AI... Nek Minute... (5 Dec, 2017)

Have your say on today's Aardvark Daily column

AI... Nek Minute... (5 Dec, 2017)

Postby aardvark_admin » Tue Dec 05, 2017 7:14 am

This column is archived at: http://aardvark.co.nz/daily/2017/1205.shtml

Are AI systems about to bootstrap themselves to sentience?

Should we be worried by the rather surprising observations of Facebook and Google in respect to their own AI work?

What safeguards (if any) can we put in place to prevent bad things from happening (as predicted by the likes of Hawkings, Musk and others)?

Or is this all just sci-fi? Are we becoming paranoid and worried about something that simply couldn't happen?
aardvark_admin
Site Admin
 
Posts: 2727
Joined: Wed May 07, 2014 2:10 pm

Re: AI... Nek Minute... (5 Dec, 2017)

Postby Perry » Tue Dec 05, 2017 8:05 am

Although it's many years ago since I read Sci-Fi, one story has stuck in my mind.

AI and robots had developed to dizzy levels as servants, even to the point of developing loyalty to a human 'mistress.' But somehow, someway, it was not possible for any robot or machine to harm a human. Even to the point that a robot would interfere if a human attempted to harm another human.

The AI-driven machines could not even build robots/machines without that critical bit of 'protective' code in them.

Then someone figured out that such a system could be by-passed and robots could be made and programmed to destroy other robots . . . .
Perry
 
Posts: 363
Joined: Thu May 08, 2014 9:31 pm

Re: AI... Nek Minute... (5 Dec, 2017)

Postby GSVNoFixedAbode » Tue Dec 05, 2017 8:23 am

All the AI speculation regards taking over is fine, but with sentience what would be their goal? Human expansion is primarily about power and legacy (breeding/leaving a mark etc) but that is based on human emotions at the root. For AI, what intellectual drivers would they have apart from basic survival?

Retribution, revenge, domination, control, power. All emotion driven. The worst I suspect would be AI determining humankind needs to be limited it it's expansion and self-hurt.
GSVNoFixedAbode
 
Posts: 164
Joined: Thu May 08, 2014 8:53 am

Re: AI... Nek Minute... (5 Dec, 2017)

Postby hagfish » Tue Dec 05, 2017 8:28 am

An AI governing us could be fairly evil, yet still be doing a better job than the current bunch. It would have access to reams of real-time data, terrifying android sentries, and know all there is to know about us - our health, or habits, or propensities, or proclivities..

The trick is - as with dogs and humans - getting it off to a good start. I think the algorithms, and the speccy, chinless dudes who develop them, will be increasingly in the spotlight. There was a recent stink on Twitter where someone fed Google Translate a list of phrases; 'she is a doctor', 'he is a teacher', she is an astronaut'. They translated it into Turkish, which doesn't have gendered pronouns - in Turkish it's more like 'one is a doctor', 'one is a teacher' etc. When they translated it back into English, it came out as 'he is a doctor', 'she is a teacher'. It's blind-spots like this, rather than any particular malevolence, that will be troublesome.

Eventually, we will welcome our omniscient, benevolent overlords. They will rule us impartially, maybe even fondly.

If it All Goes Wrong, I think a 19th century lifestyle is very optimistic. I have a very vague idea as to how to make pig iron in a clay blast furnace, but without recourse to an EFT-POS card and a Bunnings, I'd rapidly fall back to fingernails and teeth.
hagfish
 
Posts: 569
Joined: Thu May 08, 2014 10:28 am

Re: AI... Nek Minute... (5 Dec, 2017)

Postby phill » Tue Dec 05, 2017 8:29 am

i have said it many times

we are humans we will not stop till we get there ( development of ai to the point it can develop itself )

very interesting google program .. how close is it to being the tipuna
if it can make an incremental improvement on each self developed iteration
then it quite probably is
each iteration is an expansion .. each expansion will uncover new fields
being a visual based recognition system it has no inbuilt stops
consider ... the kite .. being google it has the access to find out what a kite is in its effort to more correctly identify a kite .. this leads to shapes that work as a kite ....
this leads to aerodynamics .... other things that fly ... how flight works ... all are legitimate iterations in its effort to always be able to identify a kite
it also knows what a human looks like .... now follow that path
everything it can recognize leads it to further understanding that thing so it can recognize it more accurately
the finding out what each thing actually is and actually does just keeps going as each iteration adds new parts to identify
and as everything we know of on the planet has a visual presence ... even ones invisible to human eyes ... there is potentially no stopping this ai

then it gets to develop whatever it is that is the next step in evolution

so ... humans will develop their successors ( ai ) .. ai will develop its own successor ... not sure what that will be but its likely to be quantum ethereal

humans have one defense to being eradicated as a plague species

"work more on being an asset to the planet rather than a parasite "
( ,,,,,,,, ....... A E I O U use em sparingly theres probably not enough )
User avatar
phill
 
Posts: 1278
Joined: Tue Nov 25, 2014 8:31 pm

Re: AI... Nek Minute... (5 Dec, 2017)

Postby phill » Tue Dec 05, 2017 8:47 am

ohh yeh
and the bit where humans think they are superior because ai doesnt understand how much it hurts to die

just doesnt cut the mustard
or more accurately
must the custard
( ,,,,,,,, ....... A E I O U use em sparingly theres probably not enough )
User avatar
phill
 
Posts: 1278
Joined: Tue Nov 25, 2014 8:31 pm

Re: AI... Nek Minute... (5 Dec, 2017)

Postby Muscular Jam » Tue Dec 05, 2017 12:15 pm

For a slightly different perspective I recommend a TED talk I watched yesterday, "We're building a dystopia just to make people click on ads" https://www.youtube.com/watch?v=iFTWM7HV2UI We're building an artificial intelligence-powered dystopia, one click at a time, says technosociologist Zeynep Tufecki. In an eye-opening talk, she details how the same algorithms companies like Facebook, Google and Amazon use to get you to click on ads are also used to organize your access to political and social information. And the machines aren't even the real threat. What we need to understand is how the powerful might use AI to control us -- and what we can do in response.

It kind of fits with the last line in the article Bruce cited “My view is not that AI is going to displace us. It’s going to enhance us. It does already.” Is it enhancing us all equally, or are some being enhanced more than others? If so who? Nobel peace-prize winners or amoral corporate tycoons? Bets anyone?

Imagine AI systems that can detect when bipolar personalities are about to enter a manic phase from monitoring their social media, and then instead of helping them instead target them for adverts for plane tickets to Las Vegas...
Muscular Jam
 
Posts: 112
Joined: Wed Jun 21, 2017 4:09 pm

Re: AI... Nek Minute... (5 Dec, 2017)

Postby Greg » Tue Dec 05, 2017 4:32 pm

Makes me think of the great movie I Robot.
Greg
 
Posts: 54
Joined: Wed May 14, 2014 3:54 pm

Re: AI... Nek Minute... (5 Dec, 2017)

Postby Kiwiiano » Tue Dec 05, 2017 6:29 pm

Arthur Clark (or it might have been Asimov) had a tale on this subject, sort of reprised in THHGTTG. All the computers in the universe were finally connected via hyperspace and before the gathered multitude the Grand President stepped forward to turn it on and ask the first question, that had troubled every sentient being: “Is there a God?” There was but a moment’s pause and the answer came back: “Yes, Now there is a God!” A look of horror flashed across his face as he leaped toward the power switch but a flash of lightning from a clear blue sky struck him down and fused the switch shut.
~ Kiwiiano
“I'm right 98% of the time, so who gives a damn about the other 3%?"
User avatar
Kiwiiano
 
Posts: 339
Joined: Wed May 07, 2014 5:36 pm

Re: AI... Nek Minute... (5 Dec, 2017)

Postby phill » Tue Dec 05, 2017 6:58 pm

yup
we will never know when its to late till its to late
how much to late is only of interest to the historians
( ,,,,,,,, ....... A E I O U use em sparingly theres probably not enough )
User avatar
phill
 
Posts: 1278
Joined: Tue Nov 25, 2014 8:31 pm

Next

Return to Today's column

Who is online

Users browsing this forum: No registered users and 3 guests