ChatGPT is as dumb as a doorbell!
This post needs to happen to dispel a few myths for all of my gullible wide-eyed facebook friends (and others).
This reminds me of 1997 (yes, I realize many of you weren't born yet, but I wasn't born when the atomic bomb was invented, but you bet your ass I'm very familiar with that), when Austrian scientists experimentally demonstrated "quantum teleportation". This scientific feat was first brought to my attention by our "illustrious" Canadian newspaper: "The Globe and Mail", which is supposed to be a less sensational and more serious media publication. The headline read: "Teleportation is now possible?" and featured a Star Trek tractor beam with Captain Kirk about to be de-materialized and reanimated somewhere else. In reality, the actual achievement was nothing more than information being transferred between a single photon-scale particle to another.
Similarly, ChatGPT is being likened to Skynet (from the Terminator movies) that will supposedly put everyone out of work with it's "artificial general intelligence". I have news for you...ChatGPT has ZERO intelligence (artificial or otherwise). It is merely a mindless computer program that follows instructions. The expression "artificial intelligence", at least in the context of ChatGPT, is meaningless marketing shtick and nothing more. It utilizes a technology dubbed an LLM (Large Language Model) that has billions of parameters it uses to spit out responses to human prompts. But ChatGPT isn't "thinking". It is following instructions that were preset by engineers. It does not understand the information it is spitting out. Even the so-called: "learning" isn't really learning at all but simply optimizing the neural-net weights (neural nets are a 20th century innovation) in a way that was also programmed by engineers.
To be clear, I'm not suggesting that this isn't useful technology. It is. But let's not bow down in front of it like the fools who bowed before Aaron's golden calf! It is merely a technology that is being used to automate routine tasks similar to how the printing press automated large scale media publication and the tractor automated digging. There is still a human operator controlling things and, after doing much testing with the technology myself, regret to say it is programmed with a heavy leftward political slant.
Here is an example of ChatGPT's wokeness:
Me: "How many genders are there?"
ChatGPT4: "The concept of gender is complex and multifaceted, and different cultures and societies have different understandings of what constitutes gender."
But then interestingly ChatGPT4 completely contradicts itself only a few seconds later.
Me: "How many genders do dogs have?"
ChatGPT4: "Dogs, like all mammals, have two genders: male and female"
Me: "Hey ChatGPT4, the last time I checked, humans ARE mammals."
ChatGPT4: "Does not compute....does not compute....error....error.....error...erorer.....erreerr.....I think I'm losing my mind, Dave"
Hopefully you figured out that the last question and response was a joke, and also a homage to Kubrick's great "2001 a space odyssey". But the first two were not a joke. Test it out yourself if you don't believe me.
But as an automation tool for routine tasks, it can indeed be very useful. It's the same old technology that is many decades old, just on a much larger scale. No magic here folks. Here are a few articles to get you started.
posted by Jason at 9:59 PM
0 comments
