|
This.
The incessant talk about it can get annoying, but "pissed off"? No, over the decades I've learned not to let annoying things get under my skin. It's useful.
Typically, I might try to get it to give me a short sample showing how to use some obscure API. More often than not I still have to tweak the results (I'm not expecting anyone, or an AI, to do all my work for me), but asking it in plain English what you're trying to do is often far better than trying to come up with the exact right keywords to have a search engine bring back a relevant page.
|
|
|
|
|
I hadn't heard/read anything about it all day...until now. Thanks for that! 
"Go forth into the source" - Neal Morse
"Hope is contagious"
|
|
|
|
|
I'm more p---ed off about people (read: members of the media) who have no idea how generative AI works claiming it will solve the world's problems and/or bring an end to civilization as we know it. 
Transformer ANNs in particular are very useful when trained on a vetted and well bounded corpus of text (e.g. user's manuals for software, service manuals for machinery, the works of Shakespeare, the IRS tax code, etc.) Persons who rely on the publicly available version of ChatGPT to do their homework or answer questions of any significance are only fooling themselves.
/ravi
|
|
|
|
|
Quote: Persons who rely on the publicly available version of ChatGPT to do their homework or answer questions of any significance are only fooling themselves.
My latest expirience with the ' publicly available version of ChatGPT' was to solve a math problem.
It helped me a lot, it helped me to solve it in between two hours to get a solution with Octave (the matlab alternative).
I don't see the point where 'are only fooling themselves'.
I don't know matlab/octave syntax, but by the help of ChatGPT I was able to solve in short.
Without ChatGPT: I would have to first learn matlab/octave 
|
|
|
|
|
I'm glad ChatGPT was able to help you solve your issue. I've also used it with great success to quickly figure out how to do something specific in Android (with which I'm only somewhat familiar). Before ChatGPT I would have Googled the problem and probably waded through several answers at SO before eventually reading the Android docs and hoping to find a solution.
I should have been clearer. What I meant was ChatGPT works by predicting an answer (actually predicting just the next word in a sentence) to a question, based on the LLM it's trained on. It doesn't "understand" a word of what we ask it. What I meant by "questions of significance" was a question that requires analytical thought and whose answer can't be found in an existing corpus of text.
For example, asking ChatGPT to suggest a course of therapy for a specific psychological patient won't get us anywhere, since ChatGPT doesn't understand psychiatry or how to treat patients suffering from psychosis. It simply searches for the next likely word (and consequently a phrase and sentence) based on our input. This is why it excels at tasks like summarization and perusing content in well defined and well bounded bodies of text (e.g. user guides, collections of legal and tax rules and in our case Octave and Android how-to articles).
/ravi
|
|
|
|
|
As mentioned, I was on a math/programatically thing with ChatGPT.
Especally the suggestions from ChatGPT (usually an iteration when ChatGPT delivered not the expected solution *1) ) on a question leads me to the conclusion, that it is not simply an answer based on statistics what the next word will be.
*1) But yes, sometimes ChatGPT starts to repeat itself if one claims a mistake in the answer. And that, confirms then your statement, that statistics for the next word comes into acount.... 
Anyway: Still confused some times when using ChatGPT, but also surprised how good it helps for easy code fragments 
Sorry for my English. Most probably hard to understand
|
|
|
|
|
Your English is fine! 
Requests to ChatGPT include a property (oddly enough called "temperature") that controls how much it will vary successive answers. This property simply injects a degree of randomness in its output. So it's not surprising that its answers can sometimes be completely incorrect, and not just slightly wrong.
A couple of examples:
- I asked ChatGPT what would happen if I deleted a file in a folder that was synced to OneDrive when I didn't have an internet connection. It responded by saying the file would be restored to my local storage the next time I connected to OneDrive. When I asked it the same question again, it said the file would be deleted from OneDrive when a sync was established. Two completely opposite answers in quick succession!
- I asked ChatGPT to tell me a little bit about the app TakeStock (which happens to be written by me). It responded with a very accurate description, and included the fact that it was written by me. I then asked it what other apps Ravi Bhavnani had written and it responded with a list of apps that were completely unrelated to TakeStock and that I haven't written. I've written several other popular apps, but none of those appeared in the list. In fact the apps in the list weren't even remotely similar to any of my other freeware apps.
At work we use (a paid version of) ChatGPT that has been trained on our user's manuals. Our product is pretty complex (highly configurable) and its documentation set is quite large and includes several detailed troubleshooting guides. ChatGPT has been remarkably useful by allowing end users to ask questions in plain English (and soon other languages) and do away with the need to speak with a human (although that option is always available).
/ravi
|
|
|
|
|
This exactly. Every piece of software with more than one "if/else" is now an "intelligence". I've yet to see how ChatGPT is so "intelligent" that it can raise my kids.
|
|
|
|
|
Ravi Bhavnani wrote: bring an end to civilization as we know it
That's Facebook's job (and its countless spinoffs).
And based on the results so far, it's getting pretty good at it.
|
|
|
|
|
Ravi Bhavnani wrote: Persons who rely on the publicly available version of ChatGPT to do their homework or answer questions of any significance are only fooling themselves.
People who expect to get full/complete answers from an AI are only fooling themselves.
But I often find it useful to point me in the right direction (eg, putting together a small sample that shows how to correctly use some obscure API). Even if it's not 100% correct, it's more often than not sufficient to get me on the right track.
And you're right, those who expect more...well, we tend to see them post questions here in the Q&A section, and they're never happy there also with the responses they get back...
|
|
|
|
|
The Media has become a in your face we need to be everywhere
Back in the day (yes I am 77 years old) you picked up a magazine or newspaper
and would read what you wanted. You watched or listened to these medias when you wanted to
TODAY the phone wants to tell you everything when you contact a friend
and with that phone usage you can link your usage information with your a computer
WHY is this happening it is called revenue from Ads
I have a cell phone and learned the lesson when I first had a flip phone
NO ONE gets my phone number if they demand I give them the White House Number
Drivers most companies crazy when I say I Do Not Have A Phone
Safeway has $2.00 off on Ice Cream if you scan the coupon price tag
So my question is am I being discriminated against if I do not have a phone?
I used AI to solve a C# & VB.Net question at this site LINK BELOW
How do I find the 4th Tuesday of the Month ?
The answer was very useful to understand it you need a working knowledge of coding
and most important how to ask the question
Code Generator - CodePal[^]
|
|
|
|
|
As we can't see whether your hand up is or not, I'm assuming it's not.
So here's a link to register for the Microsoft Build: AI Day [^]
You can hear about AI for a full day 
// TODO: Insert something here Top ten reasons why I'm lazy
1.
|
|
|
|
|
🙋♂️🙋♂️🙋♂️🙋♂️🙋♂️🙋♂️🙋♂️🙋♂️🙋♂️🙋♂️🙋♂️🙋♂️
Specially hearing about it, when it is wrongly used.
M.D.V.
If something has a solution... Why do we have to worry about?. If it has no solution... For what reason do we have to worry about?
Help me to understand what I'm saying, and I'll explain it better to you
Rating helpful answers is nice, but saying thanks can be even nicer.
|
|
|
|
|
Nelek wrote: Specially hearing about it, when it is wrongly used.
Somebody wrong on the internet about a technical topic? Say it ain't so!
|
|
|
|
|
In the internet?
And in the radio, in the television, in the newspaper, in the news@work...
M.D.V.
If something has a solution... Why do we have to worry about?. If it has no solution... For what reason do we have to worry about?
Help me to understand what I'm saying, and I'll explain it better to you
Rating helpful answers is nice, but saying thanks can be even nicer.
|
|
|
|
|
I'm more annoyed by the way the media have fallen upon it as something to babble endlessly and inanely about. It seems that the less they understand something, the more noise they make.
I'm reminded of the Biblical warning - "bow down not before craven idols". (I'm not religious, by the way, but it nevertheless strikes me as good advice.)
|
|
|
|
|
I used to be annoyed back when it was just used to cheat investors out of their money.
But now? Now it's the real deal, my friend.
|
|
|
|
|
|
As a developer for 44 years, I hate AI. Todays development systems have so much of that garbage in them. I hate it when it keeps changing my code while I'm trying to type something. I know what I want so please stop messing with me. It really lowers my productivity.
|
|
|
|
|
So I am not the only one fed up with all the automatic code insertions...
I was working on a C# project in VS 2022 and got so annoyed with this idiocy that I went back to VS 2019, where I am staying for the time being...
In VS 2022 I tried setting ll of the options off that may have been supporting all the code insertions but nothing worked...
Steve Naidamast
Sr. Software Engineer
Black Falcon Software, Inc.
blackfalconsoftware@outlook.com
|
|
|
|
|
I use Bing Chat GPT for help, but yeah, I think I would be better off if we could go back to the 80's man. One of my favorite games as a child was Neuromancer. Wish they would bring the original version back so I wouldn't have to do the "dark room" emulator gig.
|
|
|
|
|
I just see AI as another way to make people dumber than they already are...
I don't see the need for it in my development projects...
Steve Naidamast
Sr. Software Engineer
Black Falcon Software, Inc.
blackfalconsoftware@outlook.com
|
|
|
|
|
I see current "AI" as a tool that can be helpful if the user is aware of its limitations. But it definitely doesn't meet my definition of Artificial Intelligence.
|
|
|
|
|
yeah, not going to put it up yet again as I've been shouting and waving to get of that wagon for like... years???
|
|
|
|
|
I'm tired of hearing about it on all the tech sites I visit.
Since I retired a few years ago I haven't come across a need
to even try any of the AIs.
|
|
|
|
General 
News 
Suggestion 
Question 
Bug 
Answer 
Joke 
Praise 
Rant 
Admin 
Submit an article or tip
