AI = Unemployment?

[mr.WHO]

Then enlighten us with your wisdom.

Apparently you understand something completely different from AI than 99% from IT and engineering. Maybe it's different in Poland, but AI and informatics are like chicken and egg.

We call game AI an AI, even if we know it's not.

ChatGPT and current AI is the latest inverstor buzz like crypto and NFTs (functional technology, but misconstruded and twisted by snake oil salesmen).
Just like those two, it's easy to sell BS to the masses.


The correct term for those "Chatbots" should be Machine Learning, not AI - that's why every IT and Engineer specialist will laught at you thinking that Chatbot can become sentient AGI.

Edit:
Actually, there was a story of Google AI specialist who got convinced by Chatbot AI that it became sentient

[EGO_Aut]

Then enlighten us with your wisdom.

Apparently you understand something completely different from AI than 99% from IT and engineering. Maybe it's different in Poland, but AI and informatics are like chicken and egg.

We call game AI an AI, even if we know it's not.

ChatGPT and current AI is the latest inverstor buzz like crypto and NFTs.
Just like those two, it's easy to sell BS to the masses.


The correct term for those "Chatbots" should be Machine Learning, not AI.

A simple example of AI is chess. Impossibly large amounts of data would be required to teach all moves to a computer. Despite this, software still manages to beat the best chess players - is the software stupid or intelligent? Doesn't intelligence have something to do with making "good" (I don't want to say correct) decisions?

[mr.WHO]

A simple example of AI is chess. Impossibly large amounts of data would be required to teach all moves to a computer. Despite this, software still manages to beat the best chess players - is the software stupid or intelligent? Doesn't intelligence have something to do with making "good" (I don't want to say correct) decisions?

Intelligent and intelligence is not the same.

My vacuum cleaner is intelligent, but it has no intelligence.

[Observe]

The correct term for those "Chatbots" should be Machine Learning, not AI.

How is Machine Learning not Artificial Intelligence? Organic lifeforms learn in different ways and possess what I suppose we could call 'Natural Learning'. Much of our learning is associated with sensations, emotions, desires etc., which machines do not have in the same way and may never have. Are we talking about consciousness? If so, it'll be a long discussion, because none of us even knows what that actually is.

[mr.WHO]

How is Machine Learning not Artificial Intelligence? Organic lifeforms learn in different ways and possess what I suppose we could call 'Natural Learning'. Much of our learning is associated with sensations, emotions, desires etc., which machines do not have in the same way and may never have. Are we talking about consciousness? If so, it'll be a long discussion, because none of us even knows what that actually is.

Both Humans and animals are to some extend capable of breaking from their programing (genetics, instincts, habits) and create something new.

Machine Learning is simply mathematical/statistical process of input/output.


Let me give you an example - my Dad had a briefcase with a code lock combination from 000 to 999.
Machine learning would be to bruteforce every 999 combination until you hit a right one.
Intelligence would be finding the habits/information of a person and figuring out which of this information could be use to break the code (e.g. birthday).

You can add a code to Machine Learning to cross reference birthday database and bruteforce it as well, but again - there is no sentient intelligence, only code.

That's why current generation of "AI" is so attractive, because it can bruteforce through enough mundane data to statisticaly calculate possible output.
It's very powerful, but there is no more intelligence here, than in the calculator in your smartphone (which isn't smart, contrary to the name).

As I mention, this could be useful to find cure for cancer, but it won't be the AI who will invent the cure. It will be the people who will enter right input data to the AI and people who will be able to make the cure, based on statistical data output.

True sentient AGI would find on it's own that the cancer exist, gather research data, process it and then go to invent the cure on it's own, without anybody asking.

[Alan Phipps]

So why or when would you use supervised AI/mechanisation over a supervised traditional human workforce?

I would suggest AI/mechanisation would become more attractive due to some combination of being faster, more efficient with resources and working environment, less error-prone, lower overheads/running costs (including its spread development, installation and upkeep costs), and reduced holidays/industrial disputes/days sick. Now, as to the ethics of this, my gut feeling is that the current human workforce should first be given every opportunity to provide approaching comparable business benefits and efficiencies before any decision is made.

The reality is that a human workforce or their representatives may reject and impede all and any efficiencies and modernisations, and may even go so far as to actively sabotage such efforts. This attitude can quite possibly damage their own future employment prospects - but that is obviously not applicable to all current commercial/industrial instances and situations.

[chew-ie]

Then enlighten us with your wisdom.

Apparently you understand something completely different from AI than 99% from IT and engineering. Maybe it's different in Poland, but AI and informatics are like chicken and egg.

We call game AI an AI, even if we know it's not.

ChatGPT and current AI is the latest inverstor buzz like crypto and NFTs (functional technology, but misconstruded and twisted by snake oil salesmen).
Just like those two, it's easy to sell BS to the masses.


The correct term for those "Chatbots" should be Machine Learning, not AI - that's why every IT and Engineer specialist will laught at you thinking that Chatbot can become sentient AGI.

Edit:
Actually, there was a story of Google AI specialist who got convinced by Chatbot AI that it became sentient

"99% IT & engineering" is no homogen field. As is the rest of people who come in contact with "AI" these days. So for the sake of simplicity we agreed on using the term AI very loosely. But if we go all formal on it, there are distinct differences and classifications. Technically it's a wrong name.

Like mr.WHO said - ChatGPT and game AI is nowhere near "strong AI". Game AI is just a set of rules & means to fake decisions (state machines, path finding algorithm, fuzzy logic and the like) while machine learning is just a specialised tool for dealing with some problems as well. Oh and speaking about "AI" - ELIZA is around for a very long time now...

But if we are to talk about "strong AI" - we have to hone the terms again.

--

RE the topic - jobs in general are evolving quite fast [these days]. Like back in the industrialisation new methods for automatisation are introduced. Digitialisation is slowly integrating all the tools we know for decades and this process is not new. I still don't believe that "todays AI" can replace programmers. Besides the fact that we don't have strong AI yet all those changes will take a lot of time. At the end we are talking about humans - and laws (!) who have to change as well. And there is also the question if we really want this - do we really want to have only 1 person where could be 5. There is also the human factor - and we might be able to work solo on a job. But we miss out the benefits peers can have.

But nowadays we throw a lot of programmers onto all kind of problems who might be wrongly deployed. And in those areas automatisation tools might be the better answer. But we still need a lot of real brains to orchestrate the partial results.

A example: we truly don't need a horde of programmers these days to code yet another shop software. Or a website ... Yet it used to employ whole teams to achieve just that.
Code refactoring would be another example - there is a huge toolset around these days which shortens the time quite a bit for this particular task. So at the end we need less people (but still need some).

[mr.WHO]

So why or when would you use supervised AI/mechanisation over a supervised traditional human workforce?

I would suggest AI/mechanisation would become more attractive due to some combination of being faster, more efficient with resources and working environment, less error-prone, lower overheads/running costs (including its spread development, installation and upkeep costs), and reduced holidays/industrial disputes/days sick. Now, as to the ethics of this, my gut feeling is that the current human workforce should first be given every opportunity to provide approaching comparable business benefits and efficiencies before any decision is made.

The reality is that a human workforce or their representatives may reject and impede all and any efficiencies and modernisations, and may even go so far as to actively sabotage such efforts. This attitude can quite possibly damage their own future employment prospects - but that is obviously not applicable to all current commercial/industrial instances and situations.

The most simple and safe solution would be to implement AI first to task where there are shortages of labor anyway.
There is shortage of skilled medical diagnostic staff, no matter if you're in 3rd or 1st world we are not swimming in Oncologist. Existing human workforce could then easilly requalify to supervise/train/improve AI diagnostics and there would be possibility, as the AI proliferate, that you would still need to increase the number of human oncologist/diagnostic to supervise.

Same with various complex science fields (e.g. quantum physics, rocket science) or medical research...


...unfortunately we all know it will be first and foremost applied to low-skilled and most numerous labor, with no perspective of re-train or work retention.
If you're concept artist, Hollywood writer, actor or voiceactor...you're screwed.

[fiksal]

Good point, in such a society the government could just let the unemployed starve because they don't contribute anything to the economy (obviously I hope that no government would consider that).

If people will starve when the food is abundant perhaps the economy and government of your/our choice is wrong?

do you think that system serves the people?

[fiksal]

The correct term for those "Chatbots" should be Machine Learning, not AI.

How is Machine Learning not Artificial Intelligence? Organic lifeforms learn in different ways and possess what I suppose we could call 'Natural Learning'. Much of our learning is associated with sensations, emotions, desires etc., which machines do not have in the same way and may never have. Are we talking about consciousness? If so, it'll be a long discussion, because none of us even knows what that actually is.

I think the term AI is too wide. We call anything that can compute from a data set an AI.

If we are taking about human levels of intelligence then we have a long way to go. I am not aware we have yet mathematicaly proven that machine can think like us or that we can think like them. Still it's maybe possible eventually.

Consciousness is indeed a different topic still. It's a human concept applicable to us and animals built from similar stuff.

[schorsch_76]

It's not AI that will take jobs, it's the capital that will maximize profits by cutting costs.

If we ever switch to automation in everything we do, we should really think about who we want this to benefit - people collectively, or select few.

Overall, I am optimistic. In this world, one have to be ready to learn new trades

You know what a pessimist is?


An optimist with experience.

[fiksal]

You know what a pessimist is?


An optimist with experience.

A pessimist says - it cant get any worse then this, an optimist says - yes it can!

[mr.WHO]

I am not aware we have yet mathematicaly proven that machine can think like us or that we can think like them. Still it's maybe possible eventually.

Consciousness is indeed a different topic still. It's a human concept applicable to us and animals built from similar stuff.

That's what I'm trying to say about current AI/computers - they don't think, they calculate.
Same you can't say that analog typewritter think, it process human finger input to paper ink string.
Same you can't say hammer hitting a nail is thinking.

No matter how complex and powerful the calculation are, they won't magically turn into thinking or sentience.

It's rather difficult to define true Intelligence, but to me it require two things:
- capability to make new abstract concepts
- out of the box initiative

Both Human and animals are capable to do this:
e.g. Various birds, apes, doplhins and octopus are capable of making and using tools.
Various birds, dogs and cats are able to understand human words and connect them to concepts like shape, color, application, some animals are able to count.
Hell, even animals that we usually treat as dumb have some level of intelligence - like Cows listenting to music and showing they have a taste to various music styles.

Current AI is still lightyears behind even basic animal intelligence - the difference is that now AI is actually good at imitating, but it's just an imitation - a bit more advanced tamagochi.

[Chips]

That's what I'm trying to say about current AI/computers - they don't think, they calculate.
Same you can't say that analog typewritter think, it process human finger input to paper ink string.
Same you can't say hammer hitting a nail is thinking.

What a weird comparison.

Define "thinking"... I'll wait - because at the face of it you can throw a one liner. But delve into the terms utilised and expand on them further, and you'll get to multi paragraph. Breaking it down to the ultimate atomic parts is likely beyond most of us But once it is, we may have a basis to work from when considering "artificial" mixed with other things.

Essentially, if you're trying to have a legitimate discussion, you need to align definitions, define the parameters and so on - so everyone is on the same page, talking the same thing, within the same ballpark.
A lot of what I'm reading seems to be throwing the term "AI" around and then seemingly forgetting the entire point of it including the A; Artificial.

AI is a very very very broad term. For example, is a thermostat artificial intelligence based on the job it performs?

[mr.WHO]

Define "thinking"... I'll wait

Didn't I describe it in the rest of paragraphs?

AI is a very very very broad term. For example, is a thermostat artificial intelligence based on the job it performs?

Thermostat doesn't have a concept of temperature, only a mechanical measurement.
Nor it has a concept of what measurement is for.
Nor can improvise or conceptualize alternative use for it's role on it's own.

By the same distinction music player is not intelligent, but various animals listening to music and starting to imitate or dance or even sing in their own way are intelligent.

[Chips]

Didn't I describe it in the rest of paragraphs?

No.

Thermostat doesn't have a concept of temperature, only a mechanical measurement.
Nor it has a concept of what measurement is for.
Nor can improvise or conceptualize alternative use for it's role on it's own.

By the same distinction music player is not intelligent, but various animals listening to music and starting to imitate or dance or even sing in their own way are intelligent.

You've not defined what artificial intelligence is. "It doesn't have a concept..."
Since when does "concept" come into a definition of artificial intelligence? I specifically asked if it was an example of artificial intelligence.

As I said, it's not possible to have a meaningful discussion when we're not even talking about the same thing is it Instead of talking AI we've got "concepts" and "conceptualise". Unaware those were classic AI definitions.

By the same distinction music player is not intelligent, but various animals listening to music and starting to imitate or dance or even sing in their own way are intelligent.

So observing the animals dance or sing in their own way indicates intelligence. Okay... https://youtu.be/BTYX34OYW-E?t=40

[mr.WHO]

You've not defined what artificial intelligence is. "It doesn't have a concept..."
Since when does "concept" come into a definition of artificial intelligence? I specifically asked if it was an example of artificial intelligence.

As I said, it's not possible to have a meaningful discussion when we're not even talking about the same thing is it Instead of talking AI we've got "concepts" and "conceptualise". Unaware those were classic AI definitions.

By the same distinction music player is not intelligent, but various animals listening to music and starting to imitate or dance or even sing in their own way are intelligent.

So observing the animals dance or sing in their own way indicates intelligence. Okay... https://youtu.be/BTYX34OYW-E?t=40

To some extend "concept..." could be understand as variable in coding language.
Human coder must input and define variables for current AIs, depending on the task they need to perform.

Intelligence would be, if we just dump a data set on the AI and it would be able to "conceptualize" the variables (some AI can do this) and task itself on it's own (this AI cannot do yet).


Music player is not intelligent, it just do the pre-programmed task of playing music.

Dog is intelligent, because when he hear human music, he's able to grasp human music concept (input) by extrapolating dog concept of social pack howling as closest analog (variables) - then he try to improvise (create and execute task of howling to human music).
Talking birds are even better example, because best of them are able to dynamically conceptualize human words meaning to things like objects, shapes, color, numbers.
You can ask Parrot - "bring me red circle cookie" and Parrot will fly and bring you the red circle cookie.
Shape reckognition cameras can identify red cicle cookie, but it's not dynamic, out of the box, but static (uploaded, defined and coded by human) - it won't connect on it's own to nearest mechanical arm to bring you the cookie, unless human specifically code this interaction - thus no intelligence.


In some narrow, very specific and technical way Machine Learning can be compared to Animal training.
The difference is that AI are static (narrow minded, limited by code, inflexible, no intelligence) and Animals are dynamic (more flexible due to intelligence).

To finish this - it's more likely we will have a revolt from "Planet of the Apes" long, long looooooong before we will ever be in danger of AI becoming AGI.

[mr.WHO]

On some deep philosophical level, you could say that Humans and animals are just this - a sum of pre-defined code (sum of lived experince, interactions, evolutionary path).


If AI would be able to:
- dynamically assimilate new code and tasks
- dynamically be able to asses code/task priority/usefulness and delete redundant code to prevent overflow


Then I'd have a tought nut to crack.
Beyond some critical mass of asymilated data, we would enter uncanny territory, where we wouldn't be able to tell, if the AI is intelligent or not.
For sure it would be at least a perfect imitation of intelligence - probably a sure Nobel prize award for someone who will make it.

[fiksal]

I feel like we need to go for basics without using philosophy.


Computers are deterministic calculating machines.

As far as we know humans are not. There's nothing magical about our brains but the wires are not connected the same way as we build computers. Can we ever build a computer that can think like us, I honestly I don't know. But I would ask a question - why would we do that?


Machines are very good at what they do, they can perform complex calculations fast, a lot of them, and all to predictable results, repeatable, testable results. Thus we say deterministic. That's a very important word.

Humans actually don't and can't. We are much slower on calculations. But we are fast on solving multiple parallel problems, that we struggle even defining for AI. As far as we know we are specifically not deterministic machines.

I am not a neuro scientist, so my view on this is only from mathematics.

Chatting software, deep learning software, is adapting mathematical formulas over data to the result that we tell it we expect. Emphasis on Expect.

In other words chat bots are not seeking to find the answer, they are calculating what answer we find most probable, whether that answer is right or not, chat bot doesn't know. Unless we tell it. Chat bots in fact don't know anything it responds with exists, or that you exist.

In my opinion we can never say currently built machines are conscious or even intelligent, ever. If you want that, we need a whole different micro processor, that is a brain and not a processor.

And I will finish with the last question of why. Machines are good at tasks that we are not. Complex learning human like non deterministic machine would have same flaws as us, it'd unreliable in production of anything. There's no point to make them.

[mr.WHO]

That's well said.

You can't magically have deterministic tool turning into non deterministic...person?

Thus my example of Hammer unable to gain Intelligence by hitting enough nails.



We are nowhere near in danger of making Skynet or Cylons, but we are currently making a very good next-gen Tamagochi (with build in Google search, Wikipedia, YT algoritm, Photoshop, Voice mixer + modularity to futher add hundreds of more addon features).