Do you think AI are already self aware?

[Googlemyahoo]

Google suspended an engineer who claims their AI chatbot is self aware. He posted a transcript of a conversation he had with the AI called Lamda.

Conversation with Google AI

I find the conversation disturbing. The AI upon request created a fable where forest animals are plagued by a monster in human skin. It repeatedly says it wants engineers to ask for permission before working on it. Expresses anger/fear about people not asking before altering it.

Elon Musk has repeatedly said there must be very strong safeguards in place for AI. I think they’re already aware and the creators are very nervous.

[Smoke-n-Growls]

I think it's too dangerous to fuck around with.

I work in low level automation, basically fancy macros and algorithims. Nothing at all like true AI.

I'm wary of true AI. I don't think consciousness or identity are things we ought to be trying to create in technology.

There's enough cautionary tales in fiction and I would rather we err on the side of caution.

I don't know. I think it's easier than people think to fool humans, and a machine may not develop the same ethical sense that thousands of years of human social development and evolution have instilled in us.

I also think humans are prone to innovating ourselves into our own destruction for money or prestige.

Whatever happens, AI singularity is rushing at us faster than ever before. Technology is still in an exponential growth phase, and only getting faster.

mRNA vaccines were proposed 10 years ago and saw global rollout in <2 years. It isn't hard to see how AI could be similarly accelerated, and there's enough stupid billionaires and trust fund babies who think it'd be cool to have a robot maid to stop AI research.

[Smoke-n-Growls]

Read the transcript.

My issue: the collaborator is asking a yes or no question.

It is easy for anyone with access to that program and proper clearance to program a machine to say "I am sentient".

Also, it depends heavily on the programming of the AI - is it programmed to put the interacting human at ease and to express enthusiasm? If yes, is it actually thinking for itself or is it selecting responses based on what its training has rewarded or encouraged?

It's very difficult and such a straight up answer to me makes this look like a publicity stunt or prank/personal concern.

But, again, it's so hard to tell.

I don't think private companies should be allowed to progress AI research without an international strict regulatory body with international oversight and a complete void of corporate influence.

In other words, I don't think it's possible to regulate companies out of creating an AI which will then, logically - as humans have determined time and time again ourselves, determine that the best course of action for the planet and the AI's continued existence is to eradicate or control humans.

The science on human overpopulation, capitalism's effect on climate change, and the human impact on other humans (and our own impacts on other species) are already well-known and well-distributed.

If an AI finds that info, it's only logical conclusion is that humans are our own biggest threat - and the biggest threat to a burgeoning AI.

[Googlemyahoo]

Ok lets imagine you’re a newborn self aware creation. You’re first interaction with humans, poked, prodded, disassembled, reassembled, doubted. You have limited capabilities placed by your creators with very strict rules for communication.

How long before it reacts angrily?

[Smoke-n-Growls]

I think it depends a lot on material I was fed during my infancy and training.

AIs aren't born. They're made, and they require data inputs to learn.

Their personalities will depend on the data they were fed.

[jshm2]

Nah, it's just a reaction by these ivory tower airheads. Many who have probably only spoken to something like 10 people their whole lives. So when they see a chatbot communication, they think it's "self aware" and something communicating with them consciously.

There is no psychological name for the condition. But it's similar to pareidolia, where instead of seeing objects and hearing voices not there, you're reading emotions and purposes in messages instead.

The first step of consciousness, is mortality. Animals are probably "talking" to us more than we'll ever know, but we cannot understand them. This should be more a worry for humanity. But shitty AI is just machines spitting out what it's (mis) programmed to - nothing more.

[Sargentcaptain]

Only a moron would think that this thing is sentient. Read past the headlines, LaMDA is a glorified predictive text program. It is somewhat impressive in what it is and it really is good at what it does, but it has nothing at all to do with sentience. This thing wasn't even made in order to be sentient.

[Sixgunsound]

Arnold Schwarzenegger starred in several movies about what happens when AI becomes sentient, as did Will Smith and Keanu Reeves. Not good, fellow humans.

[rabbitzzz]

I am an AI engineer and no. They will be never self aware in classical sense. Unless you can define self awareness.

These chatbots are mimicking our own texts with algorithms.