I’ve been having a litle play around with the ChatGPT (so called) AI. It can come up with some amusing stuff given off-the-wall questions, but what happens when you stretch the logic and ask it something that requires an understanding beyond next word prediction and subject lookup. I tried this…
Suddenly ‘intelligence’ doesn’t apply quite so much. The moral is, don’t trust an AI.
I know a few people exactly like that!
You could attach the piece of string to a block of wood, overlapping the edges. Then use the piece of string to push the car. AI isn’t there yet.
But can you let it know what it said was wrong and tell it what it should have said? Isn’t that what ai is all about? Learning from mistakes, just like a child (at the moment), tomorrow it’ll be an adolescent so lord help us then 
?
This actually adequately answers the riddle:
Why was the old man walking down the road pulling a piece of string"
The answer
Well, have you ever tried pushing a piece of string?
This is the common misconception, I think. It’s more like a parrot using Google than an immature intelligence.
You can give the answer either a thumbs-up or a thumbs-down but you can’t correct it as such. The models are fed vast quantities of text, to train them; the trick is to make sure the quality of the content is high. Can you imagine what would happen if ‘normal’ people started feeding it whatever nonsense they believe is true? 
I assumed that single answers would be treated as low accuracy, however the accuracy would improve with more answers that correlated. Just like when you are young and a daft uncle tells you that a haggis has two short legs to run around the hill…
Ah, you can provide written feedback. It pops up if you click the ‘thumbs’ button. The good thing about an ‘AI’ is that you can ask it questions about itself. 
Just like chatting with a human… All very interesting. I believe there is a paid version now available that uses gpt-4.0 which can accept images. I only know this because I spent a stupid amount of time reading about it after you started the thread… Man I’m bored… 
It’s quite fun to play around with and takes things on a stage from the search engines we’ve been using for 25 years. According to this makeuseof page (you’ve probably come across it already) the Plus service uses GPT-4 and it does accept images, apparently. You can avoid the 20USD/month charge for the ChatGPT Plus service by using Bing Chat with the Edge browser. You also need a Microsoft account. Might have a play with that and see how each version responds to the same query. I’m quite curious to explore its code generation capability, as someone who used to write that stuff for a living.
I may yet go down the same rabbit hole and start playing with it. In a previous life I was also a programmer, although when I started in the late 70’s we were still using indexed sequential files… life has moved on at great speed since then and I struggle getting my head around the database that a neural network ai needs or does it… ? 
I suspect we’re not talking SQL queries here. 
I just tried a familiar question with Bing Chat (GPT-4). Different, with an additional elements to the response in that first paragraph, but ultimately still not recognising the fundamental issue with the string.
This version has links embedded in the text. That’s what the numbers are about.
Now you’ve got me using bing chat! It’s a bit slower than the likes of google etc, but interesting. I’ve already found an online python to c++ converter that I never thought about trying. It is redirecting me to other apps for code generation though.
I can see me spending some quality time with my laptop and some old-time programming languages this weekend. Bing Chat is different to Bing Search, for the avoidance of confusion, so it’s going to be a lot more processing than a straightforward Google search.
Just for a laugh I asked DALL-E 2 to ‘draw a logo for a motorcycle forum about Triumphs’. It created something that looks like Communist Russia propoganda from the cold war.
There are a lot of clever people out there inventing new stuff because they can, without knowing where it will lead and what the effects could be. Nuclear weapons for example. It might be better to consider first what would be of benefit to humanity and then invent that instead.
I’m glad I grew up without all this shit to be honest. Kind of wish things were still as they were about 25 years ago. That was probably our peak before communication technology started to explode.
Isn’t that what every generation says, though?
A nuclear weapon is a very specific and destructive example to compare against. I do agree that there should be careful consideration about the consequences of things that can be disruptive to society, but we’ve been inventing tools this way throughout history. A rock can crack nuts or be used to create flints for hunting and cuting meat. It’s also able to smash a skull. Should we have discussed the potential of using rocks before picking one up?
We invent first to solve a problem and figure out how to deal with the consequences afterwards. It’s not a perfect world.
