ChatGPT or derivations - Pros and cons

Discuss various technical topics not related to Mozilla.
User avatar
Frank Lion
Posts: 21173
Joined: April 23rd, 2004, 6:59 pm
Location: ... The Exorcist....United Kingdom
Contact:

Re: ChatGPT or derivations - Pros and cons

Post by Frank Lion »

The above post was a something and nothing test that proves nothing either way.

So, what is this ChatGPT stuff really about? As always, if you want to know what a perp is thinking then you have to think like a perp and not like you would view the same information.

We start with a well-known human failing - https://www.youtube.com/watch?v=QereR0CViMY

i.e. much as people protest otherwise, most are not as individual as they think, but are highly predictable in their behaviour and also very susceptible to peer pressure. The important word there is 'peer' - https://dictionary.cambridge.org/dictio ... glish/peer

To the perp, every human failing is something to be exploited and something to make money from. Traditionally, this was achieved by regular adverts and for the last 10 years or so, this has been done by artificially creating targeted aspirational 'role models' or 'influencers'. The idea being to impact on certain peer groups, who in turn impact on each other and so ad infin. Trouble is, like making a hit pop record, it's all a highly imprecise business with huge outlay and no guarantee on investment.

Now, let's go back, say, 20 years and eavesdrop on a perp doing a sales pitch to a whole roomful of perps and it would go something like this -
Perp wrote:"Imagine a world where we didn't have to spend a fortune on how people think. Imagine a world where instead of influencing peer groups that we ARE the peer groups! Where every thought, review, testimonial, retail decision, political opinion and the entire vox populi is controlled by us.

Ladies and gentlemen, I give you the world of artificial intelligence!"
At the moment, it's still baby steps - https://www.youtube.com/watch?v=RbTsHEPMQoo Later on, it will not be.
"The only thing necessary for the triumph of evil, is for good men to do nothing." - Edmund Burke (attrib.)
.
User avatar
Grumpus
Posts: 13239
Joined: October 19th, 2007, 4:23 am
Location: ... Da' Swamp

Re: ChatGPT or derivations - Pros and cons

Post by Grumpus »

@Kevin McFarlane - The damage caused by Windows 95, aside from the security issues (one being all the passwords in a local and remote accessible text file, was the freedom of using Dos for other programs without going through the 95 shell. There was also the mild extortion for using a software under 95. The list could go on ad nauseam.
300 Million jobs is not a number easily replaced by normal attrition or new markets if the new markets have no basis in food, clothing or shelter and relocate wealth to a small group.
The problem: You have a ditch to dig. Does the cost of developing a mechanical ditch digger and the time required balance against the hiring and paying of several thousand individuals. There lives if out of work are, even minimally improved to deal with real issues.
If the ditch digger is developed and manufactured it may create more jobs but the here and now will always take precedence over the maybe.
We have no real world particulars on ChatGPT and how the application balances against humanity and whether it is actually going to help or or cause more problems than it is worth.
This is not to say it should be made a pariah before it's time but would a few moments more of introspect hurt.
For the man who took his life there is no excuse, either for him or the chat system.

The accuracy, in some cases less than a problem, is only a small part of the AI interface.
If someone can think of it there's the very real foresight of it being applied improperly, accidentally or deliberately.
Doesn't matter what you say, it's wrong for a toaster to walk around the house and talk to you
Kevin McFarlane
Posts: 597
Joined: November 10th, 2009, 3:47 am

Re: ChatGPT or derivations - Pros and cons

Post by Kevin McFarlane »

Frank Lion wrote:
Who is Frank Lion from mozillazine?
2 seconds later....
https://chat.openai.com/chat replies
Frank Lion is a long-time contributor and moderator on the Mozillazine forum, which is a community-driven support forum for Mozilla software, including the Firefox web browser and Thunderbird email client. Frank has been a member of the forum since 2005 and has contributed to numerous discussions on a wide range of topics related to Mozilla software, including troubleshooting, tips and tricks, and feature requests. He is known for his expertise and helpfulness in assisting other users with their Mozilla-related issues.
Fast, but also not totally accurate.
They have acknowledged that ChatGPT will not necessarily get things right (it also only has data up until the end of 2021). It is possible to ask it to correct itself though. That can work, but not always.

Also the ChatGPT UI could do with updating to include references like Bing Chat and Perplexity have. Those are not foolproof, but you can at least see where it's getting its information from.

Interestingly, a week or so ago I queried Perplexity. It said essentially that the claim was fake news and that the source did not say such and such. But Perplexity also supplied references. I followed through to the original source and contra Perplexity it was NOT fake news. So Perplexity was wrong. Unfortunately, I did not realise at the time that I could have sent feedback to Perplexity notifying them of their error (which was clearly caused by political bias in this case - you could tell from the context).

I've also experienced a couple of cases in the past few days where 1 all of the various chats were rubbish but a regular search was correct (it was a bit tricky though and did need some cross-checking) and 2 Regular search was rubbish (not wrong, but couldn't offer the information in a digestible form) and chat - in this case Bing Chat - was excellent.

But I would also add that standard searches with Google and Bing are only as good as the accuracy of the pages they return - which can contain false information. There is no panacea. Users have to use their brains and crosscheck, the more so, the more important the information you are looking for.
User avatar
Grumpus
Posts: 13239
Joined: October 19th, 2007, 4:23 am
Location: ... Da' Swamp

Re: ChatGPT or derivations - Pros and cons

Post by Grumpus »

Watching the news, as objective reporting as can be found it appears like all software it's a two edge sword. On one hand an application using ChatGPT is successful while on another it's a failure.
Where the issue lies is in application, how it is being applied aside from any benign applications.
Concerns for privacy and information theft and misinformation, with so many disinformation purveyors makes it require more scrutiny and management.
Example: Leaks
This is only one of a number of issues which have popped up in the last couple of days.
There's also some reports of people with villainy on their minds and having no programming experience being able to use it to write noxious apps and programs with the help of auto writers. Some are used to save time in a positive manner.
It's like using the same knife to cut your meat and disembowel a friend. :-k
Doesn't matter what you say, it's wrong for a toaster to walk around the house and talk to you
User avatar
Grumpus
Posts: 13239
Joined: October 19th, 2007, 4:23 am
Location: ... Da' Swamp

Re: ChatGPT or derivations - Pros and cons

Post by Grumpus »

Chat and Sims article

IMO= Perceptions based on a database of past actions to determine present and future response, is at best assumptive based on past data perception.
A gold mine application for swindlers, politicians and those with less than desirable motivations. The fact this stuff goes south after a while, changing direction and adding new topics or answers based on any past inaccuracy of data acquired is most certainly a dangerous technology to govern peoples lives and conditions.

The simplest problem is overcoming double meanings, something the computer world has had a general disregard for over the years.
In most cases the determination of meaning has to be cognitive and can't rely on data without extensive and very likely prohibitive amounts of information which may still be affected by double meanings during the determination. :-"
Doesn't matter what you say, it's wrong for a toaster to walk around the house and talk to you
User avatar
Grumpus
Posts: 13239
Joined: October 19th, 2007, 4:23 am
Location: ... Da' Swamp

Re: ChatGPT or derivations - Pros and cons

Post by Grumpus »

. . . and in the movie Fantasia the Sorcerer's Apprentice has a fit of a time with the bucket brigade of brooms, just like ChatGPT.
Where the hell is the Sorcerer to straighten out the mess and the worse mess coming.
Doesn't matter what you say, it's wrong for a toaster to walk around the house and talk to you
User avatar
Grumpus
Posts: 13239
Joined: October 19th, 2007, 4:23 am
Location: ... Da' Swamp

Re: ChatGPT or derivations - Pros and cons

Post by Grumpus »

OK so lets say they deploy ChatGPT globally, OK that's been done, is there a way to decline services and networks which are using it?
Could it be mandatory to reside in a Pollyanna state of mind?
On a similar note, to the folks making art with AI: It is certainly a medium but if the machine does all the conceptualization what does it need you for other than pushing a button?
If you are using a database to support the AI, are you infringing on copyright by using the lighting from a Turner or colors of Maxfield Parrish?
Sure there are artists which do this, gather in little groups and become the Secession of this or that but it's still suspect and aimed at income and not the true spirit of creation.
Since becoming a certified antique this year (or soon) 75, remembering having to do the sketch and finish art work, color separations by hand, learning how to use lettering quills and other brushes, sculpting in clay, wood, plaster, wax and metal
It seems to me it's as business oriented and soulless to the art world as clip art. P.S. You can cram the defense of clip art.
If an artist is not needed to duplicate the very characteristics, lighting, color, brush stroke or other bits and pieces which define that particular artist's style,
doesn't that remove the human element, the instinct or oblique cognitive thought except what's recorded in a gathered database. What value does it have if it diminishes the human characteristic, the unique one time only display which provides the worthiness of the art.
Do people really need a copy of the Mona Lisa done by AI or is this solely for pocket lining? :-k
Doesn't matter what you say, it's wrong for a toaster to walk around the house and talk to you
User avatar
Grumpus
Posts: 13239
Joined: October 19th, 2007, 4:23 am
Location: ... Da' Swamp

Re: ChatGPT or derivations - Pros and cons

Post by Grumpus »

Remember the contests to catch the greased piggy?
Doesn't matter what you say, it's wrong for a toaster to walk around the house and talk to you
User avatar
Grumpus
Posts: 13239
Joined: October 19th, 2007, 4:23 am
Location: ... Da' Swamp

Re: ChatGPT or derivations - Pros and cons

Post by Grumpus »

There's going to be no end to concerns over the application of AI for all sorts of reasons but . . .
. . . unless connected to the real world, not just a database, AI is always going to be like any other programming, junk in junk out.
Concern should be over predictor like applications, medical, social direction, etc.
Not so much the foibles of the programming but more so the decisions made after application by insurance companies, educators, law enforcement and other business'
Recent and various reports of predicting someone will develop this cancer or that or possibly have a heart attack may go beyond reality and cause inaccurate decision making.
Loss of coverage, increased costs to the insured or even loss of jobs, all due to some inaccurate, virtual based, decision.
Years ago when designing spread sheet template, before you had to buy them, it was always fun to add a column for the difficult clients. Percentage increase of fee (AH Tax)
Insurance companies do what they can to reduce their risks but this may provide an excuse to be even more selective.
Doesn't matter what you say, it's wrong for a toaster to walk around the house and talk to you
Kevin McFarlane
Posts: 597
Joined: November 10th, 2009, 3:47 am

Re: ChatGPT or derivations - Pros and cons

Post by Kevin McFarlane »

Grumpus wrote:There's going to be no end to concerns over the application of AI for all sorts of reasons but . . .
. . . unless connected to the real world, not just a database, AI is always going to be like any other programming, junk in junk out.

Concern should be over predictor like applications, medical, social direction, etc.
Many of the Chats are also connected to the real world, i.e., the web, and give links (ChatGPT still doesn't though and is restricted to 2021).

AI for medical diagnosis is clearly one of the more useful applications for the tech. But presumably the doctors don't just take it at its word. E.g., the AI spots something and then the doctor looks deeper for other confirmation. But the AI just speeds up the initial search and saves time. The Chat AIs, when they are accurate, are similar in the time-saving aspects - or even when they are not accurate. For example, the other day I saw that Guido van Rossum, the inventor of Python, uses AI code generation because it saves him time. The fact that it doesn't necessarily get it right doesn't matter so much as he has the knowledge to fix up the errors. But it's still quicker than working from scratch. I'm a programmer and I can vouch for the same on a lesser scale when I'm coding in Visual Basic .NET, find an example in C#, and then run it trough an online C# -> VB translator. It gets bits wrong. But it saves time and I know enough to fix it up or I can work through the compiler errors if it's obscure.

Not so much the foibles of the programming but more so the decisions made after application by insurance companies, educators, law enforcement and other business'
Recent and various reports of predicting someone will develop this cancer or that or possibly have a heart attack may go beyond reality and cause inaccurate decision making.
Loss of coverage, increased costs to the insured or even loss of jobs, all due to some inaccurate, virtual based, decision.
Big Tech uses AI in its censorship decisions. In theory, there's supposed to be manual review but their procedures are opaque and/or inconsistent, so who can tell. But I'd say AI making physical decisions, such as taking down a YouTube channel and/or demonetising it, is far more consequential than a Chat AI returning dodgy search results. But it depends on the significance of what you're searching for. The more important it is to your life the more important it is for you to cross-check. But that applies to non-AI searching as well. Thee is no "safe" magic bullet to any of this.

Btw, since I was last here I've played around with 10 Chat AIs (or really 8, because the 10th one is kind of broken and the 9th is just a ChatGPT clone.

What I've found is

1. The Chat AIs are both more and less impressive than you expect. (By less impressive I mean they get some seemingly simple questions totally wrong.)
2. No one Chat is better than all others for all queries/prompts.
3. For some queries a standard search is better than all or most of the Chats.
4. For those Chats which give web references they do not always clearly draw information from the particular link cited).
5. They can provide a link where the Chat tells you something that actually contradicts what is in the link. So the Chat kindly provided the user with information to refute the Chat (if they were diligent enough to follow up). In this case it was obvious that the error was due to the Chat's anti-right wing bias. You could tell from the way it was worded.

Anyway, I don't know whether that makes you feel more or less reassured! :lol:
User avatar
Grumpus
Posts: 13239
Joined: October 19th, 2007, 4:23 am
Location: ... Da' Swamp

Re: ChatGPT or derivations - Pros and cons

Post by Grumpus »

Reassurance isn't an issue, if AI doesn't make errors or bad decisions causing harm of any kind, it exists a non-issue.
Unfortunately there's that group of callous souls who decide damage on percentages, leaving the 10 or 20 per cent on their own and out of consideration.
Problem is and will always be the level of accuracy in the database, links accessed and the decisions or directions implemented by the AI program.
There's no doubt automated process' help across the board. I've been replaced several times over in my almost 75 years and managed to go another way or accept the change. In some cases the loss, while improving the physical condition of the end result or process loses the grace, respect and history. Add also the loss of supporting skills and whatever grace, respect and goodwill accompanied those.
The issue whether changes with unlimited or minimally restricted conditions are a benefit to man in general, a humanist concept, exists.
The idea of AI programs having autonomy which can possibly go off the rails or having no rails take a malicious adventure of it's own design.
Of course there's always going to be the secular jackass who will declare the AI a god, start a worship cult , develop another superstition or institute secular pretensions.
At my age it's more Laconic 8-[
Doesn't matter what you say, it's wrong for a toaster to walk around the house and talk to you
Kevin McFarlane
Posts: 597
Joined: November 10th, 2009, 3:47 am

Re: ChatGPT or derivations - Pros and cons

Post by Kevin McFarlane »

Grumpus wrote:Reassurance isn't an issue, if AI doesn't make errors or bad decisions causing harm of any kind, it exists a non-issue.
And humans don't also make mistakes and bad decisions causing harm? Seems you're asking for the impossible.
User avatar
Grumpus
Posts: 13239
Joined: October 19th, 2007, 4:23 am
Location: ... Da' Swamp

Re: ChatGPT or derivations - Pros and cons

Post by Grumpus »

Asking for the impossible is what logic does, by-passing emotion, compassion and philosophy (in theory).
Otherwise what is the purpose of an AI system except to remove the chance of error, again the search for perfection.
It should be noted your comment of seeing the errors in the code because you are experienced is fine but AI globally deployed, folding my finest tinfoil chapeau, leaves a small number of technically adept in control.
A very much smaller number than what exists today in various contentious and self righteous groups.
The question should be, are these folks humanitarian guided or irreparable Sheldon's wanting to globally engineer society, a goal a number of totally wacky wealthy seem to feel they have a right to do.
Mind you, I like a dollar as much as the next person but have no desire to control anything except my lawn mower, car and computer system.
:wink:
Doesn't matter what you say, it's wrong for a toaster to walk around the house and talk to you
Kevin McFarlane
Posts: 597
Joined: November 10th, 2009, 3:47 am

Re: ChatGPT or derivations - Pros and cons

Post by Kevin McFarlane »

So bad stuff might happen. What is your solution? It seems you don't want the technically adept or "wacky wealthy" to be in control. You fear the global engineering of society. On the other hand, you seem not to want the "plebs" to have access to this tech as they might do bad stuff as well!
User avatar
Grumpus
Posts: 13239
Joined: October 19th, 2007, 4:23 am
Location: ... Da' Swamp

Re: ChatGPT or derivations - Pros and cons

Post by Grumpus »

It's not a matter of access as it's already in the wild, to late for not want, even if the code is unavailable there's always someone in the tech world who will share.
It's how I got my first copy of Kermit in 1983, left anonymously in my mailbox.
Also there are reasons for AI, you mentioned, but like social media there's the good and the bad which should be addressed before it gets out of hand.
Probably easier to control the bad on social media than something like AI in the wild though the hope for both is diminishing.
Sitting in a hospital room with a doctor who looks to be about 12 sitting in front of a computer screen and asking a question almost every 2 minutes is disconcerting at best.
A concern here is the attention to the screen more so than patient,likewise the data over experience or what a patient says.
You don't kill the dog for peeing on the rug, just train them to go in a more appropriate location.
Doesn't matter what you say, it's wrong for a toaster to walk around the house and talk to you
Post Reply