Register a SA Forums Account here!
JOINING THE SA FORUMS WILL REMOVE THIS BIG AD, THE ANNOYING UNDERLINED ADS, AND STUPID INTERSTITIAL ADS!!!

You can: log in, read the tech support FAQ, or request your lost password. This dumb message (and those ads) will appear on every screen until you register! Get rid of this crap by registering your own SA Forums Account and joining roughly 150,000 Goons, for the one-time price of $9.95! We charge money because it costs us money per month for bills, and since we don't believe in showing ads to our users, we try to make the money back through forum registrations.
 
  • Post
  • Reply
klafbang
Nov 18, 2009
Clapping Larry

Fallows posted:

I think thats where bitcoin got stuck at too

Buy my new ICO. It's like bitcoins, but instead of inefficient proof of work or anti-vegetarian proof of steak, we mine butts using proof of concept. (Whitepaper coming soon)

Adbot
ADBOT LOVES YOU

SolTerrasa
Sep 2, 2011

Risc1911 posted:


I give you another example. The https://en.wikipedia.org/wiki/Halting_problem Similar to this, ML and AI will never be able to "know that it does not know". This is getting a little philosophical but i think you understand what i mean.

I don't think you've explained this very well - right now I think you're saying that AI can't explain when it doesn't know an answer, but that is extremely common. For instance, if you feed an image of static to Google's ImageNet API, it'll return you some json that you interpret as "there's a 0.001% chance this image is a koala, and a 0.0005% chance this image is a car, and everything else is way lower than that". No halting problem violations there.

cubicle gangster
Jun 26, 2005

magda, make the tea

Risc1911 posted:

Even tasks mastered almost reflexively by infants are far beyond the capacities of contemporary AI.

I use ai based tools in work and it's having a huge impact. One of which is this - https://research.nvidia.com/publication/interactive-reconstruction-monte-carlo-image-sequences-using-recurrent-denoising
There are others too.
I dont think an infant is very good at reducing noise in a generated image, or smoothing out the results of a fluid simulation. When used to refine and clear up early stage simulations (no matter what is being simulated), AI and ML is very much 'in use' by professionals and has been for a couple of years.

but it cant put the round peg in the right hole so it's useless, right?
Did anyone ever claim thats what AI was even for?

cubicle gangster fucked around with this message at 19:15 on Nov 13, 2018

BigBadSteve
Apr 29, 2009

PhazonLink posted:

I hope Butt'ers remember that Mythbusters did a free energy episode and they then spend 1000 K$ of poo poo fiat money to buy wires and get free energy from the Earth's magnetic field.
(or they set their house on fire making that cool propane wheel thing or they try perpetual motion machines)

A stable store of power value.

Feinne
Oct 9, 2007

When you fall, get right back up again.

cubicle gangster posted:

I use ai based tools in work and it's having a huge impact. One of which is this - https://research.nvidia.com/publication/interactive-reconstruction-monte-carlo-image-sequences-using-recurrent-denoising
There are others too.
I dont think an infant is very good at reducing noise in a generated image, or smoothing out the results of a fluid simulation. When used to refine and clear up early stage simulations (no matter what is being simulated), AI and ML is very much 'in use' by professionals and has been for a couple of years.

but it cant put the round peg in the right hole so it's useless, right?
Did anyone ever claim thats what AI was even for?

I think they're speaking more generally to the broad AI claims rather than the many specific applications at which ML excels. They're saying that there are things an infant does trivially that we don't even have any idea how to start making an AI do, not saying that there is nothing an AI can do that an infant cannot do better. At least I hope that's what they're saying because the latter is an obviously indefensible claim.

naem
May 29, 2011

Risc1911 posted:

I don't think we have reached "Peak AI" yet. It's all bollocks any way.

One can certainly imagine how, in principle, systems that can detect patterns with massive data processing might find hitherto unknown ways of constructing theories that surpass those within the reach of human intelligence. And that could have all sorts of effects. But among the concerns we face, this doesn’t seem to me to rank high. Even tasks mastered almost reflexively by infants are far beyond the capacities of contemporary AI.


Quote-Unquote posted:

It was way shorter than that, when Microsoft released Tay on to Twitter and it turned into a Nazi calling for the extermination of Jews within about an hour.

My Linux Rig
Mar 27, 2010
Probation
Can't post for 6 years!

you know the funny thing I heard is that when tay was deployed in china, it turned out to be quite polite

Lemming
Apr 21, 2008

My Linux Rig posted:

you know the funny thing I heard is that when tay was deployed in china, it turned out to be quite polite

I think it was taken down and rereleased after they implemented a Nazi Inhibitor

Papa Was A Video Toaster
Jan 9, 2011





Lemming posted:

I think it was taken down and rereleased after they implemented a Nazi Inhibitor

Isn't their Twitter 1000% not anonymous?

Tacky-Ass Rococco
Sep 7, 2010

by R. Guyovich

Lemming posted:

I think it was taken down and rereleased after they implemented a Nazi Inhibitor

Can't believe Asimov forgot to add "don't be a Nazi" to the Laws of Robotics.

My Linux Rig
Mar 27, 2010
Probation
Can't post for 6 years!

Lemming posted:

I think it was taken down and rereleased after they implemented a Nazi Inhibitor

it was? I didn't hear anything about that

Serjeant Buzfuz
Dec 5, 2009

Turns out 90% of the people on Twitter are just chat bots.

LethalGeek
Nov 4, 2009

ghosTTy posted:

The power isn't wasted

Seriously: gently caress you

Lemming
Apr 21, 2008

My Linux Rig posted:

it was? I didn't hear anything about that

That's because chat bots are boring if they have Nazi Inhibitors

Fur20
Nov 14, 2007

すご▞い!
君は働か░い
フ▙▓ズなんだね!

LethalGeek posted:

Seriously: gently caress you

coiners should be chained to a wall just out of reach of food

when they ask why it's being wasted as it rots and they starve to death just say "it isn't being wasted"

EorayMel
May 30, 2015

WE GET IT. YOU LOVE GUN JESUS. Toujours des fusils Bullpup Français.

The White Dragon posted:

coiners should be chained to a wall just out of reach of food

when they ask why it's being wasted as it rots and they starve to death just say "it isn't being wasted"

Would this be a blockchain?

NtotheTC
Dec 31, 2007



My god they've already perfected an artifical Yodel delivery driver.

We need basic income NOW

Nessus
Dec 22, 2003

After a Speaker vote, you may be entitled to a valuable coupon or voucher!



Feinne posted:

I think they're speaking more generally to the broad AI claims rather than the many specific applications at which ML excels. They're saying that there are things an infant does trivially that we don't even have any idea how to start making an AI do, not saying that there is nothing an AI can do that an infant cannot do better. At least I hope that's what they're saying because the latter is an obviously indefensible claim.
Yeah I think "AI" in the sense of "automating a lot of little things or finding novel uses for computer power" vs. "AI" in the sense of "the impending ultra-gangster computer god that will revive Kurzweil's dad for him"

feelix
Nov 27, 2016
THE ONLY EXERCISE I AM UNFAMILIAR WITH IS EXERCISING MY ABILITY TO MAKE A POST PEOPLE WANT TO READ
Sentience is a glitch that arose due to biological evolution being an extremely bad optimization algorithm. Sentience does not serve to maximize the objective function of "how many copies of your genes did you pass on", and in fact probably detracts from it. The algorithms we use to optimize neural networks are much better than that, so sentience will not arise.

Dr. Fraiser Chain
May 18, 2004

Redlining my shit posting machine


feelix posted:

Sentience is a glitch that arose due to biological evolution being an extremely bad optimization algorithm. Sentience does not serve to maximize the objective function of "how many copies of your genes did you pass on", and in fact probably detracts from it. The algorithms we use to optimize neural networks are much better than that, so sentience will not arise.

Your first two sentences rely on claims you and others haven't proven. Though really every sentence has claims that are not necessarily true.

feelix
Nov 27, 2016
THE ONLY EXERCISE I AM UNFAMILIAR WITH IS EXERCISING MY ABILITY TO MAKE A POST PEOPLE WANT TO READ

Goodpancakes posted:

Your first two sentences rely on claims you and others haven't proven. Though really every sentence has claims that are not necessarily true.

drat I really thought I could provide a rigorous logical proof of my hot take on unprovable philosophy regarding undefinable terms

feelix
Nov 27, 2016
THE ONLY EXERCISE I AM UNFAMILIAR WITH IS EXERCISING MY ABILITY TO MAKE A POST PEOPLE WANT TO READ
Also who cares if an ANN achieves "sentience" whatever that means. It still won't ever do anything except what it was trained to do. Like OK, it's contemplating its own existence or whatever but the only thing it's capable of actually outputting is "13% dog 9% cat 5% goatse"

Papa Was A Video Toaster
Jan 9, 2011





Please, let's just laugh at number go down.

Khorne
May 1, 2002
Calling ML/NN stuff AI really grinds my gears. Statistics, linear algebra, and calculus are AI guys!! For real!! I'm the popular opinion in this decade!

Is your dryer not AI because it stops heating when the moisture drops? There is a degree of artificial intelligence to polling its sensor and determining what it should do based on its state.

Is your computer not AI because it reacts to your inputs and does what you tell it? Philosophically, is using repetition and statistics as programmed any different from a state machine as programmed? In science we've found statistical approaches to provide comparable outcomes to non-statistical methods.

How about just plain old statistics predicting outcomes given inputs? Is that AI? Is that equation you're using not thinking for you? A decent subset of ML is really just computational statistics that would be beyond tedious to do by hand.

What makes pattern classification or optimization algorithms AI and everything else not?

To be clear, I'd consider all of those things I listed "AI" to some extent. It's just weird seeing anything remotely related to ml be called AI when you don't see every state machine or thing traditionally used in AI called AI. In fact, only ML gets called AI now and other AI gets shunned with "but this ML approach is AI! New! Exciting!" AI is a broad word and it being co-opted by a current trend isn't cool. ML/etc certainly have AI applications. They're very useful and powerful techniques to solve the problems they apply to. But they aren't sentient, and how they are often portrayed is ridiculous. "it's learning!!", it's learning about as much as an n-body system is showing you how the world works. That is to say, it's routinely doing the exact algorithm you programmed it to do and giving you the output you would have gotten if you had infinite time to do it by hand with identical inputs.

Khorne fucked around with this message at 00:18 on Nov 14, 2018

feelix
Nov 27, 2016
THE ONLY EXERCISE I AM UNFAMILIAR WITH IS EXERCISING MY ABILITY TO MAKE A POST PEOPLE WANT TO READ

Khorne posted:

What makes pattern classification or optimization algorithms AI and everything else not?

Because they are difficult enough to understand for investors and grant-awarding agencies that calling them a sexy name will significantly increase the probability and amount of money one can get

Lemming
Apr 21, 2008

https://i.imgur.com/DWEFgPa.mp4

wid
Sep 7, 2005
Living in paradise (only bombed once)

Feinne posted:

I think they're speaking more generally to the broad AI claims rather than the many specific applications at which ML excels. They're saying that there are things an infant does trivially that we don't even have any idea how to start making an AI do, not saying that there is nothing an AI can do that an infant cannot do better. At least I hope that's what they're saying because the latter is an obviously indefensible claim.

And the flaw with that kind of thinking is the fact that there is no need to create an AI that could simulate an infant. Like, why? Infants are unnecessary to any process we need AIs to do so no one bothers to emulate them. If there were, could we make AIs that act like infants? Sure (maybe not right now but some day). AIs and ML are very good at specific tasks as long as the parameters are clearly defined. Anyone who thinks any AI developers are trying to create sentient beings is being a stupid scifi baby. It's the same as thinking any day now the Tesla automated driving assistant will find the cure to cancer. I'm not saying sentience will never appear out of AI and ML since we don't even know how sentience came about or even how to properly define it that would encompass all living beings. Could we accidentally create a sentient being? Sure, it's possible. But no one is aiming for it and no one is designing machines to be that. Stop thinking Westworld and Terminator as what AIs are supposed to be.

IPCRESS
May 27, 2012
code:
while 1
   vomit()
   cry()
   shitpants()

AI Infant Status: Programmed. Honestly I just took the 'drunk' subroutine and renamed it.

Now the real question is: Have buttcoin exchanges started offering CDOs yet?

Gobbeldygook
May 13, 2009
Hates Native American people and tries to justify their genocides.

Put this racist on ignore immediately!

IPCRESS posted:

Now the real question is: Have buttcoin exchanges started offering CDOs yet?
When the Bitcoin Savings & Trust ponzi scheme collapsed, it was quickly revealed that a bunch of other bitcoin funds had just put all of their investors money into it.

Tacky-Ass Rococco
Sep 7, 2010

by R. Guyovich

IPCRESS posted:

Now the real question is: Have buttcoin exchanges started offering CDOs yet?

CDS's, you mean? There has to be crypto-denominated debt first for there to be crypto default swaps. But, now that you mention it, crypto bonds sound like the next big scam. Like a classier ponzi scheme.

Has anyone said BitBonds yet?

Lote
Aug 5, 2001

Place your bets
Hey guys. Take a look at the Tether price. It’s down to $1USD = 0.965 tether. Volume is super low on the charts which is very bad and indicative that we may see some fireworks shortly. You may want to get your popcorn ready.

jimmyjams
Jan 10, 2001


King Kong of Megadongs
Gobblin' them mega schlongs
Makin' sure they mega long
Stroke' 'em if they mega strong
is there a tether thats backed by tether

Lote
Aug 5, 2001

Place your bets
Yes. It’s called Bitcoin

Rectal Death Adept
Jun 20, 2018

by Fluffdaddy

We are safe until there is some kind of breakthrough in battery technology. Suck my dick robots.

feelix
Nov 27, 2016
THE ONLY EXERCISE I AM UNFAMILIAR WITH IS EXERCISING MY ABILITY TO MAKE A POST PEOPLE WANT TO READ

Rectal Death Adept posted:

We are safe until there is some kind of breakthrough in battery technology. Suck my dick robots.

My dude have you seen the Matrix ?

Tacky-Ass Rococco
Sep 7, 2010

by R. Guyovich
Gotta say, it's fuckin' sweet that we're gonna get to see autonomous killbots in our lifetimes.

DerekSmartymans
Feb 14, 2005

The
Copacetic
Ascetic

Rectal Death Adept posted:

We are safe until there is some kind of breakthrough in battery technology. Suck my dick robots.

I’m sure firehoses would defend us.

Ofc if the robots are sucking dicks I vote we let them be in charge for a little while.

Viktor
Nov 12, 2005

Internet historian did a video on the chat bot uprising https://youtu.be/HsLup7yy-6I

I think we got some time! Zo chatbot keeps sending stupid gifs

My Linux Rig
Mar 27, 2010
Probation
Can't post for 6 years!

feelix posted:

Sentience is a glitch that arose due to biological evolution being an extremely bad optimization algorithm. Sentience does not serve to maximize the objective function of "how many copies of your genes did you pass on", and in fact probably detracts from it. The algorithms we use to optimize neural networks are much better than that, so sentience will not arise.

you say that while there’s 7.5 billion people in the world

If anything, sentience works too well

Adbot
ADBOT LOVES YOU

Facebook Aunt
Oct 4, 2008

wiggle wiggle




My Linux Rig posted:

you say that while there’s 7.5 billion people in the world

If anything, sentience works too well

How many bacteria?

  • 1
  • 2
  • 3
  • 4
  • 5
  • Post
  • Reply