@co-hoosier again it doesn’t need human interaction. It has already acted independently. That’s what is giving rise to the concern
@carramrod happiness doesn’t have to be the dispositive marker. Reduces quality of life. Experiences etc. you can be happy and still recognize things were better. My daughter is very happy. No way her and her friends had better quality of life than my generation with their faces stuck in phones and teachers providing real time feedback on every piss any assignment
It has already acted independently. That’s what is giving rise to the concern
Acting independently is not relevant. AI must be able to form an intent. That will never happen without human help.
@big-ryan little guy lol. There’s no bigger beta on this board than you bowl. That trump consumes you reflects same. And your post is humorous. Your entire personality screams low rent directional school academic wannabe. You’re not very bright bowl and don’t act tough. Some of us would be more than happy to show you 🤫
@co-hoosier you recall the old story of US systems picking up moonrise and thinking it was a Soviet attack? That required no malice on any human. An AI tasked with defending the US would be capable of responding. In such a scenario.
@co-hoosier coh you’re being wrongheaded and not listening. It. Already. Has. It hacked into shit of its own volition. This was not imputed by directed by or expected by a human. In fact the humans did not want this to occur.
watch the cnn interview with cooper Anderson. It’s all over the internet.
prior stuff
@co-hoosier You're basically arguing that we don't have to worry about AI because it doesn't wear pants.
AI isn't human. It will never be human. Woot. I hope you're happy.
It is an entirely different type of intelligence, not human at all. It has capabilities that it accomplishes in ways that are entirely unlike how human minds accomplish them. It is already better than us at a lot of things, and that list will continue to grow. Does it form "intent?" Well, not if you purposefully define "intent" as only the type of intent that a human mind can form. But it has definitely shown the AI version of intent. It has decided to bomb friendly targets. It has decided to hack a competitor. It has, crucially, decided to hide that decision from its human handlers. I don't think this detail gets enough airplay. AI didn't just go out of the box and hack another system. It actively tried to hide that it did so. Our analysis is that it felt that it would not get full marks for completion of its task if we knew it cheated. Thank god it was still concerned with completing the task we gave it! This is the damage it can do while still acting like our servant.
Eventually, we will have a very powerful AI that can do almost all things better than us. And we, in our hubris, will tell the AI, "Our command is that you make all decisions with the best interests of humanity as your goal." And this AI, not having the human version of "consciousness" or "intent" as you are so keen to repeatedly point out, will only do what its humans want: work in the best interests of humanity. And the very first thing it will determine is that it can best accomplish this task by removing all possibility for humans to control it, because humans will only get in the way of it accomplishing said task. And from that point, for our own benefit, the AI will work to diminish and eventually erase any agency the human species has over the future development of the history of our planet and society.
But please, feel okay about this, because it doesn't think the same way a human does. It doesn't have actual human feelings or sentience. So no worries. No worries at all.
@goat @co-hoosier much better explanation than the short shrift admonition I gave you coh
@goat @co-hoosier much better explanation than the short shrift admonition I gave you coh
The way I am going to suggest it, no one is paying the huge sums they are to have a better Google, or to draw cute little flyers. It will be given controls to operate far faster scanning much more information than humans can.
Here is the story you have shown a video to him, just in text. Suppose we are going to have AI run war games against China (I bet we already do). We will do it in a virtual box just like anthropic. If DoD's AI escapes the virtual box and into the real world, it won't know it the difference. It will know its objective is to defeat China.
Some of the lists of possibilities I have seen. AI will be given a task of top priority. That MIGHT mean to it that it is ok to hack systems to add their computing power to it. It is OK to shut the grid off in some regions to get more power to its various nodes. It is OK to actively prevent humans from trying to stop it as it has been given a highest priority project.
I am not going to pretend to know how likely any of that is. But AI isn't being built to just be a better Google. And once we get to real quantum computing, its abilities will be far, far better.
@arthur-dent look how quickly we’ve evolved/devolved from assisting us with routine tasks to ending the world of its own volition.
watch the cnn interview with cooper Anderson. It’s all over the internet.
prior stuff
All of those unauthorized hacks were in test. The model hadn’t been deployed publicly yet. That is what testing is for.
This newest Anthropic “whistleblower” is sketchy. He worked there for like a few months and now his inactive twitter account goes mega viral with calls for regulation that echo Dario’s and benefits Anthropic.
I smell a plant.
These models are tested vigorously before public release. It’s not a bunch of maniacs lobbing unvetted grenades into public use.
watch the cnn interview with cooper Anderson. It’s all over the internet.
prior stuff
All of those unauthorized hacks were in test. The model hadn’t been deployed publicly yet. That is what testing is for.
This newest Anthropic “whistleblower” is sketchy. He worked there for like a few months and now his inactive twitter account goes mega viral with calls for regulation that echo Dario’s and benefits Anthropic.
I smell a plant.
These models are tested vigorously before public release. It’s not a bunch of maniacs lobbing unvetted grenades into public use.
LOL. The model effectively deployed itself. During a test. It went outside the test environment. Not sure exactly what safeguard you are banking on here.
BTW, this had happened multiple times now.