ChatGPT: The Future of AI is Here! (1 Viewer)

[AccessBlaster]

Just another tool like any other

 

[Uncle Gizmo]

Chat appealing a fine ..

Motorist uses AI chatbot to challenge airport fine - BBC News

Shaun Bosley used ChatGPT to generate a human-like appeal to a £100 final notice letter from NCP.

www.bbc.com

 

[Jon]

If you want ChatGPT to explain things simply, just ask it to explain it as though you were five years old!

 

[Jon]

Or, something else I just tried...I like examples where people use analogies to explain things, just like Jesus did! So, ask it to explain using an anology.

 

[AccessBlaster]

ChatGPT is crazy, I've been playing around with Tkinter, Python, and SQLite3 designing little apps with GUI interfaces. ChatGPT has helped with the creative side of design by solving some of the coding hurdles that would take months to suss. I know the purest would say you need to learn what the code does before implementing it. But I already know the context of what I need, I just don't know the syntax of the code. Anyway, I'm having a blast with it! I just wished I had it 15 years ago I would have built better apps in a shorter amount of time.

 

[Jon]

I often paste code in. If you add nothing else, it will just explain it to you. In detail!

 

[Steve R.]

Here's an interesting take on AI. Recall the story, that all the works of Shakespeare can eventually be duplicated by an unlimited number of chimpanzees pounding randomly on typewriters. In the March/April 2023 issue of Analog there was a short fictional story where Analog was hit with a plagiarism lawsuit. Seems that a team of AIs had been assembled to write every conceivable story thereby making it "illegal" for any human writer to actually write a story since they had all been written already!

 

[Minty]

That's an interesting and clever story.

However, if it takes even the quickest computing power 300 years to crack a 12-character (upper and lower case) password, can you imagine how long it would take it to write every combination of words that make sense?
Even if you restricted it to a short story (say 2000 words) I can imagine it would take a while

 

[tltanhueco1990]

But until the time ChatGPT gets updated with a DB after 2021, it might get overthrown by Bard.

 

[Jon]

ChatGPT 4 has been announced and released to some.

 

[Isaac]

That's an interesting and clever story.

However, if it takes even the quickest computing power 300 years to crack a 12-character (upper and lower case) password, can you imagine how long it would take it to write every combination of words that make sense?
Even if you restricted it to a short story (say 2000 words) I can imagine it would take a while

true but , at least in the US, a lot of copyright/patent type of legal theories are anchored by some type of "mostly" or "substantially" type of tests, of varying definitions, wording and historical application.

you wouldn't have to make a literal maximum words combination. something less than that.

 

[conception_native_0123]

ChatGPT 4 has been announced and released to some.

I wonder how much better it is than v.1

 

[Jon]

For the American bar exam, it went from the 10th percentile (ChatGPT 3.5) to 90th percentile compared to humans. i.e. top 10%. It can also take in images and understand them.

As a side note, I played around with ChatGPT yesterday by asking it to create mnemonics to help you remember things. It is very good!

For example: How do I remember, using a mnemonic, that conception_native_0123 refers to Adam?

One possible mnemonic to remember that conception_native_0123 refers to Adam could be:

"Adam was the Conception, the Native, born in year 0123"

Each letter in the word "Conception" corresponds to the first letter of one of the components of the username:

• C for Conception
• N for Native
• 0 for the first digit in 0123, which represents the century (in this case, the 1st century)
• 1 for the second digit in 0123, which represents the decade (in this case, the 10s)
• 2 for the third digit in 0123, which represents the year within the decade (in this case, 22)
• 3 for the last digit in 0123, which represents the specific individual (in this case, Adam)

By using this mnemonic, you can associate each component of the username with a specific detail about Adam's birth and identity, making it easier to remember.

 

[AccessBlaster]

Katie McFarland was on Greg Gutfield the other night and they were discussing ChatGPT. She said this technology could replace lawyers and teachers.

The ramifications of this technology hasn't fully been explored.

 

[Jon]

Eventually AI will replace everything. Initially, they thought it would replace lower paid work. But in fact it is replacing the cognitive jobs instead.

 

[The_Doc_Man]

Even with v4, the folks who provide it issue warnings that it sometimes "imagines" facts and thus gets things wrong. It apparently doesn't have a built-in fact checker. However, as long as it cannot pass the Turing test we are OK. On the other hand, watch out for the day that an AI tries to file a lawsuit - which would of course require it to have standing to do so - which in turn would make it a legal person. If you thought the abortion furor was vicious, just wait for the AI person-hood campaign. You can BET that most members of elected governments will crack down on AI installations. And, of course, there will be the death penalty for an AI... death by off-switch. AND there will be lawsuits to prevent people from "pulling the plug." I won't live to see that - but my grandkids probably will.

 

[conception_native_0123]

I feel special that you remembered to use me Jon. thanks buddy!

 

[Jon]

I feel special that you remembered to use me Jon. thanks buddy!

You are special. UK members will know what I mean.

 

[Jon]

@The_Doc_Man, some think we are 5 years away from human level AGI. If you look at the output from ChatGPT, it already knows more than every human on the planet. Its knowledgebase is huge. As for hallucinations, the change from ChatGPT 3.5 to 4.0 has reduced this significantly. And humans hallucinate too. Just look at the disagreements between political sides for a start.

The Turing Test is kinda redundant nowadays, as I see it. Imagine you take a Turing Test, and the answers are so smart that they cannot possibly be from a human. You spot this God like intelligence a mile off. So, the AGI fails the Turing Test. Yet it has super-intelligence.
I see the next 5 years as a pivotal turning point in human history, where intelligent machines take things to the next level. Huge disruption is likely to take place on all sorts of levels, ranging from jobs to medicine. Exciting times!

 

[Isaac]

To be honest with you I still kind of fail to see how ChatGPT is anything more than an automation of google searches and then quickly (admittedly, VERY quickly) combing through the results, ingesting it all, and spitting something back out. As a developer in the database world, every time I use ChatGPT, I instinctively don't get the idea that it's anything special, intelligence-wise. I get that it's very special in the speed that it can pose your question, or distinct components OF your question, to Google (or whatever you want to call the search engine capacities we already had), then read all the results very quickly, then spit out a summary of those results. YES - I agree, that's extremely impressive. But it's impressive from a data storage-and-retrieval-quickly perspective, not an Intelligence one.

Even paintings, code, music. ALL of that can be dumbed down to bits of data stored, which already WERE stored. What ChatGPT does that is new is harness extreme power and speed to retrieve all that in an instant. It doesn't seem like artificial intelligence to me, it seems like an impressive array of hardware, storage and retrieval of information we already had, without exception.

I've already seen Universities successfully detected its output in plagiarism tests, something I was a bit surprised by. But then I told my daughter, it shouldn't be surprising. If you pose me a question, then show me 3 answers to the question, I could pick out ChatGPT output most of the time, and that's not even with me having a "plan" in place - just recognizing its generic output, which is an exact cobbling together of what I would have written with a few hours of google searching and writing.

Trust me, I hesitate to post this, as I expect to be pounced on from all sides - but that is just what I think.

It just seems like they built the fastest search-engine and results-parser in the world, threw the label "AI" on it, and watched for everyone's jaws to drop. ?