It doesn't matter how much data they collect. It will never be enough.
Source: I worked at a very large corporation on various AI / ML technologies, and we threw millions of machines at some of these problems. Bonus points we had access to virtually unlimited data. Despite all that, our models were only something like 80% accurate. Imagine the peak of AI, all the world's computers, only being able to see a dog 80% of the time, when toddlers can do it 98% of the time.
But you are right: Stop engaging with AI in any form. Just walk away. The only reason people are doing it is because it attracts attention, and attention attracts investors, and investors create hype.
Amazon's mechanical turk is still better than the best AI. Even the lowest class members of India do a better job at menial tasks than the best AI ever will.
This word "AI" only has power because we think it has power. It, like "3D TV", is nothing without someone willing to pony up cash to sing in the choir.
i disagree. not with the sentiment about calling 'artificial intelligence' but that i prefer a bit of code running a machine than any one or thing from india trying to grift me something.
I once thought of trying to use the generative AI to write code. But I'd have to go through the code and understand it before I could tell if it's code I'd actually run. And gen AI doesn't create concise and clear solutions, it's often convoluted, so it'd be like reading code written by an incompetent child. That's the kind of code it generates.
It will NEVER replace code hand-written by an intelligent person. We're already seeing this prophecy fulfilled in the software world. People THINK they are more productive, but they are not, because they are creating software they can't understand, they can't modify, and they can't fix.
I don't have any models running on my machine, not for anything. I know how they work, and a good rule-based system is very much superior to the best it can produce. It's way easier to "teach" a system basic rules about stuff than it is to get AI / ML to produce anything useful.
Yep! 100% agree. But how are we supposed to wake up these users if not by making AI look bad? Just talking about abstractions makes them feel dumb then they get emotional and even agressive.
they are at a stage right now where they need the large models to have more engagement from billions of people, soon.
it will be pervasive enough soon to fool most people.
which, imo, if it replaces these indian call centers. im in full support.
but im not willing to have them control energy to run their server farms for it.
lastly, let me add that if you get calls to confirm things for you accounts for banks and things.
have a digitalizer read the data to that machine.
they are training large models to recognize your voice
for your own security and safety, of course.
Source: I worked at a very large corporation on various AI / ML technologies, and we threw millions of machines at some of these problems. Bonus points we had access to virtually unlimited data. Despite all that, our models were only something like 80% accurate. Imagine the peak of AI, all the world's computers, only being able to see a dog 80% of the time, when toddlers can do it 98% of the time.
But you are right: Stop engaging with AI in any form. Just walk away. The only reason people are doing it is because it attracts attention, and attention attracts investors, and investors create hype.
they all need you; we sure as fuck dont need them
This word "AI" only has power because we think it has power. It, like "3D TV", is nothing without someone willing to pony up cash to sing in the choir.
i disagree. not with the sentiment about calling 'artificial intelligence' but that i prefer a bit of code running a machine than any one or thing from india trying to grift me something.
I once thought of trying to use the generative AI to write code. But I'd have to go through the code and understand it before I could tell if it's code I'd actually run. And gen AI doesn't create concise and clear solutions, it's often convoluted, so it'd be like reading code written by an incompetent child. That's the kind of code it generates.
It will NEVER replace code hand-written by an intelligent person. We're already seeing this prophecy fulfilled in the software world. People THINK they are more productive, but they are not, because they are creating software they can't understand, they can't modify, and they can't fix.
I don't have any models running on my machine, not for anything. I know how they work, and a good rule-based system is very much superior to the best it can produce. It's way easier to "teach" a system basic rules about stuff than it is to get AI / ML to produce anything useful.