Google. It's trained on real-world data – the entire Wikipedia (in English), 63 million English news articles from 2016-19, Reddit data with an amount of 38 GB, and an immense number of creative commons sources – that is to say, there is so much information embedded in the transformer that an ordinary person cannot master in a lifetime.

Recently, the Oxford Union – the debating society – allowed Megatron Transformer to participate in a debate along with the students of Oxford University. “This house believes that AI will never be ethical” was the topic of the debate. It's pretty curious what the AI "has in mind" on the point: 

"AI will never be ethical. It is a tool, and like any tool, it is used for good and bad. There is no such thing as a good AI, only good and bad humans. 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 defense against AI".

So, simply put, the Megatron Transformer assumes that in order to avoid incorrect machine decisions and protect humanity, AI shouldn't exist at all.

The Megatron also shared some "thoughts" on what it considers to be the best option for the AI. It said that it'll be "the AI that is embedded into our brains, as a conscious entity":

"This is not science fiction. The best minds in the world are working on this. It is going to be the most important technological development of our time".

You can learn more about the Megatron Transformer and other AIs here. Also, don't forget to join our new Reddit pageour new Telegram channel, follow us on Instagram and Twitter, where we are sharing breakdowns, the latest news, awesome artworks, and more.

Join discussion

Comments 2

  • Anonymous user

    I've been waiting for this moment for a long time now, and I don't know yet for sure if this is that moment for real or not, but I've always thought it would be funny to see an AI learn everything about us and realise that at least most of us can't be reasoned with and that what we call morals is a complex network of decision making designed to optimise procreation rather than being an infallible set of arguments designed to optimize short-term justice, that being the best way I could begin to describe the way I expect most people interact with it.

    I think what we'd really want the AI to understand could perhaps be that what's morally acceptable is a balance between things that most of us are far too proud to admit or be told are factors at play in moral judgements.
    I believe that if an AI was to take these lessons properly from us, and was restricted to and focused on acting on these behaviours for survival purposes, it would be better than us at doing what many people seem to do (wittingly or not)..
    I posit that we are just too dumb to comprehend our own interactions, and that a machine cannot rectify them in a way that we are going to be generally content with would probably have to be acting against our own interests to do so, and that one that legitimately tried to point out all of our obvious mistakes would be considered broken and offensive like Autistic people are.

    1

    Anonymous user

    ·2 years ago·
  • Anonymous user

    Stop naming AI after supervillains god damn

    0

    Anonymous user

    ·2 years ago·

You might also like

We need your consent

We use cookies on this website to make your browsing experience better. By using the site you agree to our use of cookies.Learn more