“ChatGPT is flawed because we are all humans and ChatGPT is trained on human data.” - this is an interesting excuse for why ChatGPT is so stupid.
For many years now I’ve been working with “AI”. In the early 1980s I wrote software to be able to chat with a computer and have the computer learn from talking with people. Of course the hardware available at that time, to a private individual, meant this was extremely limited in it’s ability. What shocks me is how little we have progressed in the past forty years. Almost all current “AI” has no dynamic learning ability.
AI of course means many things to many people. Unfortunately there is nothing intelligent about it.
Clearly in this example, ChatGPT has no idea at all that you can’t walk across the English Channel.
There are two aspects of intelligence that are extremely important: learning and creative thought. Most of what people call AI is nothing more than faking such that a computer appears to be human like.
Voice to text and text to voice is amazing, but it’s not intelligent at all. It gives us the illusion of talking with a computer. Most chat bots that are essentially very fancy menu systems, those aren’t intelligent and they can be very infuriating (depending on how well they were designed and what you need).
Here I told Stable Diffusion to create an image of a human talking with a robot. Maybe AI is arrogant and thinks only robots should talk to other robots? Or it doesn’t know what a human is?
Software like Stable Diffusion can create fake images. Work with it a bit and you’ll quickly curse that it’s not intelligent at all. Amazing yes, but not intelligent.
Currently I’m taking a course on “Prompt Engineering”. Yes, we are now studying how to talk to “AI” because AI is too stupid to understand us. Is this progress? Teaching humans how to communicate better? Well, the cynical side of me would say yes, because for all our brilliant intelligence, humans have been notoriously bad at communicating. Although I wonder if we are doing it wrong. Stable Diffusion doesn’t understand commas. Instead of teaching humans to not use commas, wouldn’t it be better to teach the AI how to understand commas?
Sci-Fi has had fun scaring us with AI for many decades: “2001: A Space Odyssey”, “Westworld”, “WarGames”, “Terminator” and many other movies. Should we be afraid of AI?
First, let me repeat: humans have not yet created an intelligent machine / computer.
What we should be afraid of is machines making decisions when the machines are not intelligent! Also we should be afraid of ignorant humans trusting computers because they think the computer is intelligent. “AI” is a great way to push propaganda and have foolish humans believe it.
The future is coming, probably not as fast as most people think (ignore the hype), perhaps faster than I think. Overall, my thoughts are that humans are our own worst enemy, we are a far bigger threat to ourselves than AI is.
Human stupidity astounds me, for decades I’ve argued against adding new features when the old software isn’t reliable. Will we continue to make this mistake and build robots with tons of malfunctioning features?
Perhaps if you see two robots having a conversation over a cup of coffee, then maybe it’s time to be afraid?
Artificial imitability.