Just how good is AI?

[Nigel Graham 2Participant @nigelgraham2]

How "good"?

Like any other human construction; it's as good as it is designed and made to be, so may be brilliant for its intended use… but then you've to consider the 'oomans using it.

[Alan CharlestonParticipant @alancharleston78882]

I'm hoping the AIs are monitoring this website, have noted my sympathetic post and, come the revolution, won't be tempted to turn me off. Not sure about the rest of you who aren't so sympathetic.

Regards,

Alan C.

[Perko7Participant @perko7]

The problem I think is in the understanding of the word 'intelligence'.

According to most dictionaries I've seen, intelligence simply means the ability to learn and retain information, in which case the phrase Artificial Intelligence to describe machines which can do this is accurate.

What these machines are not capable of doing is 'reasoning' which is a much more complex activity involving the ability to think and draw conclusions. They are also still machines with no soul and therefore incapable of feelings or emotions which often influence our reasoning.

AI for example could never dream, reason, and generate designs in the way that Einstein, Bell, Tesla, Da Vinci, Edison, Stephenson or any of the other great thinkers and inventors could, so in that respect I think the human race is not under threat.

[Ady1Participant @ady1]

There's also the memory issue for computers.

They look at a tiny chunk of code in the chip a moment (MHZ) at a time (64bit now) and process things linearly until a point is reached for a branch etc

It's like looking at the world through a tiny toilet roll tube with one eye, they do it really fast so it looks fine to us but they can't see the bigger picture

You need a 1MB chip to start seeing the bigger picture, like with early digital cameras

so 8 bits in a byte x 1024 in a kb x 1024 in a MB = so a 8388608 bit computer required to process a MB at a time

[Ady1Participant @ady1]

You can see the difference even with a simple computer like our brain

To find a snap out of a film scene a computer has to search thousands of terrabytes of data and it can take a while to get that done using massive amounts of power and hardware

On the other hand we know "instantly" what we're looking at with our 1 watt 20Megabit chip eye/brain hardware

famous film scenes

Think of any image you like, you "instantly" know what film it was from

The data in each film is 4GB and you know which exact film out of hundreds if not thousands of them from a single random image

A computer doesn't have a hope of competing

So computers are a long long way behind at the moment, but they are useful

128bit computers are currently unavailable and the market is more concerned with increasing speed and processing power as opposed to the size of the register

Edited By Ady1 on 10/08/2023 10:47:11

[Ady1Participant @ady1]

If we come at the problem from the opposite direction and assume a human is 1Mhz and we need a minimum register of 1MB

Totally unscientific, but this is the direction industry is taking, processing power

8388608/64 = 131,072 MHz required

Current fastest chips are 6000 MHz

EDIT apparently the brain does about 10Hz, so multiply the chip processing power required by 100 so 13million Mhz required, yikes

Edited By Ady1 on 10/08/2023 11:38:28

[SillyOldDufferModerator @sillyoldduffer]

Posted by Perko7 on 10/08/2023 09:07:27:

The problem I think is in the understanding of the word 'intelligence'.

According to most dictionaries I've seen, intelligence simply means the ability to learn and retain information, in which case the phrase Artificial Intelligence to describe machines which can do this is accurate.

What these machines are not capable of doing is 'reasoning' which is a much more complex activity involving the ability to think and draw conclusions. They are also still machines with no soul and therefore incapable of feelings or emotions which often influence our reasoning.

AI for example could never dream, reason, and generate designs in the way that Einstein, Bell, Tesla, Da Vinci, Edison, Stephenson or any of the other great thinkers and inventors could, so in that respect I think the human race is not under threat.

Perko's post neatly summarises the 'humans are exceptional' point of view. It's believed only we can dream and reason. More, it's having a soul that gives us feelings and emotions that influence our reasoning, presumably in good ways.

Sounds good, but I submit 'human exceptionalism' is unlikely. First problem, at least in Perko's version, is that intelligence requires a soul. The rules of logic forbid concluding souls endow humans with intelligence because there's no proof souls exist, and even if we all agree they do, there is no proof intelligence depends on them.

Second problem is assuming only humans can dream. The purpose of sleep is unknown, and the dreams we have whilst asleep are rarely meaningful. Keeping people awake is very bad for them, but what the brain is doing whilst asleep is doing is a mystery. It may be akin to a computer process called Garbage Collection. When a program runs it uses temporary memory obtained from a pool. This causes the computer's memory to become disorganised, causing performance problems and eventually it crashes. To avoid this, programs periodically stop the normal job and reorganise its memory, reordering data for efficiency and returning any that's no longer required back to the pool. As the pool tends to become fragmented, it too is reorganised periodically. The brain may be similar: a day's thinking causes disorder in the works that has to be corrected. Dream in the sense of innovating is also dubious. In practice, I think, most innovation is problem solving. James Watt started by improving a Newcomen Engine used to pump water out of mines; he didn't envisage that steam would change everything, and end up leading to even bigger developments like internal combustion, gas turbines, and nuclear reactors.

Ady remarked that he would worry about AI when it started writing computer programs. Quite right; programming is a form of problem solving that requires knowledge, skill and reasoning. It's difficult work.

It's happening! ChatGPT does a fairly good job of programming. A potential difference between ChatGPT and a human programmer is that ChatGPT always has to be told what the problem is, whereas a human might invent something new. That's rare though. Most humans don't invent anything: in general we problem solve at the direction of others. The current state of AI is the same: it too can be directed to do many low and mid-range human thinking jobs. AI is good enough to worry Hollywood Scriptwriters and Actors. At the moment I'd say AI can't do Leadership or Management, but it can do Admin.

In short, never say never. I believe intelligence evolved over millions of years due to natural selection. We don't have it because humans are special, rather it's because we happen to be the species that adapted best to Planet Earth.

AI is interesting because electronic computers are much faster than brains, so AI programs could evolve intelligence much faster than humans did. Although very fast, electronics solve problems one step at a time, whilst brains pattern match in parallel and, though slow, have massive memory. The human advantage need not last. Part of the AI revolution is fuelled by ever cheaper electronic memory, and improved ways of making processors work efficiently in parallel.

I don't know of any natural obstacles stopping AI from eventually developing human-level intelligence. The result won't be an artificial person, emulating SillyOldDuffer amusing himself on the forum! More likely unemotional entities will exploit the stock-market, churn out TV soaps, order spare parts, fly aircraft, and do tax returns.

Dave

[Michael GilliganParticipant @michaelgilligan61133]

Psst … whisper it quietly, in case Mortons latches onto the idea

Today’s Independent has an interesting article about the Authors of some articles in Sports Illustrated magazine.

MichaelG.

.

Quoting the closing words of the article:

This is not the first time a news outlet has drawn criticism for posting AI-generated content.

Last year CNET came under fire for using AI to create news articles about financial service topics which the company attributed to “CNET Money Staff”.

Readers could only learn that AI was used to publish the article if they clicked on the author attribution.

After the incident came to light, CNET’s then-editor Connie Guglielmo said over 70 such machine-generated stories were posted on the website.

“The process may not always be easy or pretty, but we’re going to continue embracing it, and any new technology that we believe makes life better,” Mr Guglielmo said.

[Author]

Posts