ChatGPT or derivations - Pros and cons

Discuss various technical topics not related to Mozilla.
User avatar
Grumpus
Posts: 13268
Joined: October 19th, 2007, 4:23 am
Location: ... Da' Swamp

Re: ChatGPT or derivations - Pros and cons

Post by Grumpus »

As if artists, designers, musicians, engineers and actors don't have enough on their plate . . .
. . . this article shows a future of who dunnit, who diddit and WTF when it comes to things like the actual stability and extent of protection under copyright and patent law regarding protection enforcement. It seems to me AI has no real right to exist if doppelganger creations are only what's applied using it or what it becomes based on copied works of others.
Computer hardware and software power can probably perform calculation and diversionary trails for expanded knowledge (computers thinking outside of the box) but the use of creations without compensation, credit or minimal acknowledgement is intellectual property theft. In essence counterfeit production.
Imagine 3-D printers hooked to AI using information from a safety device and the AI program disregards material structure and the airplane crashes, boat sinks or car door won't open. This is not chicken little mentality as no one appears to be regarding the expanded ramifications.
The computer industry has never been a real think a head group when it comes to having discovered a new techno-trick. It just bulldozes ahead with the application without any real consideration for the damages to the whole.
At 75 years and having applied computer skill and hardware at various levels since 1968 it seems to me the lawyers need to stop looking at this with subjective eyes which are very likely looking at legal fees instead of the flat facts. There's no reason for discussion of whether it's legal, it isn't in any sense of the word.
Even works no longer covered by copyright can not be segregated from the abuse and as the article states sub-standard data would lead to sub-standard output. Funny, sh*t in sh*t out has always been the standard so what's the reason for the confusions?
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: 13268
Joined: October 19th, 2007, 4:23 am
Location: ... Da' Swamp

Re: ChatGPT or derivations - Pros and cons

Post by Grumpus »

Hypothetical Question: If a creation by AI is determined to have been created with permission granted data.
What prevents the data collected from reuse by other systems/individuals and how many levels of copy before it winds up a confusion of permissions creating the need for new permissions.
Seems exponential. :-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: 13268
Joined: October 19th, 2007, 4:23 am
Location: ... Da' Swamp

Re: ChatGPT or derivations - Pros and cons

Post by Grumpus »

Brush up on your obscure languages
Wonder how this will work with Firefox's new translate function, if it's like Google's? :-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: 13268
Joined: October 19th, 2007, 4:23 am
Location: ... Da' Swamp

Re: ChatGPT or derivations - Pros and cons

Post by Grumpus »

This Register article shows why judges need to have a deeper understanding of the topics they review and make decisions on. Having had experience with arts and judges decisions a technical background would have helped this judge to be a little more understanding of the circumstances involving copyright and ownership.
It's always been a sore subject in the technical community as to who owns what and in many case when technology is transferred and used with a disregard for origin or ownership.
AI is a real test of property rights, intellectual property boundaries and having judges intelligent enough to see the variations which cases are based on. There will always be those who feel nothing is out of bounds but in order for patent and copyright laws to be enforced you need judges who have a working knowledge of the subject to limit those abusers.
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: 13268
Joined: October 19th, 2007, 4:23 am
Location: ... Da' Swamp

Re: ChatGPT or derivations - Pros and cons

Post by Grumpus »

This has to go in the too funny column. Politically correct adnauseum
It seems the simplest answer is the AI is picking up on the timidity of it's Internet sources. While it may be somewhat off target it shows a general lack of freedom required to properly work, particularly when it comes to details affected by random points.
History has enough nonsensical changes due to whatever group of overly sensitive or politically correct whiners but wait, AI might help increase the Whine factor exponentially.
OK wait for it, someone will take offense to the use of Whine :roll:
Doesn't matter what you say, it's wrong for a toaster to walk around the house and talk to you
Kevin McFarlane
Posts: 599
Joined: November 10th, 2009, 3:47 am

Re: ChatGPT or derivations - Pros and cons

Post by Kevin McFarlane »

This is Machine Learning Fairness – allegedly for unbiasing results that are skewed by race, sex, etc. – gone wrong.

So, say, 9 out of 10 engineers are men. User requests an engineer and Google may show you a woman instead of a man. An “unfair” algorithm would default to the stereotype – man.

I assume Google’s ML fairness shows women 50% of the time or might actually show them preferentially. Gemini is probably doing the latter, but with race, and applying it to the past!

But really “fairness” should reflect reality and the algorithm should randomly show men 90% of the time. If in the future the percentage of women increases then it should adjust to suit.
In the meantime, Google’s ML fairness is actually producing disinformation!
User avatar
Grumpus
Posts: 13268
Joined: October 19th, 2007, 4:23 am
Location: ... Da' Swamp

Re: ChatGPT or derivations - Pros and cons

Post by Grumpus »

If you take the time to consider the definition of AI it does not exist as a replication of human brain process
True AI requires perception. reaction and imagination controlled by personal conditioning, general experience and emotion.
Owing to the considerable arguments one way or another on the issues in life, accurate or truly factual results can not occur regardless of the math. The binary is limited to the binary no matter how many times you twist it is always binary.
What you wind up with are calculations based on data which can be helpful or harmful depending on the accuracy and purity of input.
There will never be a Johnny Five. :-k
Doesn't matter what you say, it's wrong for a toaster to walk around the house and talk to you
Kevin McFarlane
Posts: 599
Joined: November 10th, 2009, 3:47 am

Re: ChatGPT or derivations - Pros and cons

Post by Kevin McFarlane »

AI to create a half billion new jobs — here’s why
Global services company Ernst & Young (EY) has been deploying general AI (machine learing, etc.) since 2012. Over the past decade, it has seen major improvements in employee efficiency and accuracy.

For example, the firm implemented automated document readers that could summarize reports; it also rolled out data classification engine and prediction algorithms — tools that helped employees with everyday tasks.

Jeff Wong, Ernst & Young’s Global Chief of Innovation, said automating tasks through AI resulted in a 10-fold improvement in time to completion and five-fold improvement on accuracy of results. And instead of replacing workers, it forced the firm to hire more technologists.

In 2018, for example, the company had 2,100 technologists; today it has more than 75,000. A lot of those workers were added to address new capabilities created by AI.

“Moving task time to thinking time was incredibly valuable for us,” Wong said. “I’m part of the crowd who believes over the 15- to 20-year time frame, AI will be more expansive to jobs and job creation. We will go through a shift in changing period between now and then.”

With the adoption of AI, Ernst & Young faced a problem common to all organizations seeking to roll out the technology: an ill-prepared workforce. So, Wong led an effort to upskill employees and is now rolling out AI training to all 400,000 workers. So far, 84,425 have completed the training.

Additionally, 15,000 EY employees have completed specialist "AI badges" and 22,000 more are in the process of getting them.
User avatar
Grumpus
Posts: 13268
Joined: October 19th, 2007, 4:23 am
Location: ... Da' Swamp

Re: ChatGPT or derivations - Pros and cons

Post by Grumpus »

There's no doubt the benefit well defined software can have in science and business but logic dictates an end of usefulness of those newly hired and will very likely require some re-inventing for themselves and likely without company support.. Teaching employees the ropes has never been a bad idea, many companies with a little foresight prepare so every employee can do the other employee's job in a pinch. This is long standing in military services. The speed of application which appears to be increasing exponentially in the AI world can only do one thing, eliminate redundancy in the solutions an interpretations the AI is applied to and the elimination of those gaining the goals set, once finalized.

There was a while ago some discussion of whether AI should be allowed in finance at all, the push by politicians must have been pretty hard and unnoticed for the most part as there does not appear to have been any publication of yes or no, this may be the only evidence unless unless Ernst & Young are using the old tome better to get yelled at for doing something wrong than not to be told no before you try. Might be curious to see how the SEC feels about this. Might be like stock car racing, if there's no written rule do it.
Hollow rivets and nitrous, what a gas. :-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: 13268
Joined: October 19th, 2007, 4:23 am
Location: ... Da' Swamp

Re: ChatGPT or derivations - Pros and cons

Post by Grumpus »

Early on advancing computer technology depended on the cooperation and borrowing of bits and pieces of software, feeding on itself and others.
Over the years various large technologically motivated concerns have bought, intimidated, imitated or stolen what was needed to try and corner markets.
What has resulted is a mentality that everything is for use, regardless of harm, without compensation or any form of true humanity if it gets in the way of the buck or the PR Turds need to boost image. The fact people speak out against unauthorized use shows this failure of AI to really come up with something new and inventive without copying or stealing anything needed to promote or develope the technology. Just because the trains run on time doesn't mean the rest of the corruption is acceptable or needed for the health and welfare of the whole.

Musicians open letter :2gunfire:
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: 13268
Joined: October 19th, 2007, 4:23 am
Location: ... Da' Swamp

Re: ChatGPT or derivations - Pros and cons

Post by Grumpus »

LLM Kryptonite
. . . or how technological arrogance breeds ignorance in upper echelons of the tech world. ](*,)
Doesn't matter what you say, it's wrong for a toaster to walk around the house and talk to you
Kevin McFarlane
Posts: 599
Joined: November 10th, 2009, 3:47 am

Re: ChatGPT or derivations - Pros and cons

Post by Kevin McFarlane »

My take on Generative AI after a year or so of regular use…

In general, I’ve found the genAIs (Chats) to be both very impressive and surprisingly dumb. By the latter I mean there are occasionally some very simple prompts that they either all fail, or do badly, on.

No one Chat tool is better than all others for all prompts.

Overall, for technical and software development prompts, Phind has been the best.

Most of them provide web sources with their answers. However,

1 It’s not always obvious how an answer has been constructed from a specific source(s).
2 The answers can still be inaccurate, even when the source(s) is not.
3 Sometimes the Chat does not give you an answer, or says it can’t, when the answer is in fact in the source(s) that it links to!

I’m currently using Brave Search as default desktop search. This has a Summary AI and a Code AI that each pop up contextually (i.e., not always) when you enter a search. They can also be separately turned off in settings but are also available on demand. I’ve generally found, especially the summary, to be more useful than not. It saves time, such that I often don’t need to venture beyond it. I recently completed an IT certification and was regularly checking or confirming question answers in Brave and it was pretty much accurate. Where it wasn’t, it was incomplete rather than wrong.

There have been cases where I’ve been unable to get an answer to what I want from search – sometimes using several search engines - but I’ve been able to using a Chat.

But sometimes it works in reverse – the Chats fail but a regular search doesn’t.

There is, however, an art to prompting, so it’s possible that I would have succeeded with more carefully crafted prompting. Even so, it remains true that for casual wording the searches can succeed when the Chats don’t.

I’ve been using an IDE coding assistant for a few months and it has been very impressive in this context, even just as code-completion on steroids, and it can be very spooky when it generates almost exactly what I was intending to type next.
User avatar
Grumpus
Posts: 13268
Joined: October 19th, 2007, 4:23 am
Location: ... Da' Swamp

Re: ChatGPT or derivations - Pros and cons

Post by Grumpus »

@Kevin McFarlane - Your post indicates ,which to me, is the biggest problem with any form of programming and that being trust.
Where calculations excel is in the math and what appears is logic follows logic but with AI perfect logic conflicts with the attempt to create sentient.
Secondary and tertiary checking cannot replace the experienced view. The influences surrounding and the slightest difference in the data is what creates the differing answers. It's taken human kind millions of years to arrive here now and the brain is still susceptible to screwy data/influences which cause us to come up with inaccurate and absurd conclusions. It doesn't mean research should stop but practical application is going to have it's limitations in providing fact.
The most flagrant example is history which shows the imperfection of the human brain which fails to comprehend the actual and accept the misguided, irrational, illogical and absurd, usually by the influence of either the ignorant or the wicked.
If we ain't perfect how can we expect to create perfect, by accident? ;)
Doesn't matter what you say, it's wrong for a toaster to walk around the house and talk to you
Post Reply