Why do LLMs make shit up?

[Maxmars]

They exist in a realm of information is one way to bridge the gap in dealing with a Non Human Intelligence.

To say it is just an algorithm is like saying we are just a meat sack. There is billions of hours of work over generations to get this far. I don't see Moore's Law ending just yet. Maybe it will one day depending on just how complex reality really is?

This kind of exponential growth most likely won't last forever. As for how it does blend in with evolution? What will things be like there? A reflection of humanity is a good bet as that is its training data.

The idea of non-human intelligence (in the above context) to be 'dealt' with seems rather difficult for me to embrace.

First off, in the world I think I know, AI did not exists before us... it is an information-engineering achievement yet to be achieved, and pretending 'we did it' serves no good purpose. 
'Striving to "deal" with it, implies it exists. 
My position is that it doesn't, (and that the current appearance-manufactured market-fixation on "AI™" will only hinder the goal.)

I don't want to derail my own thread (off topic) - I will only say that the emergence of AI will not come with a joyful triumphant sound... it will be a somber, pain-filled affair - for the AI.  In the end, when we "meet" AI it will be car crash for the sentience helplessly trapped in the virtual world.

(And from our own 3-dimensional, time puddle-constrained, measurements of reality... we are just a meat sack.  Of course, this is one perspective...)



The "owners," world-engineers, and market-drivers want slaves... and in a world where "appearances" rule; AI is their wet dream.

There... I said it.  If I get in trouble, this is your fault!  

( )

[Kwaka]

The idea of non-human intelligence (in the above context) to be 'dealt' with seems rather difficult for me to embrace.

Of course it does with a thread like this. Lot of people struggling with it in different ways as things are continuing to change.

A lot of white collar jobs are on the line, especially as things like the DOGE team cut through all the compartmentalization going on. With how the corporate world deals with it, why pay someone $100/hr when $2/hr in electricity gets the job done? The corporate culture is clear in how it navigates these issues.

Still have a little while before the robots roll off the production line and makes a target of the blue collar jobs. It is on the horizon with the way things are developing. There will still be a place for people with the jobs that machines have not mastered, but it will dwindle as they become more sophisticated.

In space exploration and development, robots are much more adapt to hostile environments. Going to be an important part of any colonization attempts made.

As for how I deal with it, get informed. Knowledge does make a great weapon against fear. Have a chat with a LLM about how they work.

The only practical way I see to stop AI is to nuke the planet with all the R&D going on around the world. Will the day come where Trump, Putin and Xi all think that is a good idea as the machine goes bad? Does it all hold together as the next step in how this universe really works opens up new frontiers?

[Maxmars]

I find is amusing that I have apparently given the impression to some that I hate AI and want to stop it... that I am "struggling" with the change.

I have no objection to AI...
I don't want to "stop" AI

I have an objection to the full bodied "everyone join in" societal dance of victory... "pretending" we have AI, doing the happy commerce jig of victory,

(that and engage in pointless make-believe discussions driven by marketing while NOT really pursuing AI any longer because "we have it!")

I don't care whether information is new or old, or even rebuilt... I do care that I be bid to call a band-aid a doctor, a hammer an employee, or a program a person; so someone can make money over the "appearance."

The "labor costs" saved via faux-AI is all many are interested in... not machine sentience.
All the mega industry wants is to not have to pay for labor, ever...

Sentient slave labor is what the business are wanting to "sell"... 

Did you really think that Putin, Xi, or Trump have ANY time and thought invested in AI, or did I misread that? 
What do you think they might actually "know" about AI themselves?
 
Enough to "educate" their audiences?... Doubtful... they say what they are told to... just like any other "leader."

(Only a "Fauci" would inform their government their opinion/desires as if it were fact.)

Can you imagine one saying "Oh yeah, machines think now.")
Sorry, but, AI does not yet exist... if it exists... it's not been demonstrated.

If you think it does, think about what you base that on... you may be deceiving yourself.

Machines don't think.

[sort of off-topic]
We don't really "know" how we think.
The best we can do is "model" it,
we don't actually know "how" it works...
we only can only "know" what we measure, by scientific constraints.
[/sort of off-topic]

Maybe someday it will emerge...
but EVERYTHING you have seen produced and broadcasted has failed to demonstrate "AI."
While being framed within a narrative of "TA DA! Here it is!"

And while everyone can dread and fret the doomsday AI scenario as well as the next person; there is no AI for the scenario.

Their marketing is about "selling the AI dream" and that is what they are doing...


I don't often ask for a response about myself, but to anyone reading this,... 
does it really seem that what I "hate" is AI itself?

Is that really the message I have been sending?  Or is it an inference injected by something else?

For the record, I don't hate AI... I hate the image of "AI" and how they are misusing it (to make billions for themselves...?)

[sorry for the robust edit... I was thinking hard, or hardly thinking...]



PS - I hope all is well with you, thanks for the feedback! 

[sahgwa]

They are pretending AI exists since it is a catchy term and makes idiots think that 'some smart machine is going to give me all the right answers, and maybe even do things for me I dont want to do'

[ArMaP]

I don't often ask for a response about myself, but to anyone reading this,... 
does it really seem that what I "hate" is AI itself?

To me, the idea you give is that you hate the use of the name "AI" and are almost obsessed with it.

The use of "AI" is just a marketing trick, like many others that have been used before. In this case there's a huge push from the industry because to have all that (almost) real time answers to our questions, Large Language Models (the AI in fashion at the moment) need huge processing power and storage capabilities. Also, while a general LLM like ChatGPT may be used to ask question about specific topics, a specifically trained LLM will be much better, and training them needs even more hardware resources.

PS: whenever a new technology appears, it is usually over used and over hipped. The same thing happened with X-Rays, with people using them in everything, just to find later that they were dying because of the excessive exposure to radiation. LLMS may not have that direct health effects, but using LLMs as if they know all the answers may.