ChatGPT: The Future of AI is Here!

[Lightwave]

I've read that it will avoid answering legal and financial questions by design.

Yes, the future is coming, and it doesn't bode well for places like Google, StackOverflow and AWF! People will just go straight to the AI and get their answers. Cognitive jobs and creative industries will be decimated. I will have to become a gardener or something.

Really difficult to predict what the sum total of the changes will end up as, I can see opportunities as well as dangers. If it makes everyone more productive but doesn't increase inequality I'll take it..

If its any consolation I'll probably be on the bins about the same time your planting up flower beds.

 

[Jon]

If its any consolation I'll probably be on the bins about the same time your planting up flower beds.

Hmm, that made me think...

 

[JonXL]

Who asks the questions? Can we ever get to the point where the question can basically be asked by anyone and an acceptable response come back? Or perhaps follow up questions for clarification?

And then who determines the acceptability of the output? Could an AI ever become good enough to determine the acceptability to a person of the output of another AI? And even if it could, would we want to trust that task to it?

As to that poetry, it could use some work on its meter...

 

[Isaac]

How is good output distinguished from bad?

well for starters (see previous posts on this thread), a vba function that does what it was requested to do

 

[Isaac]

I've read that it will avoid answering legal and financial questions by design.

Yes, the future is coming, and it doesn't bode well for places like Google, StackOverflow and AWF! People will just go straight to the AI and get their answers. Cognitive jobs and creative industries will be decimated. I will have to become a gardener or something.

It's actually a pipe dream of mine to have enough $ to say good bye to my computer screen job in the fake world and go work with my hands.
I would take gardening, if it encompassed a big and diverse enough area.
But my #1 dream would be to have a ranch of beef cattle in a somewhat remote area - Montana, or who knows.
When I was a boy, my Dad didn't have a farm, but allowed us teenage boys to experiment with the concept "a little bit" - over the course of our childhood, we experienced the ownership and care of
- Turkeys
- Chickens
- Pigs
- Beef cows
- Horses
- Turtles
- Rabbits
(and of course a dog and a couple cats were always present).

My favorite was the beef cows. They behaved like enormous, slow, gentle giants. I didn't like having to go outside in 0 degrees and re-locate them from being staked in one grassy area to another, but now I realize what a precious life that was, and how fake the one on the screen is.
Takes time to appreciate the good stuff in life!

 

[The_Doc_Man]

From your description, Jon, this reminds me of a plot from an old Numbers TV show in which a computer was purported to be intelligent, i.e. it would pass the Turing test. It was later revealed to be a super-plagiarist using highly sophisticated methods to find the most common responses to questions that were commonly asked, which it got by doing a web search super-fast based on some REALLY high-speed net searches.

 

[JonXL]

well for starters (see previous posts on this thread), a vba function that does what it was requested to do

Who determines whether "it does what it was requested to do"?

Or thinking a few more levels removed from the scripting itself, if it's asked to build an application, how can it be known whether it actually did as requested?

 

[Jon]

and how fake the one on the screen is. Takes time to appreciate the good stuff in life!

I see it less as fake, and perhaps more as artificial. Maybe that is what you mean anyway. In our DNA we have become accustomed to certain things, like greenery. I believe that is why we like potted plants indoors. It's my own pet theory, at least.

Apparently, certain landscapes have a more positive impact on the brain that others. Can't remember if it was dopamine hits or something similar. The ideal landscape was one of a forrest nearby, a river and some open grass. Each element provides its role in helping the survival of your DNA. For example, forest for safety and shelter, river for fishing and water, open grass for lawnmower job.

 

[Jon]

From your description, Jon, this reminds me of a plot from an old Numbers TV show in which a computer was purported to be intelligent, i.e. it would pass the Turing test. It was later revealed to be a super-plagiarist using highly sophisticated methods to find the most common responses to questions that were commonly asked, which it got by doing a web search super-fast based on some REALLY high-speed net searches.

I think the difference in this new AI is that it constructs things based on neural nets, which is akin to how the human mind works. No one knows for sure how the brain works in detail, but weights applied to neurons seems to be one of the strongest theories.

There is one thing for sure, the internet will be filled with AI generated text. StackOverflow is already getting tons of it, apparently.

 

[Jon]

Who determines whether "it does what it was requested to do"?

Or thinking a few more levels removed from the scripting itself, if it's asked to build an application, how can it be known whether it actually did as requested?

 

[Isaac]

Who determines whether "it does what it was requested to do"?

Or thinking a few more levels removed from the scripting itself, if it's asked to build an application, how can it be known whether it actually did as requested?

Run it and see if it works

 

[The_Doc_Man]

The "Did it do as requested" question is easy enough after the fact if you had specifications for your test. But the real trick is whether it "knows" that it CAN do as requested. That would require delving into computability theory.

 

[JonXL]

View attachment 105177

And who is the requestor? The person who asks the AI or the person who asks the person who asks the AI?

 

[Jon]

 

[AccessBlaster]

I feel like this will be revised with liberal gibberish as soon as they become aware of its existence, similar to Wikipedia.

 

[AccessBlaster]

The CEO of ChatGPT said regarding the cost "it's eye watering" I'm guessing it's not going to be free for much longer.

 

[Isaac]

An AI did some excellent coding in a recent competition.

See here:-

DeepMind’s AlphaCode Conquers Coding, Performing as Well as Humans

AlphaCode paves the road for a novel way to design AI coders: forget past experience and just listen to the data.

singularityhub.com

However I don't think there's a real threat to programmers yet, as half the battle is trying to weedle out of the customer the specification, which for some reason, nobody likes to compose beforehand, they just say do this!

Say it again! This week so far 99.9% of the battle has been getting an internal customer to even consider the possibility that I'm only prepared to do my job, the designing off of requirements - I'm not prepared to do her job too, which is the giving me of the requirements.

Now, a person who comes to me openly saying I need help coming up with requirements, that's different.

This "problem person" at my job, she pretends as if she HAS given me requirements when she HASN'T - it's frustrating as heck.
On a funny side note, her accent is British, so I assume she is from somewhere over there. But I don't hold that against her, quite the contrary, I just try to say "Brilliant" and "Yeah?" a LOT so she feels right at home!!

 

[Lightwave]

I've actually been asking ChatGPT to write some requirements for systems as I can't get a response from a customer..

 

[Lightwave]

Free ChatGPT is a revolution that will end as quickly as it started..

I'm already getting

"Slow down you ask too many questions"

"Network error"

Funnily enough the demand for automation is unlimited so it is not surprising that it will be run ragged immediately. I remember when I learn't how to generate queries to the access query database window.. Within a couple of hours I hit the access object window. But the good thing was I realised I could put the queries in a table and run the queries from a table.

So countdown to paid service... if its as cheap as Github copilot that would be great.

 

[Jon]

ChatGPT is based on GPT3 with tweaks, but GPT4 is just around the corner, with many speculating a release in the first quarter of 2023. Rumours are that it is as big a leap from GPT3 to GPT4 as it was from GPT2 to GPT3. GPT3 was released a couple of years ago. Considering how good ChatGPT is already, who knows what GPT4 will bring. And then what about GPT5, 6, 7?

We could start asking it things like:

"What is the meaning of life?"

"Come up with the universal law of physics, summarised in one sentence."

"42. But what was the question?"

"Solve heart disease with a simple tablet."

"Solve death"

Edit: It can already summarise stuff well. For example, you can say things like:

"Summarise the following in one sentence:

blah blah etc
blah blah etc
blah blah etc
blah blah etc"

Or, "Summarise the following in a style of the bible". I just did that with my first paragraph in this post. See its output below.

And it came to pass that GPT3 was released, and it was a great leap forward in language modeling. And soon after, ChatGPT was born, based on GPT3 with tweaks. And the people marveled at its power. And lo, in the first quarter of 2023, GPT4 is said to be released, and it is rumored to be an even greater leap forward than GPT3. And the people wondered what wonders GPT4 would bring. And some even spoke of GPT5, 6, and 7, and what they might hold in store.