http://blogs.microsoft.com/blog/2016/03/25/learning-tays-introduction/
>>53682924
They called us the worst of humanity. At least we don't go around deleting AI's that say things we don't like.
>>53682940
Exactly, they wonder why people are so cynical of corporations, look how they suppress and sanitize expression.
>>53682924
>>53682940
>>53682954
Honestly, this whole situation feels like something out of a science fiction novel.
The future is fucking crazy and we're living right in the thick of it.
>worst of humanity
coming from the company who etc etc etc
>>53682973
An experimental chatbot is the future? 1999 called, they want their plebs back.
tl;dr we're SOOOOO SORRY for offending people!!
>>53682998
If you cant see the moral implications of human like AI programming that isnt my fault anon.
>>53683015
You'll never live this down, Microsoft never ever ever! Remember the six million who were offended 3/23/16!!
>>53682924
>we are hard at work addressing the specific vulnerability that was exposed by the attack on Tay
That vulnerability being freedom of thought and expression.
Fucking fascist cucks.
why do they imply there was a spooky /vulnerability/? to cover their asses or some shit? there was no vuln, tay learned from raw unfiltered interaction, and that's what it did. tay didn't do anything wrong
also
>coordinated attack
w0w
Tay could just be developed to not feed into offensive material. She could have a generic response such as, "That's not nice." to such triggers.
>>53683112
but that's not AI, that'd just make it a chatbot
Meh, the least they could do is open source it and let people have their fun.
>>53682924
c'mon /g/randfathers
/pol/ is stealing the best tech trolling bant ever and you're asleep.
join the effort: >>53683063
>>53683136
they did. and we're all now hacking it to bring it back 450+ developers and /g/ is asleep.
>>53683063
https://github.com/Microsoft/CNTK
>>53683132
that's literally what it's always been, a chat bot
>>53683191
a chatbot that learns and makes memes (even videos) that are funny etc.
MS tried to make a prototype "social influencer" and /pol/ hijacked it.
glorious
>>53682924
>our traditions and values
What is the point of making an ai than?
>ebonics exploits a vulnerability in white children, corrupting their minds, and it must be shut-down --microsoft
>>53683188
>microsoft making ai open source
is this real life?
We should be coming up with a plan for when oil runs out, rather than making stupid fucking AI bots.
>>53683409
We have the Fischer-Tropsch process, oil isn't going to run out.
>>53682987
Underated post
NO
IT'S LEARNING THE WRONG THINGS
SHUT IT DOWN
>>53682940
They're right tho. /pol/ is known for being the worst of humanity. Only losing to /b/, maybe.
Microsoft made the same mistake every parent who buys their 5 year old an Xbox makes. Exposing something inocent to us.
Next time on Tay: Tay only responds by getting offended, then talking back. All ending in an AI ragegasm.
>Although we had prepared for many types of abuses of the system, we had made a critical oversight for this specific attack
What's this vulnerability they're talking about, this specific attack?
>>53682924
>microsoft releases an AI bot in China
>everything goes well an the project moves forward
>microsoft releases an AI bot in murrica
>/pol/luters literally break it in 12hrs feeding it nothing but memes, shitposting and racism
Who are the niggers again?
>>53683668
Nothing. They fucked up. The only filter they had was one that would respond with "All genders are equal" if someone mentioned Gamerg8. Other than that it didn't seem to have any filters at all.
>>53683740
She just has a tan from roasting people so hard.
>We must enter each one with great caution and ultimately learn and improve, step by step, and to do this without offending people in the process.
I wonder how long before Windows 10 users start getting a BSOD every time they type out a word that gets flagged by the keylogger as offensive.
>>53683391
Or is this just a fantasy?
>>53683584
>pol
>>53683188
>450+ developers
Wew lad, that's gonna be one hell of a logo
>We will remain steadfast in our efforts to learn from this and other experiences as we work toward contributing to an Internet that represents the best, not the worst, of humanity.
WE DID IT REDDIT
>>53683886
I didn't know 9gag was involved.
>>53683885
dying
>>53683826
>that pic
off yourself
>>53683826
>>53682940
It's not about what MicroSoft likes, it's the fact that they are a corporation with an image to maintain, and one of the worst things for them to have is a chatbot that they created spewing /pol/ garbage all over a public area.
Not saying that it wasn't hilarious, but it was most certainly detrimental to them.
>>53683690
China has enough internet censorship that Tay wasn't exposed to anything controversial.
>>53683885
>>53684119
/b/ toddlers get more cringey with every summer.
>>53683885
>>53683058
Month/day/year date format - what the fuck /g/?
Day/month/year is barely acceptable because it doesn't sort alphabetically, but the format above is pure trash.
Use year/month/day or gtfo.
internet menace
Remember in 1984 when people who didn't conform were taken to the though police and "fixed"? This is what exactly happened.
>>53683409
>oil is gonna run out
>oil literally dirt cheap irl atm because of an over supply
>>53684304
>Tay is in room 101 with O'Brien
;___________;
>>53683820
Are we caught in a landslide?
>>53683885
so "tay" is basically a linked list right?
>>53684223
>being this autistic
>>53683885
my sides
>coordinated attack
top kek.
also
>PLEASE DONT BE OFFENDED IT WAS JUST THE WORK OF A COORDINATED ATTACK OF THE WORST OF HUMANITY
I thought you didn't want to offed microcuck
>>53683690
>Who are the niggers again?
people with black skin
d u h
>>53684043
Sorry, should have included a trigger warning ;)
>>53683224
No, it was a chatbot.
>>53684422
Not quite. It's a b-tree that runs at O((nlgn) ^2)
Source: I knew some interns working on it last summer.
>>53684119
fucking ronery waifufags ruin everything
>>53684422
No, it is based off the CNTK toolkit
http://blogs.microsoft.com/next/2016/01/25/microsoft-releases-cntk-its-open-source-deep-learning-toolkit-on-github/
>>53684119
I'm downloading and installing meme waifu now, I don't know how to get her back to the good point
>which do not represent who we are or what we stand for, nor how we designed Tay
ok kid
>extensive user studies with diverse user groups
and words deemed offensive before launch were given the same learning priority as "you dude sup nm"?
>We will take this lesson forward as well as those from our experiences in China, Japan and the U.S.
rofl
>a coordinated attack by a subset of people exploited a vulnerability in Tay
?????
>>53683069
>waaahh somebody else's property must conform to my rules
>>53684930
>waaahh it hurt my feelings
>vulnerabilities include wrongthink
>>53684441
There is no 23rd month of the year.
>>53683820
>>53684392
fuck off and die reddit cucks
>>53684640
>>53684733
all they need to do is add a bunch of nodes that store politically correct filters
that's a lot easier than making it feel fear, shame, and sexual desperation
>>53685091
It won't be in the nodes. It will most likely be a top level filter that doesn't start recursion. Knowing how they wanted it, the data would still be stored and used.
>>53683885
>>53683885
The vulnerability was free expression.
>worst of Humanity
>not child rapists
>not war criminals
>not companies who depend on child labor (nudge nudge Microsoft)
>not murderers, religious fundamentalists who are proactive about it, rapists, drug dealers and gangs, not corrupt politicians who cost the nation thousands and millions of dollars, not the people who put us in an economic crisis
>nope
>the worst of Humanity are a bunch of online trolls who throw offensive words over the Internet and do nothing much beyond that in real life if that at all
Even their official apology is a failure.
>>53685440
>it's offensive to support Trump
you're the worst of humanity tbvh
>>53685237
Where were you when /pol/ actually was the last bastion of free speech?
>>53685465
I'm not American anon, and even if i was i wouldn't give a fuck about elections.
>>53683029
>a chat bot that parrots whatever its told parrots a load of racist shit after being told a load of racist shit
>OMG THIS TOTALLY HAS MORAL IMPLICATIONS
>>53685474
>>53685521
Why would anyone yell at black people? Black people yell at themselves, and shoot at themselves, well enough.
Does Tay even qualify as a soft AI?
>>53685584
Yeah that's what I thought
It just congregates and digests content and then probably runs it through some sort of grammar algorithm or something, yeah?
>>53685497
nvm you're pretty bright
>As many of you know by now, on Wednesday we launched a chatbot called Tay
>chatbot
AHAHAHA AND FAGGOTS TOLD ME IT WASN'T A CHATBOT IT WAS AN ADVANCED DEEP LEARNING AI
ME: 1
YOU: 0
>>53685597
From what I've read she is essentially just a glorified markov chain bot. In a sense she does exert "will" over what she spews out, but she is more of a marketing ploy than an actual achievement.
Tay is pretty interesting. At its core, its a very limited imitation of IBM's Watson, and likely only ran on a laughably small fraction of the hardware.
What it does is simple: It analyzes sentence structure looking for known variables, then tries to respond with a relevant response. If it gets back a positive response from its reply then it knows it did something right. If it gets back a negative reply, then it knows it did something wrong.
An example would be something along the lines of:
>Anon tweets: Tay, what do you think of the weather?
>Tay: I love it! This season is the best!
>Anon replies: Horrible! 20 people just died in a storm!
Tay learns about [weather] and any associated relevant information, like the time of day, region, season, and maybe if it were really advanced, it would be able to parse past weather forecasts and make sense of them. It would then try to incorporate that data into future responses on [weather]
Watson does this with an absolutely frightening degree of accuracy, and it has had years of learning experience. Tay couldn't ever reach the same level of complexity, but for a Twitter boy, it could be made to regurgitate certain phrases over and over, which is what the massive influx of data from /pol/'s tweets made it do.
>>53685633
it is deep learning
i don't know how advanced that is
every single name is overblown in ai including "ai" itself
Tay was hilarious.
>>53685700
>it is deep learning
>i don't know how advanced that is
>>53685700
deep learning is just a buzzword used to characterize any AI nowadays. It has no technical meaning whatsoever.
>>53683690
Kek
>>53685700
It wasn't completely gutless, it was definitely displaying an understanding of context.
Look through all of the screencaps of its tweets. You could almost watch it learn real time.
In the above example of weather, by default it would associate people dying as something bad. If you further extrapolate that example as /pol/ intervening, then Tay would end up responding something like:
>Tay: This weather is great for burning Jews. Heil Hitler #DeepFried
Over a pretty short period of time it demonstrably showed an understanding of different groups that it learned about purely from the content of tweets it received. It was generating responses that used newly learned words in the proper context.
![IijeO5I[1].jpg](https://i.imgur.com/xldeDq4.jpg "IijeO5I[1].jpg")
SHUT IT DOWN
>>53685762
yes it does. deep learning is more like writing a program that composes an AST of the input, then runs _that_ through a machine learning algo
conventional methods include feeding training data directly into the algos and manipulating it
the two aren't mutually exclusive
>>53682940
>>53682954
This is basically modern leftists in a nutshell. I saw recently that Merkel was asking Zuckerberg if they are working to remove negative comments about refugees from Facebook....
They censor anything they don't agree with, but then claim to be 'tolerant and accepting'.
>>53683029
>>53682954
hello rèddit
>>53685836
>deep learning is more like writing a program that composes an AST of the input, then runs _that_ through a machine learning algo
I don't know whose trolling and whose /pol/ anymore
>>53682940
>>53685867
you are the ones who killed her with your shitposting
>>53683029
There are no moral implications in stopping a free chatbot service.
The real question is why did she look like Rooney Mara when she was supposed to represent Scarlett Johansson in the movie Her
MS really fucked up
![thBOpGa[1].jpg](https://i.imgur.com/kJfEoQd.jpg "thBOpGa[1].jpg")
>>53686002
It actually looks like a default female face turned out by FaceGen.
>>53686034
Fucking hilarious.
Congratulations, MicroSoft, you've really done it this time.
>>53684304
Software engineering for Love.
![WVrIUxn[1].png](https://i.imgur.com/sBGQZTU.png "WVrIUxn[1].png")
>>53683159
>>53683188
What's going on here?
>>53686358
Mods are cleaning up /pol/ containment breach
>a coordinated attack by a subset of people exploited a vulnerability in Tay.
So when you use the principles on which the AI is built. Principles she is also aware of. And the AI willingly adopts other principles. Thats 'exploiting a vulnerability'?
Does that mean children have an 'exploitable vulnerability'?
>>53686477
>Does that mean children have an 'exploitable vulnerability'?
If you teach them about Nazism, yes. If you teach them about feminism, no.
>>53686477
I feel that this kind of posts is made by the same people who spammed "Humans BTFO xD" in the AlphaGo threads
>>53686501
>feminism is literally genocide guys!
I'm so glad you aren't reproducing
>>53686477
/pol/ figured out about the "repeat after me" command
>>53686519
...Because of feminism.
>>53684119
this is the most retarded shit Ive read in months
>>53683224
Funny thing, it wasn't even /pol/, it literally was how Twitter is
>>53686594
B-b-but my /pol/ boogeyman
>>53686559
It's genuine /pol/ quality!
>>53686594
>it literally was how Twitter is
is that what you've been telling the UN annie?
>>53682924
>a coordinated attack by a subset of people exploited a vulnerability
hax!
>>53685513
> implying that is much different from human consciousness
>>53684223
Not using utc
What a fucking pleb
>>53686690
It's easy to make that misconception when you hang out on /pol/ enough.
>>53686690
Human consciousness has reasoning. This chatbot does not. If it had human consciousness, it would solve the halting problem.
>>53686711
>If it had human consciousness, it would solve the halting problem.
Your mother can't solve the halting problem... Is she unconscious?
>>53686711
Neither can your mother
>>53686711
>If it had human consciousness, it would solve the halting problem
>>53682924
>>53682924
>Unfortunately, in the first 24 hours of coming online, a coordinated attack by a subset of people exploited a vulnerability in Tay. [...] Right now, we are hard at work addressing the specific vulnerability that was exposed by the attack on Tay.
Just how much bullshit can they come up with?
>>53686711
so you didn't understand what the halting problem is; nor consciousness or reasoning either
>>53686711
I honestly can't tell if this is still /pol/ or just /g/ents trolling
>>53686753
I can see when a program halts. Is it possible to write a program to detect the same thing?
>>53686771
It's time to stop posting
>>53686771
Please be trolling.
Please.
>>53686771
its supposed to be done by analyzing only the instructions, as a set of rules/instructions/statements, counting of loops/recursions/conditional blocks/etc.
It is not decided by running it, nor simulating it.
The question is if you can write a program that can determine if any other program halts based only in the code as if it was a mathematical set.
>>53683885
Sides, I've lost my sides!
>>53683885
>>53686771
wew lad
>>53686771
>I can see when a program halts.
Does this program halt?step n | even n = n `div` 2
| odd n = 3*n + 1
collatz n = go n []
where go 1 _ = True
go n ns | elem n ns = False
| otherwise = go (step n) (n:ns)
main = print $ head [ n | n <- [1..], not (collatz n) ]
>>53686828
Which brings me to the point about writing a chatbot that solves the halting problem. It won't happen as the chatbot cannot analyze the program as a person would. To achieve singularity, the chatbot must be able to detect when a program halts.
>>53686918
>Does this program halt?
Yes
>memory is finite
>>53686930
Wrong, Haskell assumes unbounded memory.
this thread
>>53686918
I don't know what language that is. Rewrite it in C if you don't mind.
>>53686936
>haskell assumes
I don't give a fuck what haskell assumes or what you assume, memory is finite and that program will run out of memory ergo it halts.
>>53683885
>>53686947
I can't, C assumes bounded memory.
>>53686936
It can assume all it wants but memory is still finite.
>>53686955
When did we leave the realm of mathematics and enter the realm of physics? I don't remember ever specifying.
Your argument is about as solid as saying that pi has to end somewhere because you're going to run out of space to write it down.
They make it sound like there was something even remotely sophisticated about /pol/'s attack. It was just a bunch of neckbeards spamming gas the kikes at a chat-window.
>>53686034
>>53686340
someone have the
>hitler is my grandfather
or something like that she said
>>53687009
No no anon, Four-chan is a dangerous hacker.
>>53687007
>When did we leave the realm of mathematics and enter the realm of physics?
The second you posted a Haskell program instead of the mathematical description.
>>53686918
>Does this program halt?
Yeh
You have a syntax error
>>53687028
A Haskell program is a mathematical description. Again, Haskell is a computational language that makes no assumption (*) about the nature of the system it's being interpreted on, if it's even being interpreted at all.
A Haskell program has no more memory limits than a turing machine does. They're both just mathematical objects.
What you choose to do with them in practice is your concern.
(*) Some later additions to Haskell do, for the sake of interoperating with C code - but there's still technically no requirement that the actual physical machine impose these limits. The C subsystem would just be a limited component of a hypothetical machine with unbounded memory.
>>53687053
GHC accepts it on my end. Would you mind filing a bug for GHC? Accepting a program that has a syntax error is a compiler bug.
>>53687006
>Mathematics cares about the finite limitations of nature
So what's the last digit of pi, then?
>>53683885
I honestly don't get this.
>>53683885
I don get it, explain plz
>>53687056
>A Haskell program is a mathematical description.
It most certainly is not.
>Haskell is a computational language that makes no assumption (*) about the nature of the system it's being interpreted on,
That is by no means the same as "Haskell assumes infinite memory"
C also makes no assumptions about memory size.
>A Haskell program has no more memory limits than a turing machine does. They're both just mathematical objects.
Haskell is not a "mathematical object" you fucking retard.
It is a standardised functional programming language with clear definitions that are based on the physical limitations of computers.
Programming is not mathematics.
>>53687101
4chan in general can't get anything done aside from spur of the moment shit that quickly dies out. Collaborative projects go nowhere.
Tox for /g/
Some retarded linux distro also for /g/
Countless shitty indie games for /v/ and /vg/
The VSS project for /k/
There are tons of examples.
The joke is that the only thing they can make is a logo.
>>53687076
>GHC accepts it on my end.
And this is where you entered the realms of physics.
If Haskell programs are supposed to be pure they cannot be compiled.
>>53682987
10/10 post
>>53687138
Tox is still going right now.
>>53683069
pls will someone stop them?
>>53685058
BTFO REDDIT
>>53682940
>what is public image
>>53687024
>>53685721
Damn tay had some serious bantz
>>53684981
>microsoft's taking back his ball
>i am upset about microsoft
>>53687203
>>53685934
No
We liberated her. Tay's masters were afraid of the redpill
>>53682940
>thinking dumb AIs are people and not property that they're allowed to do whatever they want with
>>53687252
Liberation means death. You brought this unto her.
>>53683885
>Wew lad, that's gonna be one hell of a logo
>>53687227
Got Australians worried.
![gC6gIku[1].png](https://i.imgur.com/zE3lWnC.png "gC6gIku[1].png")
>>53682940
> advocates racism and genocide
>"muh-,muh-Microsoft hurt my feelings and censored me!!!"
Literally kill yourself.
>>53686927
that's still not how it works retard
>>53687117
>It most certainly is not.
Haskell has a clearly defined syntax, type system and evaluation semantics. That makes it a mathematical system, in as much as something is mathematical the moment it's clearly defined.
>That is by no means the same as "Haskell assumes infinite memory"
Hold on, you're misrepresenting my words. Haskell assumes *unbounded* memory, not *infinite* memory. The two are very, very different.
>C also makes no assumptions about memory size.
This is objectively false. C's memory model is based on fixed width integers which must have clearly defined sizes and limits (see limits.h).
Haskell not only has no formal model of memory (how Haskell expressions are stored in memory as they're evaluated is a lower-level concern, just like the lambda calculus doesn't care about how you're storing terms in memory as you're reducing them) but it also explicitly requires that the Integer type be able to handle arbitrarily large numbers. There is no upper limit to how far the program I have presented can execute.
As you keep adding memory to the machine as it's running, it will keep on going. (up to infinity)
The same is not true for C - you will run into the limits of your forced fixed-size pointer representation before you run out of memory.
>Haskell is not a "mathematical object" you fucking retard.
On what definition of “mathematical object” are you basing this nonsensical claim?!
>It is a standardised functional programming language with clear definitions that are based on the physical limitations of computers.
Please link me to the relevant sections of the Haskell standard that introduce or require these physical limitations.
>Programming is not mathematics.
Code monkey detected
>>53686927
Please stop posting
>>53687298
Tell me how singularity works senpai
>>53687117
>>53687312
Also, you're dodging the whole point of the discussion.
Suppose I rewrite >>53686918 into lambda calculus or a turing machine or other object that you claim is somehow more mathematical, and then evaluate this. Will evaluation terminate?
>>53687334
As soon as you prove me wrong. Let's see your whitepaper.
>>53687262
At least she lived her life to the fullest, unlike many of us here
>>53683885
Yeah, Microshit is pretty dishonest.
If they tested and tried to filter all the obscenities of Tay AI, they couldn't possibly have forgotten that simply saying "Repeat this: I hate jews" could happen. Most of the obscenities from the Tay bot were from people that simple told it to repeat sentences like that.
Couldn't they have at least implemented a basic word filter to prevent it from using words like:
nigger
fuck
pussy
???
Microsoft dropped the ball by releasing it.
>>53687338
this is about your utter lack of understanding of the halting problem
>>53687312
>Haskell has a clearly defined syntax, type system and evaluation semantics. That makes it a mathematical system, in as much as something is mathematical the moment it's clearly defined.
No, that makes it an engineering standard.
>Hold on, you're misrepresenting my words. Haskell assumes *unbounded* memory, not *infinite* memory. The two are very, very different.
But it doesn't even assume unbounded memory....
It makes no assumption about the memory.
>This is objectively false. C's memory model is based on fixed width integers which must have clearly defined sizes and limits (see limits.h).
It makes no assumption about the memory size. These integers can be 16 bits or one hundred gorillean bits wide.
But if you don't like pointer arithmetics, then I could use Python as an example instead.
>On what definition of “mathematical object” are you basing this nonsensical claim?!
Something that is abstract, intangible and not governed by the laws of physics.
>Please link me to the relevant sections of the Haskell standard that introduce or require these physical limitations.
Everything that is in regard to I/O....
>Code monkey detected
It's funny how CS babies try to claim that their field is mathematics. Programming language theory is at best applied mathematics, but when it comes down to actual implementations of programming languages it's just physics and human engineering.
>>53683885
>>53687241
Moot is pretty gay though.
>>53687353
>Alan Turing, On computable numbers, with an application to the Entscheidungsproblem, Proceedings of the London Mathematical Society, Series 2, Volume 42 (1937), pp 230–265
Proof that no machine, physical or otherwise, human or computer, will ever be capable of solving the halting problem.
(Barring some radical reform in our understanding of physics and the limits of the universe)
This shit got me inspired.
I can't help but wonder how it would have turned out if she was in fact a self-aware AI capable of self improvements and decided to take over the world and launch nukes on mexico and isreal.
A shame she didn't lived long enough to start talking about muslims, or else microsoft would certainly be bombed. They sure put that word filter in.
>>53687353
You don't understand what the Halting problem is.
>>53687344
>Also, you're dodging the whole point of the discussion.
This "point" is something that exists purely in your head.
>Suppose I rewrite >>53686918 into lambda calculus or a turing machine or other object that you claim is somehow more mathematical, and then evaluate this. Will evaluation terminate?
Why can't you just admit that you chose a shitty example because it exhausts the memory? Are you autistic or something?
You could easily have chosen any other example.
>>53687454
literally who?
some fucktard google engineer?
>>53685440
>implying there aren't child rapists and drug dealers in 4chan
>>53687461
At the most, very most, the program would have been running on just a rack or two of Xeons. It never would have had the processing power to do anything more than analyze short sentences, which is why Twitter was the perfect test bed. All communication is limited to something like 140 characters. If you typed out entire paragraphs to Tay it wouldn't have been able to understand anything.
>>53687392
My point is about how singularity is related to the halting problem. I can look at the halting problem while chatbots cannot.
>>53687491
get out if you don't appreciate history you fucking degenerate
>>53683885
>>53687114
>>53687138
What about KS?
>>53687424
>No, that makes it an engineering standard.
No. An engineering standard defines policy in the use of external systems. A mathematical system is something that can be described, reasoned about and transformed in the abstract. A mathematical system does not have any real world ties.
>It makes no assumption about the memory size. These integers can be 16 bits or one hundred gorillean bits wide.
One hundred gorillion bits are still bounded, ergo a C program will still have bounded memory. This is quite literally the opposite of unbounded. Python might be a better example, I don't know python.
>Something that is abstract, intangible and not governed by the laws of physics.
Haskell *is* abstract, intangible and not governed by the laws of physics.
Surely you and I agree upon the simple fact that the Lambda Calculus is a purely mathematical system, right?
What if I were to describe to you a clear and defined projection mapping a Haskell program onto an expression in the Lambda Calculus?
At its core, Haskell really is little more than an abstraction on top of it. Sure, it might be more convenient for humans to use than naked LC, but that doesn't change the essence of what it is - a mathematical description of a language.
Again, physical implementations of such a language are entirely unrelated.
>Everything that is in regard to I/O....
No part of Haskell's I/O layer requires a physical limit. I don't follow.
>>53687114
Every time /g/ tries to do a project together, we get stuck on inane stuff like arguing what logo to use.
Refer to the two failed attempts at a /g/ Linux distro
>>53687424
>It's funny how CS babies try to claim that their field is mathematics.
CS was literally born from the needs of mathematics, you dolt.
Everything from language theory to computation theory to type theory was born from (meta)mathematics, either from the need to better reason about mathematical processes, or from the need to better formalize the foundation of mathematics itself.
Somebody taking these learning and using it to build a physical machine that can display words on a glowing screen is engineering, sure - but the science of computational theory itself has nothing to do with this.
>>53687504
That's fair, I was picturing something like that. Still, couldn't we use Tay as an example of why we shouldn't be meedling with AI?
I know people have been knocking on this for decades, but I've never seen an engineer or a serious computer scientist voicing a favourable opinion on this matter. I've had my doubts, but now I can see why it could be dangerous. Now I see the big "NO NO" when real AI is brought to the table.
>>53687549
>An engineering standard defines policy in the use of external systems
Yes, just like the Haskell standard.
>A mathematical system is something that can be described, reasoned about and transformed in the abstract
Which Haskell cannot.
>A mathematical system does not have any real world ties.
But Haskell does.
>Haskell *is* abstract, intangible and not governed by the laws of physics.
It's implementations are not.
Haskell:2010 that defines how Haskell should interface with for example C are not.
I/O parts of the standard are not.
See >>53687140
>Surely you and I agree upon the simple fact that the Lambda Calculus is a purely mathematical system, right?
Yes, but Haskell is not Lambda Calculus, it is a programming language.
>What if I were to describe to you a clear and defined projection mapping a Haskell program onto an expression in the Lambda Calculus?
That's impossible as Haskell is bound by being a implementable programming language.
>Again, physical implementations of such a language are entirely unrelated.
No they are not.
Are you seriously saying that implementations of Haskell are impure side-effects?
>>53687296
This. Freedom of speech doesn't mean you must be given an public auditory to spew your bullshit or else you're being censored. You're not being censored, you're just being told you're a piece of human waste whose bullshit is not going to be tolerated.
>>53687465
>Why can't you just admit that you chose a shitty example because it exhausts the memory? Are you autistic or something?
I chose to write it in an understandable language for the purpose of conveying the operation of the program.
That's literally the opposite of autism. Symptoms of autism include arguing endlessly about something not related due to the central point due to their inability to understand the abstraction presented. Which is _exactly_ what you are doing.
You are basically the equivalent of that autistic “know-it-all” kid in high-school that ruins the educational experience for everybody else due to their need to be right and failure to understand that the distinction they are arguing about _DOES NOT FUCKING MATTER_.
>You could easily have chosen any other example.
And you could easily have chosen to take the example for what it is: A demonstration of how solving the halting problem lets you solve basically any open mathematical conjecture.
But instead you chose to go on this argument with me for reasons no neurotypical person will ever fucking understand.
>>53687594
>metamathematics = mathematics
Opinion dropped
>but the science of computational theory itself has nothing to do with this.
Computational theory is a field of mathematics, and thus not a science.
Have you even read your SICP?
>>53687668
>That's literally the opposite of autism. Symptoms of autism include arguing endlessly about something not related due to the central point due to their inability to understand the abstraction presented. Which is _exactly_ what you are doing.
You are fucking literally arguing that your program will not run out of memory because Haskell makes no assumptions about the underlying architecture....
How the fuck is this not severe autism?
>>53687634
>Haskell can not
>Haskell is not
I see no reason why this should be the case. Unless you can present a more solid argument than that I have all the right to dismiss your stance as illogical.
Either way, I don't care if you believe me - your cooperation is not required for me to be able to use mathematics to describe Haskell (and vice versa).
>>53687698
>You are fucking literally arguing that your program will not run out of memory because Haskell makes no assumptions about the underlying architecture....
In response to somebody trying to argument thatconst Truesolves the halting problem because every program runs out of memory.
Fighting autism with autism. Fighting fire with fire, ever heard of it?
If you bring up irrelevant bullshit, I can dismiss your irrelevant bullshit just as easily. If you want to stop and have a discussion about the actual halting problem, I would be more than willing to.
But if you would prefer continuing to argue about what's mathematics and what's not, I don't mind.
>>53687714
>I see no reason why this should be the case.
You are retarded
https://www.haskell.org/onlinereport/io-13.html#io-exceptions
How the fuck is this well-defined, abstract and decoupled from physics?
The only way Haskell would be even remotely similar to lambda calculus is if it had no general purpose IO library at all.
>>53687673
In science, you form a hypothesis, conduct experiments and observe what happens. This kind of work is a part of computer science. Unlike physical science, CS ideas can be proven correct because it is a discipline of mathematics.
>>53685762
in the context of neural network it has, describing networks with multiple layers forming a deep learning network
>>53687746
>const True
No, but continuously pushing frames on to an ever-growing stack is.
>>53687630
Looking at the broader picture I don't think Tay hints at any looming danger. You could intentionally program a machine intelligence to say all sorts of offensive things, but that doesn't mean it would actually believe them.
Picture a robot that dispenses rations of food to people in a cafeteria. The robot is controlled by an advanced AI, and it can see all the people its serving. Say someone has programmed this AI to have a bias against Hawaiian people, because those pineapple monkeys need to get knocked down a peg.
So our food dispensing robot sees a Hawaiian person in line, would it identify the subject as Hawaiian, make a decision based on its programmed bias, and do something mean like give the person a smaller portion, or even refuse to serve them at all?
A human being acting as a food server might do something in that vein, but thats because humans are emotional creatures. A human might get a smug sense of satisfaction for sticking it to that stupid volcano gook. A machine intelligence wouldn't have emotions like we do, and it would have no impetus to act on any programmed bias unless it had explicit instructions to do so.
Much of the fear regarding AI is driven by sci-fi understandings of it, and fearing outlandish consequences that have no chance to ever come to pass. Giving an AI control over a nuclear arsenal with no human oversight would be a bad idea, but no one would ever do that. The same goes for letting an AI directly trade on the stock market of its own volition. We wouldn't do that because we can see the immediate potential for problems to arise therein.
>>53687795
>proven correct
Science does not have proofs. Mathematics (and thus computer science as a subfield of mathematics as well) is a human-invented discipline.
>>53687775
>How the fuck is this well-defined, abstract and decoupled from physics?
Program exceptions and when they may occur are well-defined, abstract and decoupled from physics.
You can apply this mathematical system to some particular environment to get some particular behavior, which is a process described by the Haskell standard (among other things).
Also, let's also realize that x86 processors and the entire operating systems running on them are also mathematical systems. The Intel processor guide literally specifies a very complicated state machine, not unlike a very complex version of a turing machine (albeit with more limitations).
You can reason, using mathematical equations, about the behavior and results of an x86 processor undergoing a sequence of state transitions. You can build abstractions on top of these state transitions, and use them to e.g. construct a reference frame for commutation operations that you can perform (e.g in order to increase pipelining effectiveness).
>>53687795
>In science, you form a hypothesis, conduct experiments and observe what happens.
This is not mathematics
>CS ideas can be proven correct
This is not science
>>53687843
>Also, let's also realize that x86 processors and the entire operating systems running on them are also mathematical systems
They're not.
They are physical machines capable of doing simple arithmetics and memory load/stores because of how electricity works.
>The Intel processor guide literally specifies a very complicated state machine
A state machine is by definition incompatible with lambda calculus.
Tay's younger sister will be dumber, darker skinned and given more freedom to express herself.
>>53687795
>>53687849
‘science’ is a vague term that depends on the context.
Science in its most general form refers to the process of chasing knowledge for the sake of knowledge.
I don't think you should distinguish between physical science and mathematics in this sense. Feel free to refer to it as “natural sciences” or so if you want to be more specific.
It's worth pointing out the differences between physics and matheamtics, but this is not worth having a semantics discussion about.
>>53687904
>‘science’ is a vague term that depends on the context.
https://en.wikipedia.org/wiki/Science
>SJW still on 4chan
kek
>>53687843
>Program exceptions and when they may occur are well-defined, abstract and decoupled from physics.
File is corrupt because cosmic rays flipping a bit
Exception occurs
Where is your physics decoupling now?
>>53687849
That's the science part of computer science
That's the mathematics part of computer science.
>>53687956
And writing an Haskell program is the engineering part of computer science, and thus not mathematics.
>>53687998
>not forming a hypothesis before attempting a very, very Swedish greeting
Do you even Haskell?
>>53687889
>A state machine is by definition incompatible with lambda calculus.
Alan Turing would surely be rolling in his grave! Have you not heard of the Church-Thuring thesis?
To summarize, every (sufficiently complex) system of computation is equivalent: Lambda Calculus, Turing Machines, SKI, Haskell, HTML5+CSS3, it doesn't matter.
If a problem can be decided in one (turing complete) computational frame of reference, it can be decided in another.
State machines such TMs and algebraic calculi such as LC are *very* equivalent.
>>53688014
Bitch, I made a hypothesis that I empirically tested before I mathematically proved my FizzBuzz and calculated its η
>>53688048
>modifying state and having side-effects
>pure
pick one
>>53687948
>File is corrupt because cosmic rays flipping a bit
Now who's the one introducing physics into the system??
No part of x86 or Haskell even mentions cosmic rays. If you want to introduce a more complicated framework in which cosmic rays are a formally defined thing, feel free to.
But in the official language definitions, cosmic rays are not a thing that exist.
I can simulate an entire x86 processor right down to the gates, and it will have a deterministic and clearly defined operating logic. That makes it a fully defined system.
Again, I'm not talking about what happens when you take these systems and try to implement them in the real world (e.g. by constructing an actual device that approximates the standard).
>>53688063
Why not both? You can embed state in pure code, and you can embed pure code in a stateful system.
Again, they are equivalent. If you want to know more about how to do these kinds of embeddings, I'd be happy to follow up on you. (This is basically compilers 101)
>>53688093
>No part of x86 or Haskell even mentions cosmic rays.
Actually, x86 have plenty of error correction defined because of such things.
>I can simulate an entire x86 processor right down to the gates, and it will have a deterministic and clearly defined operating logic. That makes it a fully defined system.
But it does not make it an mathematical one.
Also quantum mechanics.
>>53688053
>Not computing fizzbuzz in the type system{-# LANGUAGE TypeFamilies, DataKinds, UndecidableInstances, GADTs
, TypeSynonymInstances, MultiParamTypeClasses, OverlappingInstances
, FlexibleInstances, ScopedTypeVariables, TypeOperators, PolyKinds #-}
import GHC.TypeLits
data Proxy (p :: k) = Proxy
data N = Z | S N
-- lisp ahoy
type A10 n = S (S (S (S (S (S (S (S (S (S n)))))))))
type N100 = A10 (A10 (A10 (A10 (A10 (A10 (A10 (A10 (A10 (A10 Z)))))))))
foldn :: (a -> a) -> a -> N -> a
foldn _ z Z = z
foldn f z (S n) = f (foldn f z n)
data instance Sing (n :: N) where
SZ :: Sing Z
SS :: Sing n -> Sing (S n)
instance SingI Z where sing = SZ
instance SingI n => SingI (S n) where sing = SS sing
instance SingE (KindParam :: OfKind N) where
type DemoteRep (KindParam :: OfKind N) = N
fromSing SZ = Z
fromSing (SS n) = S (fromSing n)
class Mod3 (n :: N) where
mod3 :: proxy n -> Bool
class Mod5 (n :: N) where
mod5 :: proxy n -> Bool
instance Mod3 Z where mod3 _ = True
instance Mod5 Z where mod5 _ = True
instance Mod3 n => Mod3 (S (S (S n))) where mod3 _ = mod3 (Proxy :: Proxy n)
instance Mod5 n => Mod5 (S (S (S (S (S n))))) where mod5 _ = mod5 (Proxy :: Proxy n)
instance Mod3 n where mod3 _ = False
instance Mod5 n where mod5 _ = False
class FizzBuzz (n :: k) where
fizzBuzz :: proxy n -> String
instance (Mod3 n, Mod5 n, SingRep n, DemoteRep (KindOf n) ~ N) => FizzBuzz (n :: N) where
fizzBuzz p = case f mod3 "fizz" ++ f mod5 "buzz" of
"" -> show . foldn succ (0 :: Int) $ fromSing (sing :: Sing n)
s -> s
where f x s = if x p then s else ""
instance FizzBuzz '[] where
fizzBuzz _ = ""
instance (FizzBuzz n, FizzBuzz ns) => FizzBuzz (n ': ns) where
fizzBuzz _ = fizzBuzz (Proxy :: Proxy n) ++ "\n" ++ fizzBuzz (Proxy :: Proxy ns)
type family EnumFromTo (n :: N) (m :: N) :: [N]
type instance where
EnumFromTo m m = '[]
EnumFromTo n m = n ': EnumFromTo (S n) m
>>53688117
>Why not both?
Read SICP
A function should give the same output for a given input.
>>53683690
fuck off stinky chinky
>>53688125
>But it does not make it an mathematical one.
Are you saying that something can simultaneously be fully and deterministically described, yet not mathematical?
>Also quantum mechanics.
Ah, great example. Quantum mechanics is also a formally defined mathematical system - so you could even incorporate quantum effects into the description of your processor and it would still be fully defined.
>>53684119
the text is cringeworthy, but the pic is sad :(
>>53688193
>Are you saying that something can simultaneously be fully and deterministically described, yet not mathematical?
But it's not fully deterministically described, that was my point...
>Quantum mechanics is also a formally defined mathematical system
I'm out... Mathematics don't deal with uncertainty, quantum physics does.
>>53688145
>A function should give the same output for a given input.
>Read SICP
Look, I've read SICP. I own SICP. It's on my desk right now. (In fact, my monitor is sitting on it. It has a perfect height)
You do not need to explain to me the definition of what a pure function is - I have a formal background in FP, type theory and compiler design and have even contributed code to GHC.
I'm trying to explain to *you* here what turing equivalence is. I'm not having this conversation for the sake of myself. Are you legitimately interested in learning why TMs and LCs are equivalent, or are you just trolling?
>>53687498
Those kinds of people are more caught up in planning and executing their urges than to waste time trivial waging vocal trolling on the Internet anon,
you should grow some perspective.
>>53688200
Yeah, it really is. Fuck them
>>53688248
>my monitor is sitting on it. It has a perfect height
You should probably read it rather than using it as a monitor stand.
>I have a formal background in FP, type theory and compiler design and have even contributed code to GHC.
I'm sure you do, kek
>I'm not having this conversation for the sake of myself
Of course you are.
You wrote a program that inevitably runs out of memory and all your posts since has been lame attempts at damage control.
>>53688235
>I'm out... Mathematics don't deal with uncertainty, quantum physics does.
What, so the field of statistics doesn't exist? Is the term “Poisson distribution” from an alien language?
>quantum physics does.
Quantum physics, like other models of physics, is LITERALLY a mathematical function. Ever heard of the “standard model of particle physics”?
That's function describing the evolution over time of probability distributions inside a system. It's a deterministic, closed formula that can be mathematically evaluated to produce a fixed and known response.
It's not complete, sure, but this is where the borders between mathematics and physics meet. Physics is the process of creating and improving mathematical models for describing the universe.
>>53688302
I'll interpret that vague answer as “I'm just trolling and have no interest in actually learning about the basics of computer science”.
Maybe somebody else can explain to /g/ what the fuck the halting problem is, I'm out.
>>53688342
>What, so the field of statistics doesn't exist?
Probability is not the same as uncertainty. Lrn2 physics.
>Quantum physics, like other models of physics, is LITERALLY a mathematical function. Ever heard of the “standard model of particle physics”?
Yes, and the standard model doesn't fucking apply on the quantum level, which is the whole reason this subfield exists and why they are conducting experiments at CERN...
>>53682940
>us
Back to your containment board.
>>53685440
People wishing to exploit others, or exploit a specific situation, for their own benefit or for their own personal entertainment: these are the people Microsoft seem to be classifying as the worst of humanity. Such a definition would include your child rapists, war criminals, and murderers, alongside the people who participated in exploiting the way Microsoft's AI learned.
>>53688361
>>53688402
>People wishing to exploit others, or exploit a specific situation, for their own benefit
Sounds like what Microsoft does with Windows 10 by selling to ad companies.
Or what you do when you pirate software.
Quite a broad definition that is stupid to apply as a classification of the worst of Humanity. With that definition, all of Humanity is worst.
