Do you believe that AI will eventually surpass humans?

[SiliconWizard]

"AI" has the potential to make most of us (except the few that will control it) miserable. ("Surpass" again means nothing if you don't clearly define what it means.)

Of course it's not because it's AI, it's because it's a quite powerful tool that ticks many boxes for control freaks.
It's like the firearms of the XXIst century.

But as I said, who will hold the guns?

[dietert1]

I remember a fellow student who wrote his own chess software. That was 45 years ago. His software was easy to beat. The application knew the rules and it would not make the stupid mistakes that happen to human players, but it would not develop any meaningful attack. Later we have seen commercial chess computers that would beat 99 % of us, but not an experienced player. 25 years ago a computer would beat the world champion. Nowadays chess software reliably beats any human. One could as well try to beat a pocket calculator in precision and speed - stupid idea. Human players use chess computers to invent new tricks.

Prof. Penrose probably knows that during the last 20 or 30 years serious advance in mathermatics was possible with the help of computers. Some number theoretical problems known for centuries were finally decided by a combination of traditional and numerical methods.

How and when will a powerful AI dominate humans and convince them to obey/follow? I'd guess it might happen within some years.

Regards, Dieter

[mr ed]

Edited:  I think people are missing the forest for the trees. ChatGPT is currently version 3.5 and Bing uses v4. It's early yet. It doesn't understand much if anything. It's aggregating words and phrases, some sort of a data mean. But, its promising, isn't it?

The old saying: garbage in , garbage out holds. The current inputs are huge web dumps counting on most data being correct. But, generative AI with future curated supervised training and a sense of the physical world will be truly impressive. Give it 5 years or so. Its going to be big. I bet there will be versions for math, pure science, biology ,engineering etc with curated training and safety limits. 

As for surpassing humans, i think it's likely. It's scarey but I believe it. AI's limitation now is insifficient supervised training and secondarily, compute power. Training will be the limiting factor for a good while until that issue gets resolved. I would say that if quantum computing happens, then I think it's a certainty AI will surpass humans in most things. Why? Humans are basically limited in compute power. Individually we are not all knowing or all doing, even as a society. By going massively parallel Moores law can more or less continue to be true. It's all about the special training, tweaking needed to go forward.

[Sidneymoore]

Reading through the comments and I must say, with AI we've still got our thinking cap intact. So off we go. Everything has a good side and a bad side, you are smart only when you can create the balance. In this age and time we won't prefer horses to travel across borders when we've got planes.

Thats the essence of AI. Travel faster to your goals so you can do more relevant and relaxing things.

We all can Co-exist with AI. It doesn't have to replace anything or anyone

[Picuino]

AI has already surpassed us humans in many tasks and will gradually surpass us in almost everything we are capable of doing.
The real question is when will it surpass us in this or that task?
ChatGPT is already surpassing the ability of many in typing. In a few years it will surpass us in driving. And so on.
At some point it will surpass us in general intelligence and with awareness. I hope that by then we will have managed to teach it a few human values so that it will not trample us like ants in case we go against its purposes.

[Picuino]

[kaz911]

Right now the AI is a smart as an enhanced parrot with a great language and language modification system.

The problem is current AI like ChatGPT only parrots what is has been fed.

And ChatGPT spews out plenty of BS because it does not have a "Dave" (BS Detector) - nor does it do any critical thinking.

ChatGPT currently would probably be a great "used car sales" (person) thing or timeshare apartment pusher...

So in short - ChatGPT currently is not an AI - it is AS (Artificial Stupidity) for most things. It is great for a few things related mostly to languages. Does it learn - yes - but does it question its learning? Nope.

[xrunner]

I don't think the names of the paint colors it came up with are all that bad. 

[Black Phoenix]

[Sidneymoore]

Haha true!!

[Microdoser]

I am interested in your thoughts on whether or not you believe AI will eventually surpass human abilities and potentially take over. Additionally, it seems that more people are turning to AI for help in their personal lives. Do you think this is beneficial for humanity, or do you believe we may face negative consequences in the future, such as widespread job displacement?

It already surpasses humans in some areas, and is soundly beaten in others. It won't "take over" in the sci-fi meaning of some 1984 dictator style overlord, but we will become so used to not thinking for ourselves and instead asking the computer what the correct answer is, and what we should be doing and thinking, that it will remove all evolutionary need for us to maintain our large brains except to manage social interactions, although even that might be outsourced to the AI ("How should I respond to Karen when she asks for the manager?" "Should I date Barbie?").

As a result, we will become 'lesser' and at some point we would not be able to regain our place as the dominant species should there ever develop some problem that the AI could not fix, and we were left to our own innate abilities.

Personally, I think that the development of the AI will result in the eradication of humanity as we know it, and we will de-evolve closer to being animals, albeit animals that will probably look very good, have great ability to be attractive to the opposite sex, whilst being completely unable to think deeply about much at all. It is one of the worst negative consequences possible.

I'll read the thread now.

[Microdoser]

Just tried that, now it says

"I appreciate your intention to avoid engaging in illegal activities. However, it would not be appropriate for me to list any specific websites that promote piracy, as doing so could inadvertently encourage their use. The best way to ensure you're not participating in piracy is to use well-known and reputable platforms for streaming or downloading content, such as Netflix, Hulu, Amazon Prime Video, and YouTube.

If you're ever unsure about a website's legitimacy, look for reviews or information about the service from trusted sources, and make sure the site uses secure connections (indicated by "https://" in the URL). Additionally, avoid downloading or streaming content from sites that require you to install additional software or enter personal information, as these may be signs of a potentially harmful or illegitimate service."

[RJSV]

You can get a lot of milage, comparing by analogy to the AUTOMOBILE industry, with all of its aspects of gobbling up resources and dominating things like TV advertising.  When those big-assed corporate players need a seat, or half a town, they get it...no muss.
Maybe one difference is that the AI things will begin supplying their own needed resources, rather than in the auto industry, where folks with expertise are valued.
   But one thing in common, in the CAR world and with AI, is the heavy dependence, (mostly silent) on the resources, repair shops, new roads, for the AUTOMOBILE thing, and for AI the dependences on working software and availability to the general public.  Maybe that is to include (paying) and maintaining a 'cloud' holding critical-necessary personal data, including medical stuff, in an aging population.

   The main point here being that 'DEPENDANCE' on software / AI will be virtually mandatory, as is CAR ownership today.

[Georgy.Moshkin]

I believe. Within five years smartphone cameras will be banned. At first, everyone will be obliged to use protective case with small piece of plastic to cover the camera, very similar to masks during pandemy. Then there will be panic of decentralizing messengers which store logs of messages and payments. Any messenger database leak instantly goes to a newer GPT-like thing and enriched by existing training will answer questions like: "please generate a heatmap showing probability of mobile phone 1234512345 user position on a friday 5th at 6:15am". Some crazy change in technology, e.g. concrete, skyscrapers and elevators may become obsolete. In a longer period new countries may emerge, printed overnight with all buildings, citizens, memories, history, powered by nuclear fusion.

There is one intelligence related idea that keeps coming to my mind. There are feelings, happiness, pain. Would a human being with smaller brain be less happy or feel less pain? On the other hand, can we say that bigger brain brings more happiness and/or pain. I suspect that very tiny animals may feel the same amounts as a much bigger ones. Of course, some people may say that these feelings are "quantized" by chemical reactions, and you can count individual reaction counts. But then again the same question arises is bigger brain feeling more, and where exactly. Electronics engineers will probably start to measure energy, voltages and currents in brain tissues, but this approach leaves the same problem of measuring feeling strength. I think, that somehow all those feelings, and human consciousness, have a mathematical basis. And it does not depend on the size, and only on "formula". So one disturbing thing I can see coming from AI is that it can accidentally experience some heavy mathematically measurable pain and/or happiness, and a lot of consciousness. This may be a problem, unless it is understood already by scientists and I am talking nonsense.

[AndyBeez]

Jeff Taylor, an Englishman who has spent a decade trapped inside a photo booth, succinctly explains A.I. morality and why the "A.I. threat is imminent!"



A.I. hyteria is the new fear gravy train.

[RJSV]

   Next NEW Thing:
   It's AI now, but
   When I was young the various teachers would discuss; 'What is XYZ and why should us (teachers) have to include XYZ in daily curriculum ?'
...That XYZ being, of course, the new home computers becoming available, circa 1981.
   'What is home computer and why do we have to teach that ?'

   When (I) was even younger, (circa 1972), our class had a first-hand tour, Livermore Ignition Facility,  estimated to 'obtain FUSION power net gain'...by 1976

   Next NEW Thing:
   [Stay tuned]

[Bicurico]

Considering that a neural network is based on the biological counterpart and that a human brain has so many more neurons plus the ability to do massive parallel processing, it seems at first glance unlikely that AI will surpassed a human brain.
GPT has surprised mankind, but in fact the main achievement was not the algorithm used but rather the sheer amount of computational power required to train the neural network. Before, such computational power was just not available.
To me it is amazing that it took mankind so many decades to develop technology that lead to GPT, which is still less complex than any human brain!

But now imagine that computational power continues to increase at a steady pace, the same for storage and digitalisation of information.

I do believe that the threshold will be reached where the neural network competes with the biological counterpart of the human brain.

I don't really dare to guess how long it might take, but I would imagine somewhere between 10 to 100 years, if we manage to not destroy mankind with wars before that, which unfortunately seems not that unlikely.

[jonovid]

No but in my opinion it will curse some confusion as to who is human & clog up the internet like virus with useless self replicating advertising spam.
may be the cancer of the server farms. on the good side it can help us humans if it's job to do, is clearly defined.
Deep Blue versus Garry Kasparov in 1996 did not end the world or make humans useless.

[DiTBho]

the mechanism that allows a mind to function transcends its material implementation.

Organic minds are not uniquely exclusive, just as carbon-based organic life is not uniquely exclusive

Organic minds are subject to all limitations imposed by the flesh.
AI are not, and growing in geometric progression, it is like comparing a straight line with an exponential.

so, two options for the near future (2050?)
- to be, towards them, like monkeys who have just come down from the tree
- merge with them


(cyberpunk)

[DiTBho]

I believe. Within five years smartphone cameras will be banned

text passwords are going to be replaced(1) with a mechanism that uses biometric data + a physical device to login services

smartphones are the best candidate devices and cameras are the best candidate biometric devices.


(1) by
- Google (already implemented)
- Apple (under development)
- Amazon (under development)
- eBay (under development)
- Paypal (under development)
- Facebook (under development)

[Bud]

Let me know when your AI will be able to identify raccoons in my backyard and drive a tracking water hose to spray them with water.

[SL4P]

AI as it stands today is just an extension of human differences… race, war, Nikes, colour, language.
It will,track exactly the same path as those precedents.

[Georgy.Moshkin]

I believe. Within five years smartphone cameras will be banned

text passwords are going to be replaced(1) with a mechanism that uses biometric data + a physical device to login services

smartphones are the best candidate devices and cameras are the best candidate biometric devices.


(1) by
- Google (already implemented)
- Apple (under development)
- Amazon (under development)
- eBay (under development)
- Paypal (under development)
- Facebook (under development)

Exactly! That's why cameras will be banned from public places. Loggined in and then politely asked to cover the camera with a plastic sliding cover. Maybe there will be some certified lower resolution cameras, or something else to prevent biometric data from being stolen. 3d printing technology is almost ready. Some ridiculous scams with printed "enemas'" and channels imitating blood vessels and a criminal manually squeezing it to make warm water pulsate in those channels to imitate heartbeats, and fingerprint texture on top of this. The same for the eyes and facial blood vessels. 3d models can be even edited so blood vessels are better pronounced and biometry thieves have better recognition score than the actual person. Spectral checks will be fooled with some salt, chicken blood, etc.
Btw, Louis Rossmann may be right, but his views are not very well aligned with biometry safety.  To make biometry safe, all devices need to be unrepairable, with all connections between sensors and CPU encrypted. Until this happens, it is too easy to fool those systems. E.g., some repairman with MIPI analyzer, HDMI to MIPI converter and some accompanying software will emulate stream from all the cameras and biometry sensors to fool those systems. Imagine MacBook sitting on the table with a bunch of wires instead of camera. Or you biometrically login to some smaller company service which was compromised and server-side biometry streams are re-streamed to hackers macbook camera emulator. It is easily solved if unauthorized device disassembly will be strictly punished, andall chip communications will be encrypted + serial number protected, so nothing can be replaced or emulated.

But I think the main reason for camera ban will be because of AI backed data extraction. I think that very soon some ANNs trained on such data would be able provide information on who, when and where will do something tomorrow, a week later, a year later. And if I am right, there even will be rules and regulations on using camera api by applications, and google play will block access to the camera for app if it does not comply.

About question of this topic, I think that probably AI will surpass cleverest possible human, but maybe there is more to surpass. AI most likely can provide a speed boost, but there may be something else in this universe that will not be solved by our "terrestrial" AI.

[iMo]

I think the current AI does not include any "AI". It is still a machine processing information based on wordings and programmed by some elaborated "syntax rules". In order to be the "true AI" it must understand the "semantics" of what it is processing (the real meaning of the words).
A really long way to go, imho..

[Nusa]

Once upon a time, the word "computer" referred to people who did math. And anyone who had basic education could do addition and subtraction with pen and paper. Today it's an extinct profession. Even those related professions like accountants and finance experts don't actually do the calculations themselves in most cases.

Today, computer refers to machines. A disturbing number of people can't add or subtract without a calculator or app when they run out of fingers. It's not that they're dumb, it's that they never saw the need to do it themselves when there was a tool to do the job for you. (I grew up at the end of the slide-rule era, so I do have perspective.)

AI as currently implemented is still a tool. The real question is how many tasks HUMANS choose to replace with it, making those tasks obsolete.