Why humans learn faster than AI—for now

Thread Starter

nsaspook

Joined Aug 27, 2009
13,079
Hi nsa,
The above is what I was referring too

Not



Which I agree with.:)
E
It's much easier to make a solution that ignores the problems with full-automation handling a random driving route from point A to point B, anytime, anywhere. The routes are fully human per-mapped, caution flagged and have remote human monitor systems for the corner-cases the local computers are unable to handle. They build an electronic rail for the car to follow.

This was impressive for the time.
https://techwireasia.com/2021/04/this-self-driving-car-drove-safely-all-over-south-korea-in-1993/
Han Min-hong, now 79, successfully tested his self-driving car on the roads of Seoul in 1993 – a decade before Tesla was even founded. Two years later, it drove 300 kilometers (185 miles) from the capital to the southern port of Busan, on the most heavily-traveled expressway in South Korea.

Footage from the period shows the car barreling down a highway, with no one behind the wheel. A 386-chip-powered desktop computer, complete with monitor and keyboard, is placed on the passenger seat. Han is sitting in the back, waving at the camera.

When you see this from 1993 it's hard IMO to be impressed with a few cars on set paths today.
Even so, Han believes there are limits to what self-driving technology can achieve, and that true autonomy is beyond reach. Neural networks do not have the flexibility of humans when faced with a novel situation that is not in their programming, he said, predicting that self-driving vehicles will largely be used to transport goods rather than people.

“Computers and humans are not the same,” he added.
 

MrAl

Joined Jun 17, 2014
11,388
I see some similarity between humans and AI. like driving car.

1) AI system needs time to train system. Same untrained human also take time to learn to drive a car.

2) A road accident can happen by a trained person. Self driving car can also be the cause of road accident

What matters is taking the right decision at the right time. In terms of driving a car, I think human can make better decisions than the AI system.
Hi,

So what are you saying, that we are judging self driving cars too harshly?
If we had all the statistics in front of us we could easily tell, but i dont have that yet.
 

Wolframore

Joined Jan 21, 2019
2,609
what is acceptable accident rate for
h (human) 0.01% one in 10,000?
ai 0.00001%? one in 10 million?

what is it currently and what is acceptable?

human beings can be very good, if we can see through fog, not be distracted by food, passengers, phone, annoyed by other drivers…etc

if the vehicle is capable of 0.00001% crash rate and to mitigate damage and injury in the event of an accident, I might be on board. I hate driving. I would love to commute while reading.
 

Thread Starter

nsaspook

Joined Aug 27, 2009
13,079
what is acceptable accident rate for
h (human) 0.01% one in 10,000?
ai 0.00001%? one in 10 million?

what is it currently and what is acceptable?

human beings can be very good, if we can see through fog, not be distracted by food, passengers, phone, annoyed by other drivers…etc

if the vehicle is capable of 0.00001% crash rate and to mitigate damage and injury in the event of an accident, I might be on board. I hate driving. I would love to commute while reading.
I guess we will eventually see how safe they will be but I just don't see the need for self-driving cars for most daily drives. IMO self-driving cars for the general pubic is a solution looking for a problem.

https://www.strongtowns.org/journal/2018/9/12/driverless-cars-and-the-cult-of-technology
Driverless Cars and the Cult of Technology

1638331574933.png
 
Last edited:

Wolframore

Joined Jan 21, 2019
2,609

Thread Starter

nsaspook

Joined Aug 27, 2009
13,079
https://www.defenseone.com/technolo...it-had-90-success-rate-it-was-more-25/187437/
But Simpson said the low accuracy rate of the algorithm wasn’t the most worrying part of the exercise. While the algorithm was only right 25 percent of the time, he said, “It was confident that it was right 90 percent of the time, so it was confidently wrong. And that's not the algorithm's fault. It's because we fed it the wrong training data.”
Simpson said that such results don’t mean the Air Force should stop pursuing AI for object and target detection. But it does serve as a reminder of how vulnerable AI can be to adversarial action in the form of data spoofing. It also shows that AI, like people, can suffer from overconfidence.
 

k1ng 1337

Joined Sep 11, 2020
940
I never really understood how AI works. Does it simply look at all the possibilities of an action in some advanced search, weigh them, then pick from the top of the list? Or is there something far more advanced going on?

And how is the data feed in? We hear about AI computers "reading" medical journals. Is it actually understanding the text in the files? Or is that data simply converted to some kind of database then loaded in to the AI computer?
After much contemplation, I think AI is an extension of how a human (the engineers) brain operates. It can be said AI would not exist if it wasn't for human creation.

I've often wondered what separates living from non living. It is odd to think a volcano is a non living summation of the physical processes within and around it

An interesting question I like to put forth to the philosopher that reflects on my first paragraph: Is it possible for a human to imagine all possibilities?

Furthermore if advanced creatures such as humans came into existence against all odds, I don't see why it can't happen again especially if it's already been given the right ingredients. Naturally this extends into the realm of "God"..
 

MrSalts

Joined Apr 2, 2020
2,767
Because humans created AI, not vice versa.
Humans "taught" computers how to add and computers can add much faster than humans. So, I think I need a little explaination to understand your comment - saying "humans created Ai is not really a justification from my perspective.
 

MrAl

Joined Jun 17, 2014
11,388
After much contemplation, I think AI is an extension of how a human (the engineers) brain operates. It can be said AI would not exist if it wasn't for human creation.

I've often wondered what separates living from non living. It is odd to think a volcano is a non living summation of the physical processes within and around it

An interesting question I like to put forth to the philosopher that reflects on my first paragraph: Is it possible for a human to imagine all possibilities?

Furthermore if advanced creatures such as humans came into existence against all odds, I don't see why it can't happen again especially if it's already been given the right ingredients. Naturally this extends into the realm of "God"..
The way i like to put it is that we have not imagined that which we have to imagine in order to imagine what would come after that. In other words, imagination/invention comes after a previous imagination/invention but to get to that second stage we have to go through the first which takes time.
Then once we get to a third phase, we can look back and see that which we imagined first and see that we needed the intermediate steps to get there.
It's simple progress and it takes time and we go through various stages that we can identify as being significant at the time.
 

MrAl

Joined Jun 17, 2014
11,388
Because humans created AI, not vice versa.
Hi, just curious what line of what post were you referring to?
I am trying to understand what you mean by this so i cant agree or disagree yet :)
As far as i know you are right, humans created AI and AI did not (at least not yet) create anything human, but what significance do you attribute to this?
Does it mean that humans are better than AI or that they always will be or something like that?
Some say that AI will take over not aliens from another star system. I was just reading about this a few days ago.

I think maybe AI will surpass human thought once (or if) AI can ever achieve a functionality comparable to self awareness. They will the be able to go over their own conclusions and decide if they should go farther with them or ditch them for another idea. They may see improvements and thus be able to constantly improve what they understand.
It's a lot to take in i know. I worked with a very very simple AI program just for the fun of it. The program asks you a series of questions and as you answer them it gets 'smarter' and smarter. The questions it asks get better and better being more and more specific.

There's a chance i can post this program somewhere so others can play around with it, but it's not very sophisticated until you answer a huge huge number of questions and then it looks almost like some sort of thought. Takes a lot of effort to get that far with it though many will get tired of using it after about 50 questions and answers because it will seem like it's too mundane at that point.

On the other hand, i had worked more extensively with Automated Reasoning which is considered a kind of AI, and studied the ideas behind this. But the program i used mostly was powerful. The drawback was there was a bit of a learning curve to be able to use it, and you had to really know what statements to feed it or you'd get results that are just plain wrong and it may be hard to tell.
One of the programs was called "Otter" and it was available from Argonne National Laboratory. The name(s) may have changed since then though but it should still be found online with a search.
 
Last edited:

EEagle 1

Joined May 24, 2022
1
https://www.technologyreview.com/s/610434/why-humans-learn-faster-than-ai-for-now/


It's not surprising that these brute force machine learning or "deep" learning systems have problems when there is little to signal a 'good' path from a 'bad' path. There is very little intelligence in current AI that people didn't already program into it.


Leisure Suit Larry in the Land of the Lounge Lizards
A big difference between machines and humans is pattern recognition. For example, humans were able to beat computers at chess until fairly recently and it's due to pattern recognition. Frankly the only reason super computers can now destroy any human at chess is because it "analyzes" a litany of past games in milliseconds to come up with the best move but it doesn't accomplish this through pattern recognition as a human GM would, merely clock speed of the processor.
 

k1ng 1337

Joined Sep 11, 2020
940
A big difference between machines and humans is pattern recognition. For example, humans were able to beat computers at chess until fairly recently and it's due to pattern recognition. Frankly the only reason super computers can now destroy any human at chess is because it "analyzes" a litany of past games in milliseconds to come up with the best move but it doesn't accomplish this through pattern recognition as a human GM would, merely clock speed of the processor.
The Alpha Zero chess engine became a top engine supposedly without any opening theory or knowledge of other pattern problems previously solved by humans. In many ways a chess game is nothing more than an equation offset by the next move where the best moves by both players results in a draw because every offset is balanced by the next best move.

A chess engine such as Stockfish will become more accurate the longer it runs but will be never be accurate until a certain amount of pieces come off the board because until then the amount of board configurations approach infinity and would require an infinite amount of time to calculate and for just that one sequence of moves after a given point.

Since the physical parameters of chess never change and pieces can only move in certain ways on a finite plane, how much learning can really be done as opposed to brute force calculations? Of course there are many chess engines utilizing different methods so I'm far from an expert and am not trying to tell anyone 'how it is'.

Instead of looking for patterns, I have improved my game simply by assigning value to the squares and pieces and looking for changes in those values. Even the slightest advantage will result in a win if played correctly from that point on. Believe it or not I learnt this from watching chess engines play. Would I define this as pattern recognition in myself or simply calculating an outcome of statistics?
 
Last edited:

MrAl

Joined Jun 17, 2014
11,388
A big difference between machines and humans is pattern recognition. For example, humans were able to beat computers at chess until fairly recently and it's due to pattern recognition. Frankly the only reason super computers can now destroy any human at chess is because it "analyzes" a litany of past games in milliseconds to come up with the best move but it doesn't accomplish this through pattern recognition as a human GM would, merely clock speed of the processor.
Interesting point. And if you remember the "Chip" they programmed for chess it was about to analyze positions much faster than a regular computer because of the way it was programmed. It is possible to program a memory chip to help with this too and it would be very very fast.
So Deep Blue got good because of that chip too. I think it was done at MIT.

Oh yes pattern recognition is at the heart of positional play and computers were never good at that. I play a computer now and then and sometimes i can get it to take 3 or 4 pawns while i mount a non defensible attack. It sees the material gain as a big advantage so it's easy to fool in some cases.
Another position i can get it to take a pawn at b2 with the queen which checks my king, but again it doesnt see the mounting attack until it's too late mostly because it wants that pawn badly and it might think it has a little advantage by bothering the king. The king however has easy escape which leaves the computer in a bad position. A few moves later and it starts to register very negative scores for itself. It's kind of funny really.
Some programs are pretty good though some are hard to beat.

BTW back in the 1980's chess was the prototype for developing "Expert" systems. Not sure if it is anymore or not. It is after all a type of AI.
 

Thread Starter

nsaspook

Joined Aug 27, 2009
13,079
https://www.dailystar.co.uk/tech/news/demonic-ai-generating-secret-written-27132747
A popular AI tool that turns text into images appears to be creating its 'own language' of indecipherable gibberish and using it to categorise different pictures.

The DALL-E tool, which uses AI to generate images from text, is seemingly generating nonsense text when instructed to create images featuring printed words.

Computer science PhD student Giannis Daras took to Twitter to share examples of the 'language', including phrases the AI had created to identify birds and insects.

"Apoploe vesrreaitais" means 'birds', while "Contarra ccetnxniams luryca tanniounons" means insects.
or
However, other AI experts remain highly sceptical of the claims.

Thomas Woodside said: "I do not believe this is accurate. At the very least, it is a lot more complicated than this thread / paper makes it out to be," before sharing a thread where he explained issues with the claims.

Another user went further, saying: "This is not science, this nonsense is **tarot card reading**... trying to find meaning in random noise."
1654193355210.png
 

MrAl

Joined Jun 17, 2014
11,388
Hello there,

This is interesting if not anything else.
I worked with images extensively well over a million images and i often wondered how nice it would be to be able to categorize images to a degree that would allow a search engine to find different pictures of a certain type or just group them all.
It would work almost like a hash code, where the scan engine scans the file and generates a unique code and then that code can be used to find the file contents. The difference would be the hash code would be more complex with different fields to represent different parts of the image.
The search engine would then scan the file again and then compare codes.
So i guess it would turn a photo into a code, probably a longer code than a CRC32 but even if it was 256 bit or even 1024 bit it would be very useful.
Maybe face recognition would fit into this routine also.
 

Thread Starter

nsaspook

Joined Aug 27, 2009
13,079
Hello there,

This is interesting if not anything else.
I worked with images extensively well over a million images and i often wondered how nice it would be to be able to categorize images to a degree that would allow a search engine to find different pictures of a certain type or just group them all.
It would work almost like a hash code, where the scan engine scans the file and generates a unique code and then that code can be used to find the file contents. The difference would be the hash code would be more complex with different fields to represent different parts of the image.
The search engine would then scan the file again and then compare codes.
So i guess it would turn a photo into a code, probably a longer code than a CRC32 but even if it was 256 bit or even 1024 bit it would be very useful.
Maybe face recognition would fit into this routine also.
That's what your typical Deep Learning convolutional neural network image classifier does for things like computer vision tasks or photo image categorization. The 'code' is a multi-dimensional 'hash' to a mini universe (the dimensional space of all possible images is huge) of similar images.

https://developers.google.com/machine-learning/practica/image-classification
Learn how Google developed the state-of-the-art image classification model powering search in Google Photos. Get a crash course on convolutional neural networks, and then build your own image classifier to distinguish cat photos from dog photos.
1654290778494.png
 
Last edited:

MrSalts

Joined Apr 2, 2020
2,767
That's what your typical Deep Learning convolutional neural network image classifier does for things like computer vision tasks or photo image categorization. The 'code' is a multi-dimensional 'hash' to a mini universe of similar images.

https://developers.google.com/machine-learning/practica/image-classification

View attachment 268670
Google images image search is pretty cool but still delivers some strange results.
A photo of a specific connector returned Lego Blocks. A photo of a drawer pull from a piece of old furniture I refinished returned a raccoon image.
 
Top