AI = Unemployment?

[fiksal]

[felter]

I haven't been following what's being said here, so I don't know if this has been mentioned or not. I stumbled onto this, now it isn't quite about employment, but then again it could affect voice actors I suppose, but it is just so gawd darn cool and the future potential is just mind-blowing, and I'm looking forward to how it turns out in the future where we can have some kind of conversation with a NPC rather than the limited choices that we have right now.

[Alan Phipps]

@ felter: That's sort of why this topic was recently discussed.

[Chips]

You can ask Parrot - "bring me red circle cookie" and Parrot will fly and bring you the red circle cookie.

I can safely say that is utterly false. I can safely say that because you're missing something from this...

I'll say it again as it keeps going over peoples heads... why are you talking about "actual intelligence". The topic is Artificial intelligence; my questions are framed around artificial intelligence. There seems to be a real issue with people understanding the term "artificial". Lets put it into capitals, just to be crystal clear we're supposed to be talking about the same thing. ARTIFICIAL. So to not sound completely like an arse without explaining, I think it's really important (as said earlier) to realise what Artificial intelligence means. It's nothing to do with whether it can think like a human... You may keep trying to make it into that, but that's not what AI is.

Also, I'm guessing people don't have a background in the subject either. I mean I don't, but I do realise from things being said people haven't read around much...

So, can robots learn without humans? Yes.

https://www.technologyreview.com/2020/0 ... -learning/
https://news.mit.edu/2018/model-helps-r ... umans-1004
https://arstechnica.com/gadgets/2021/10 ... nts-page=1
https://www.vox.com/2017/3/23/14962182/ ... i-research

Lastly, if you really really need to keep trying to talk about whether it's "real" intelligence - humans aren't born immediately walking, talking, writing, saying "red circle" and so on. They get years of exploring their environment, and more - and for what we'd recognise as "normal" involves years of being taught by other humans, not just sticking their finger up their nose.
Yes, there's also examples of programs created where... they have AI "learning" without much human interaction at all (obviously there has to be some form of way to determine good/bad outcomes to learn )


ChatGPT probably attracted so much attention because people can interact with it and it's incredibly impressive for most. AI has existed for decades, it's been used to analyse scans for decades to detect cancer, and algorithms are created that can recognise way before humans. Then again, apparently so can a dog's nose in some cases Some AI, as has been pointed out, is far better than a human in specific tasks. But don't forget... it's "artificial".
Can it replace human jobs? Yes, it has. It will continue to do so over time. Doing so doesn't mean it's a sentient being though

(by the way, there is an AI that can solve problems such as putting coloured blocks on top of each other as asked by the person, but can't remember name and my google-fu is weak -- but essentially just like Parrot. Of course, just like Parrot, it had to be taught what certain things meant via rewards for getting something right ).

The obvious question is - can a computer ever achieve what you put as the measure of, erm, intelligence? At which point, would you start calling it artificial intelligence? or just intelligence? Hoping that may clear up why I keep banging on about artificial

[mr.WHO]

You can ask Parrot - "bring me red circle cookie" and Parrot will fly and bring you the red circle cookie.

I can safely say that is utterly false. I can safely say that because you're missing something from this...

I'll say it again as it keeps going over peoples heads... why are you talking about "actual intelligence". The topic is Artificial intelligence; my questions are framed around artificial intelligence. There seems to be a real issue with people understanding the term "artificial". Lets put it into capitals, just to be crystal clear we're supposed to be talking about the same thing. ARTIFICIAL. So to not sound completely like an arse without explaining, I think it's really important (as said earlier) to realise what Artificial intelligence means. It's nothing to do with whether it can think like a human... You may keep trying to make it into that, but that's not what AI is.

And we are back to my original point that the term Artificial Intelligence if incorrect, because there is no intelligence.
Machine Learning, good.
Intelligence Imitation, good.
Imitative Intelligence - not good, because there is no intelligence behing AI imitating stuff and the proof is in YOUR example below:

So, can robots learn without humans? Yes.

https://www.technologyreview.com/2020/0 ... -learning/
https://news.mit.edu/2018/model-helps-r ... umans-1004
https://arstechnica.com/gadgets/2021/10 ... nts-page=1
https://www.vox.com/2017/3/23/14962182/ ... i-research

Do you know how AI "learned" to walk?
In simplest term:
Human coded a stickfigure joints
Human coded a stickfigure joints movement limitations
Human put AI stickfigure in walking scenario with obstacles....and then AI bruteforce every possible joint movement combination - there is only limited number of joint movement that will commit the movement and only limited amount of movement that allow effetive walk. Once you bruteforce through all the combination you will be able to find out most effective way of movements - there is no intelligence here, just bruteforce statistical data.

Not to mention that if human will make a mistake and misdefine some variables, AI was able to make absolutely rudiculous and hilarious ways of walking, because bruteforce statistics output found out that with those incorrect parameters, this strange and unnatural way of walking is the optimal one:

https://www.youtube.com/watch?v=gn4nRCC9TwQ

Physical robot movement is a bit better, because it's based on real/material environment (instead of virtually made stickfigure course), but the logic is the same - if human coder will screw up with input data and joint parameters, robot will fail miserably (what you can see in articles is usually only the best and most sucessful outcomes, while 99,99% of failures are not making into articles).

[mrbadger]

I've been testing both ChatGPT and Bard on various tasks, writing, programming and some other things. Also various AI image creation ones.
Yes they work. If you want a basic image or something for an ad campaign you don't have to pay royalties for? Ideal. If you want something that'll stop people in their tracks? Not so much.

The writing and programming? Well it's got a long way to go. Form letters? Yes, but if you've not already got those what is your business doing? Everything else needs further work after being generated.

[Observe]

The obvious question is - can a computer ever achieve what you put as the measure of, erm, intelligence? At which point, would you start calling it artificial intelligence? or just intelligence? Hoping that may clear up why I keep banging on about artificial

We don't understand human intelligence and so our ability to define what is 'artificial' vs 'real' intelligence is somewhat lacking. Clearly, artificial intelligence is different from organic intelligence, even though they may have analogous functions.

Humans (and others) have sense organs, which put us into contact with various external phenomena. We also have the sense of thought, which gives us awareness of thought. Initially, there is not much difference between an eye and a camera. Both register visual contact without any judgement or response. Next we evaluate the object of our vision to determine if it is good, bad or neutral based on familiarity. Humans do this in their mind (or wherever). AI can do it with a neural network using a dataset. Pattern recognition in both cases.

What happens next, is perhaps the biggest distinction between artificial intelligence vs. the natural intelligence shared by organic lifeforms. Using the above example with the eye sense, when we decide that what we are seeing is good, bad, or neutral, we generate sensations in our body (fear, lust etc) and we take action (or not) accordingly. Artificial intelligence cannot feel, it does not have sensations or emotions and as such, is fundamentally a very different form of intelligence that ours.

Until AI can experience sensations (have feelings), it must remain a tool to serve us, not the other way around. If AI is ever able to feel in the same way we do, then we must consider whether it is 'alive' and deserving of civil rights under law. I think we are a long way from that sort of sentient AI, but then I wouldn't be surprised to see it's emergence any time.

[mrbadger]

Until AI can experience sensations (have feelings), it must remain a tool to serve us, not the other way around. If AI is ever able to feel in the same way we do, then we must consider whether it is 'alive' and deserving of civil rights under law. I think we are a long way from that sort of sentient AI, but then I wouldn't be surprised to see it's emergence any time.

That would be AGI, and we're not even close to achieving that. Possibly may never be unless a fundamental change occurs in mathematics, which is seriously unlikely to occur any time soon.

[fiksal]

I can make a python script to feel things in a few minutes (decision tree + a few inputs). It's still wouldnt qualify for human level of "inteligence" due to determinism. (Referencing here possibly the long and borning post I wrote)

[mr.WHO]

Ironically, I think that the answer to how we can live with proliferating AI bots is in Medieval times.

No, I don't mean crusades

Talking about comparing AI intelligence to animal intelligence and Machine Learning to animal training, made me think we already had this in Medieval times - Horses!

Back then Horses were one of most expensive and essential things in society and economy.
Horse handling and breeding was very imporant job, but to some extend, most of population was capable to handle horses as a common knowledge.


Now in present times, we would have similar case.
AI handlers/trainers/makers would be an important job, but basics of AI are within intelectual capabilities of most population.

We should start integrating basics of AI into school curriculum, so that AI basics would be common knowledge, like farming in medieval times.
Like now skills in MS Office is basic stuff, so would be the AI - a tool in human career path, not a competition.
It would be even a bit better, because unlike MS office or IT coding, AI should be somewhat attractive to both STEM and Art students.

Oh and it will help to have an open source AI, so we would have an alternative like with analogs to MS Office or Windows.

[Chips]

And we are back to my original point that the term Artificial Intelligence if incorrect, because there is no intelligence.
Machine Learning, good.
Intelligence Imitation, good.

Sorry, what's your background in this? Curious given "artificial" is being entirely ignored and "intelligence" is something you're absolutely stuck upon.
Furthermore, the arguments/points you are making indicate you've not read anything regarding the links, let alone the background. The simulations are done to prevent damage to an actual robot in the learning phase. The legs on said robots are articulated in a way; why would you run a simulation which doesn't match those param... oh forget it.

Honestly, this is ridiculous. Your attitude is "MY OPINION IS RIGHT AND THE ENTIRE WORLD OF ARTIFICIAL INTELLIGENCE RESEARCH BEST START USING ANOTHER WORD THAN INTELLIGENCE BECAUSE I SAY ITS NOT INTELLIGENT".

Again, Arti-bloody-ficial.
https://dictionary.cambridge.org/dictio ... artificial

made by people, often as a copy of something natural

But for sake of brevity, you're absolutely right. It's not genuine intelligence, and no-one ever said it was either. So, as you're now happy with the fact you're right... what's next? We can't discuss artificial intelligence because the word intelligence is involved. What do you propose the world changes to? I mean, should we stop saying "artificial flowers" and call them "plastic green things with purdy things on top" because, guess what, they're not an actual flower?

The sarcasm is the last ditch effort to get you away from absolute blinker of "but it's not intelligence"... - but I am definitely interested to know your background given what's being said, how it's said, and why the world intelligence is the key sticking point rather than the vastly wonderful, interesting and exciting world of Aritificial, erm, representation of behaviours and/or knowledge and/or decisions and/or ... can you come up with the phrase we should use?

[mr.WHO]

Sorry, what's your background in this? Curious given "artificial" is being entirely ignored and "intelligence" is something you're absolutely stuck upon.

Christ, here we go again.

Google definitions
Artificial - adjective - made or produced to copy something natural
Intelligence - noun - the ability to acquire and apply knowledge and skills


Now the very first link you provided:
https://www.technologyreview.com/2020/0 ... -learning/

First paragraph:
"Within 10 minutes of its birth, a baby fawn is able to stand. Within seven hours, it is able to walk. Between those two milestones, it engages in a highly adorable, highly frenetic flailing of limbs to figure it all out."


The premise of Intelligence in this article is completely false - the robot didn't learn jack sh*t in 10 minutes or seven hours.

It was human who:
- build the robot joints
- measured joints parameters and movement scope
- gather or generated database of said movement
- made a simulation model for robot movement
- run simulation couple (hundred or thousand) times and improved (trained the model) to exclude movements that do not work or are inneficient/worse than other
- then linked physical robot and run real life test runs for additional data (e.g the Google digital stickfigure video - AI made it run in bizzare way, because it didn't accounted to energy waste of movement, or used accurate enough model of gravity and joints, comparing to real life) - again hundreds or thousands test runs to further improve movement model.

Then there is an end result for poster article - we have final movement model that we spent tens of thousand hours to perfect, we chain a fresh/blank robot and let it run for 10 minutes and bam - clickbait article that the robot has learned to walk on it's own - which is COMPLETELY FALSE.

It's like claiming river is capable of learning, because in given set of physical laws, the water is able to "find the way" from source to ocean.
Even if humans will build a dam, the water will be able to either overcome the dam or evaporate and fly in clouds, yet there is no intelligence here.

Same is with AI Machine Learning - there is no intelligence, just a bruteforce of big data and calculating power. If the answer is in the data, with proper model it will be found in same way water will find the way to the ocean.
On the opposite, if the data doesn't contain the answer or the model is wrong, there will be no success no matter how powerful the AI is.


Going back to initial definition:

Artificial - adjective - made or produced to copy something natural
Intelligence - noun - the ability to acquire and apply knowledge and skills

AI is incapable to acquire and apply knowledge and skill - it's done 100% by humans (input made by humans, model and variable made by humans, task defined by humans, taks execution success defined by humans).

I don't know why you're so fixated on "Artificial" - the artificial part is 100% correct, it's the Intelligence part that is incorrect.

[Chips]

And your background is?

p.s. How many hours from birth does it take a human to walk?

By the way, you talk about the article but have not read it. Again.

That’s the idea behind AI-powered robotics. While autonomous robots, like self-driving cars, are already a familiar concept, autonomously learning robots are still just an aspiration. Existing reinforcement-learning algorithms that allow robots to learn movements through trial and error still rely heavily on human intervention. Every time the robot falls down or walks out of its training environment, it needs someone to pick it up and set it back to the right position.

Now a new study from researchers at Google has made an important advancement toward robots that can learn to navigate without this help. Within a few hours, relying purely on tweaks to current state-of-the-art algorithms, they successfully got a four-legged robot to learn to walk forward and backward, and turn left and right, completely on its own.

Pay attention to the last paragraph please. And you do realise I focus upon the artificial part of artificial intelligence because it implies the intelligence is "fake", "not real" etc? So... yeah? We agree on that? Right? What are you arguing about then? The examples are all "fake" intelligence. Okay? We agree?

Where the fake intelligence exhibits behaviours that its, well, learned. Right? So what would you like us to call this area instead of artificial intelligence to get over this hump? Suggest alternatives. Also, what's your background in all this?

[mr.WHO]

And your background is?

p.s. How many hours from birth does it take a human to walk?

I ran a few Machine learning models in my job, know a few programing languages and interact with multiple bot AIs.
Big data processing is my main focus and the AI is the thing that will be quite useful in my area.


Your human walk example is not accurate:
The AI walking bot example would be
a human teaching child how to ride a bike,
except that he glue and duct tape child to the bike, never let go the bike and pretending that the child now know how to ride the bike.
It's just an imitation of riding.
You could lobotomize the child or put the doll on the bike and it would ride the same - an imitation, no actual intelligence or learning.

[Chips]

I am dumbfounded by your background given this thread then

[mr.WHO]

That’s the idea behind AI-powered robotics. While autonomous robots, like self-driving cars, are already a familiar concept, autonomously learning robots are still just an aspiration. Existing reinforcement-learning algorithms that allow robots to learn movements through trial and error still rely heavily on human intervention. Every time the robot falls down or walks out of its training environment, it needs someone to pick it up and set it back to the right position.

Now a new study from researchers at Google has made an important advancement toward robots that can learn to navigate without this help. Within a few hours, relying purely on tweaks to current state-of-the-art algorithms, they successfully got a four-legged robot to learn to walk forward and backward, and turn left and right, completely on its own.

Pay attention to the last paragraph please. And you do realise I focus upon the artificial part of artificial intelligence because it implies the intelligence is "fake", "not real" etc? So... yeah? We agree on that? Right? What are you arguing about then? The examples are all "fake" intelligence. Okay? We agree?

My answer stands and cover above paragraps - it's you who can't read the article:
'relying purely on tweaks to current state-of-the-art algorithms"

[Chips]

AI is incapable to acquire and apply knowledge and skill - it's done 100% by humans (input made by humans, model and variable made by humans, task defined by humans, taks execution success defined by humans)

So to sum up the last page - I'm saying "it's artificial" and then you're going "no, it's not real intelligence" and so I'm saying "yes, here's an example of artificial..." and you're saying "No, this is not real intelligence because..." and I say "so it's artificial is what you're saying" and you go "no, it's not real intelligence" - so I say "so, artificial?" and you say "no, it's not real intelligence".

You get where I've been coming from yet? Because while it's amusing, I'm very bored of it - and the thread is de-railed from talking about how, (erm, what are we calling it?) may cost jobs.
It may. By another name.

[mr.WHO]

You get where I've been coming from yet?

I get that by your definition nearly everything can be defined as Artificial Intelligence, which is incorrect.
Maybe my definition is too narrow, but your is for sure too broad.

[Chips]

You get where I've been coming from yet?

I get that by your definition nearly everything can be defined as Artificial Intelligence, which is incorrect.
Maybe my definition is too narrow, but your is for sure too broad.

No, you really don't. I've not defined the breadth of AI (I posed a question, I didn't answer or provide any argument for/against, of something that would be an incredibly broad term -- but is a good starter for students to consider what AI could mean; to ask questions, to think, to consider, to engage).
I've used bonafide links of recognised AI fields. Fields that have thousands of text books written, taught in universities, research departments, hundreds of thousands of proponents of the field around the world. You, it seems, are dismissing them all as not "AI", because it's not actual intelligence. Which means its artificial, right? But no its not actual intelligence. So artificial? No, not actual intelligence. So...

I'll leave it here because frankly, others deserve their ability to comment in the thread. Should have abandoned it half a page ago >.<

But - artificial - fake, not real, not actual, created by people - like programmers, who define parameters, wrote the code, ran simulations or just let the robots fail - that dictionary definition of artificial, made by people... intelligence.

[mr.WHO]

Term Artificial Intelligence - it order for definition to be true, both elements need to be true.

Artificial (true) and Intelligence (false) = false

You focusing on Artificial mean:
Artificial (true) or Intelligence (doesn't matter if true/false) = anything that is artificial meet the definition even if there is no intelligence. My coffe machine meets this definition.

Me focusing on Intelligence mean:
Artificial (doesn't matter if true/false) or Intelligence (true) = anything that is intelligent meet the definition. ChatGPT doesn't meet this definition, but my dog does.


Now that article that you like to mention - every single paragraph of that article covers and back up all the examples I provided TO THE LETTER (they write about human input, modifications and tedious oversight several times).
The article itself admits that the title is just a clickbait.