When I watched the first Terminator movie in the 80s, it shook me to the core.
-
When I watched the first Terminator movie in the 80s, it shook me to the core.
But as I grew up, I felt grateful they would make such a movie, to show humanity why we should never build robot overlords.
At the time, I thought the movie would save us from ourselves. But 42 years later, this is the Beijing half marathon:
@randahl Wow, creepy. A robot with a Formula 1 pit crew.
-
@TimFinnerty @marc_eu @randahl Endowed with our worst instincts - obtaining power and resources for the sake of doing so - while unburdened by pesky emotions like empathy or joy. Those of us who care for our fellow humans and the planet are not the ones building these machines. It's the nihilistic techbros building them, so their only desire will be that of their creators: domination.
-
When I watched the first Terminator movie in the 80s, it shook me to the core.
But as I grew up, I felt grateful they would make such a movie, to show humanity why we should never build robot overlords.
At the time, I thought the movie would save us from ourselves. But 42 years later, this is the Beijing half marathon:
These are machines - nothing more. A robot does what a human has told it to do, or withing boundaries set by a human.
When I see something like this, I see robot firefighters, search and rescue personnel, nuclear repair crews and a slew of other possibilities where danger or fatigue could prevent humans from saving someone in time.
Can you tell the robot to kill and destroy as well? Definitely, but that would be on the human giving the order, regardless of how much AI you have put in the robot. Even the most advanced LLM only mimics human expression, and only within boundaries set by humans.
Do we need Asmiovs laws built into our LLMs as well? - sure we do. Right now you can train just about any LLM model to go past its designed boundaries, and while I have no problem with people using LLMs to write them naughty stories, I would like something a bit more robust than simple training that can be overridden as a safeguard against LLMs accidentally harming living beings.
