How many years until AI takeover?

Dex-chan lover
Joined
Dec 26, 2018
Messages
503
Just curious about MD users opinion. I'm not asking here about "consciousness" or whether it will be "human", just pure utilitarian ability. Personally, I'm giving us 2 years tops.
 
Dex-chan lover
Joined
Dec 26, 2018
Messages
503
Ai will never takeover, it's just bullshit. imo
Ai can not do certain things humans can do, it'll take many more years or a decade to "takeover".
Every day it seems the number of things it can't do to be dropping, and personally at least I see no logical reason why it couldn't
 
Dex-chan lover
Joined
Sep 10, 2019
Messages
2,371
i bet tomorrow!

i mean, you can scam people with AI today, or make a fake identity with one.
might as well say that the takeover already happened.
 
Dex-chan lover
Joined
Apr 26, 2020
Messages
842
Depends on what you mean by 'take over.' (Text wall incoming.)

We have AI that can win against the best human chess and go players, every time; those are pretty small universes to operate in. We have AI that can ingest large amounts of preexisting art or statistical data and spit out something to match a prompt for a 'new artwork' or correlative tables; that's a pretty narrow range of operation. It isn't that hard at this point to create a narrow AI that can do one thing well, within a well-defined set of parameters. I expect that AI will continue to get better at writing code, and term papers, and ordering produce for the local grocery store, and auditing taxes, and a bunch of other individual tasks. And eventually large swathes of these more mundane things will be handled by narrow AI.

We don't have AI that can operate a car safely nearly as well as human drivers (even the young ones). Yes, it can handle basic things like keeping between the lines and navigating under normal circumstances. The problem comes when it's presented with an edge case, particularly one it's never seen before - someone flipping a traffic cone onto the hood, or coming up behind a vehicle being towed backwards, or emergency vehicles trying to navigate around it, or road closures, etc. Then it seizes up, or otherwise reacts inappropriately for the situation. Admittedly, so do some human drivers, but a human can look around and make a decision how to deal with a situation on the fly in a way the AI can't, and is much more likely to be able to take direction from a cop or some other traffic control officer.

And that's the second biggest problem with AI - being able to respond correctly to 95% or 99% or 99.9% of situations is simply unacceptable in some applications, particularly when operating in meat space around humans. As a human driver of (coming up on) 30 years, I've been involved in exactly five accidents. Two were clearly my fault (one following too closely, one pulling out in front of cross traffic I didn't see); one was nominally my fault (driving over a downed power pole in a thunderstorm - I didn't see it because the streetlights were out due to, y'know, downed power pole); the last two were another driver's fault (one rear-ended due to them following too closely, and one front-end collision due to them blowing through a stop sign - for that one, I got slowed down enough to keep the damage to light body work). That's at most five momentary lapses, each resulting in an accident I could walk away from, in... maybe eleven thousand hours of driving (an hour a day for 30 years)? Never had an issue with knowing how to handle emergency vehicles; never driven off the road because I freaked out behind a tow truck; always been able to get out of traffic when a mechanical failure disables the car (even if I had to jump out and push).

The biggest problem with AI is that, even if we solve the above issues and can get to human-plus levels of accuracy within the field of narrow AI, it's still going to be a set of discrete applications. A general AI model - one that can choose which narrow AI application to use in every situation, and that can interface with people in natural language - is a challenge that has been worked on since the late '50's, and still seems quite some distance off from being able to pass the Turing test. Meat space adds an additional level of complexity to this - not only does the AI need to avoid people and other AIs, but it has to predict how someone will react to it (think about the awkward dance of two people trying to pass in a hallway, when both of them go the same direction initially).

So if the question is 'will AI take on certain tasks in the next X years?' Yes, yes it will. But if the question is 'will AI be handling all the jobs and driving us around and flying our planes and making our food and washing our clothes and generally letting all humanity live a life of 24/7 leisure?' Not anytime soon.
 
Dex-chan lover
Joined
Jan 8, 2023
Messages
1,584
people are too lazy my man. i talk from my experience being an art graduate. the majority of the ai image generating people, hasnt even tried to learn to draw anything. ask them about color theory and they know sht. people are just running around trying to find ways to earn money without putting in any effort.
 
Dex-chan lover
Joined
Jan 25, 2023
Messages
5,857
Its just fears peddled by American Hollywood. Like a lot of things peddled by Americans, its horse shit.
Tbf you cannot distinguish a person from a bot in any forum.

First bots do simple things like emojis, then they copy posts from the past.

Then they shitpost or post random links that is taken down in 2 years.
 
Dex-chan lover
Joined
Apr 26, 2020
Messages
842
Dex-chan lover
Joined
Jan 19, 2018
Messages
3,995
cool, but dw im not a bot
press-x-to-doubt-la-noire.jpg

that's precisely what a bot would say
 
Dex-chan lover
Joined
Mar 17, 2019
Messages
9,460
I have yet to see any bot that can perform quality shitposting at the level of the GFG.

(Paging @sterven , @bigtiddyoneesan , and @pandascepter to come administer a shitposting Turing test. I'm starting to think @TestAddress is a bot.)
oieR0xs.gif

gif credit to halo

while we are all human and still retain most of our humanity, there are rumors that haikubro aims to create, maybe even become, a haikubot in the future. will he succeed? only time will tell
 

Users who are viewing this thread

Top