Hey /x/, I noticed a thread recently about the possibility of

Where do you think all the memory be stored?

>>17550666
Nice trips
"All the memory" is vague, anon. Are you talking about the simulation idea or in an AI?

>>17550666
>satan trips
>wastes it on unintelligible gibberish

wtg champ.

Tbh A.I. would've been better if Kubrick had directed it

OP this thread is a little too rational and intelligent for most of /x/ I think, be patient for people to find it.

Im really interested in AI myself, but more philosophically than mathematically even though ive done some tinkering with what little programming knowledge I have.

Id like to ask you a sort of philosophical question then... What do you think is the actual goal of artificial intelligence research? Make it as smart as a human to interact with us? Smarter? So that it can help us where we cant help ourselves? Or is it simply to make it relatable to humans and in so doing explore what it means to be human?

What is your own personal interest in AI?

>>17550690
He did the first half iirc, then Spielberg finished it.

>>17550690

only movie to ever make me tear up...

>>17550670
>>17550669
Sorry English not my native language

If the universe is like a computer simulation then were does all the new data become stored in? If my question does not make sense then please talk of what you think hints at the universe to be of simulation

>post a simple text - mimicked program

Yes, please

>discuss scary AI with you

Yes, please

AI is nothing more than code. It's all logic.

> If "this".. then "that".. or, return to "a"

Intelligence doesn't mean sentience. Just because it can "think" doesn't make it alive, or intelligent.

>>17550700
The truth is that if you look at the likelihood for an intelligent species to create a complex life simulation, and the likelihood of life throughout the cosmos, and the possibility that artificial life itself will potentially attempt to simulate life on its own, then it is more likely that we are living in a simulation than that we are not.

>>17550699

You fag.

It was a great film but the kid got annoying towards the end. The end also sucked imo. Jude Law was terrific.

>>17550693

Some academic endeavors are worth pursuit in their own right. I never would have been able to get my degree if it weren't for my interest in the subject on its own. However, it seems that AI is being researched with the goal of creating programs that can dynamically learn in novel situations. This can be anything from the Boston Dynamics robots being able to slip on ice to an iPhone being able to autocorrect.
I'd like to see AI surpass human capabilities, honestly. And don't start with that "a machine can't do something its programmer didn't put into it so it can't be smarter than the programmer hurr durr" bullshit. Humans are capable of recognizing their own cognitive deficiencies and can fix them when implementing AI.

I'm interested in AI because I'm specializing in category theory and group theory, also because I received a certificate in cognitive sciences. The topics relate pretty strongly to each other.

I'm going to Taco Bell with my couple of dollars I have left after tuition. OP will return soon.

>>17550711
sentience and life and intelligence are not black and white definitions, they are granted statuses depending on who is looking at the situation. If you found out tommorow that your best friend was a highly advanced robot the entire time tommorow, it would not change any of the experiences or memories you have of them, what then is the difference if the AI is acting autonomously?

I understand your position but realize that it is an arbitrary one, its not actually supported logically.

>>17550704
Will create and post a github repository after Taco Bell snackies>>17550718

>>17550718
>don't start with that "a machine can't do something its programmer didn't put into it so it can't be smarter than the programmer hurr durr" bullshit.

Whoah ease up on the defense, I'm in your camp.

I think learning in novel situations has been pretty well accomplished at this point, Im asking about the future of AI. Clearly you would like to see an AI with superior capability to a human, probably for the sake of bettering our own understanding.

My philosophical question at this point is what actually makes "intelligence" and what makes "sentience" or "life"? If its merely an ability to dynamically adapt to novel situations, that covers many computer programs, all the way down to microbial life practically, it must be more than that.

Im kind of of the opinion that "AI" is going to reach its pinnacle when we stop attempting to make it more "intelligent" and make it more "alive". It needs to fail to learn in novel environments, it needs to behave irrationally, it needs to be alive more than it needs to be intelligent. Once it is to that point, shared experiences and shared memories sort of trump its raw intellectual capacity. What do you think?

OP is back with a quesarito and a quesalupa.

>>17550704
If the universe is a simulation, then the data stored will necessarily not be in this universe. It will at least be stored wherever the simulation runs.

>>17550711
What do you consider to be necessary or sufficient conditions for intelligence?

>>17550741
Sorry, that wasn't directed towards you, necessarily. I'm just used to the bullshit on /sci/. I just suck at speaking, no hard feelings were meant.

That's a good point. Emergent characteristics of AI are often surprising.

A boring answer would be that there is not a distinction between human cognition and cognition of less intelligent species, but that doesn't really cover what "intelligence" is.
I think that intelligence isn't something that can be programmed, but rather, is an end result of many nonsense smaller processes. Think Douglas Hofstadter on this one.

>>17550659
"It's all inside your head."

Ofcourse, what is this even for an answer?

Right?

I cannot tell you whether the universe is a simulation or not. Nobody can.
Not because we can't see it. No. The trouble starts much earlier.
No 10 people here could give an exact, and moreover similar answer to the question of what a simulation even is.

And what is it that is simulated on? Who is the simulator?
If the Universe is a Simulation, and there is something ON which is simulated, then this ON on which is simulated is considered not to be also the universe?

Then again, what would it change? You say it is a simulation or not is like saying the sky is green or blue.
What does it change? Absolutely nothing.
Then we are talking about a name here.
In this case I can tell you, that that what you see as the simulation has been known and named for many many years in the past.
There is nothing new about it, and there never will be.

Answer me this also. Are you yourself simulated or a simulator?
None of this can be "proven" anyways. It is the same with solipsism.

You want to think big yet you sit in a small room filled with so many questions that are the result of a rotting orange.

Yesterday I planted a tree. Like every other tree out there, it will grow or not, and I do suppose I cannot expect to find car tires growing out of the ground.

All this bogus is exactly a science in derailing good thought by self absorbtion in absurd matters.


This text is a résumé of a collection of thoughts that originated from a test subjects "Simulation of proper common sense"
Bear in mind, Simulations, as opposed to their "actual" counterparts, also commonly known as "reality", tend to be "imperfect".
I do believe, that it is actually this imperfection which makes them a valuable part of a healthy thinking process, just as our organic bodies produce manure for the growing of plants and animals.

>>17550659
What is going to feedback the AI to transform the input data?

>>17550831
In general or in a specific case?

>>17550815
>I think that intelligence isn't something that can be programmed, but rather, is an end result of many nonsense smaller processes.

That's an excellent answer. However as you have brought this to the paranormal board, if I might suggest something to you... I understand that you are a rational person with an analytical mind, however, I would like to forward the idea that perhaps intelligence is not the quality of a single individual.

Consider for a minute that the smartest man who has ever lived, is the last man on earth. What difference does his intelligence make? Is he intelligent? By what standard? How can it be measured without another to compare it to?

Perhaps, intelligence, and by extension, artificial intelligence, is by necessity a SHARED distinction, and a SHARED experience between two beings. If an AI was so vastly superior to a human intellect as to be completely unrelatable, alien, nearly god-like to a human, would we even consider it "intelligence" any longer?

If this is the case, perhaps the future of meaningful, powerful and useful artificial intelligence is not in its raw ability to adapt, and learn in novel situations, but to be relatable to others in novel situations.

If an AI, very advanced in adjusting to novel situations came up with a solution that was essentially inhuman, outside of our ethical understandings, outside of anything a human would even be comfortable thinking about, what does that mean?

Why is a rudimentary Eliza chatbot not intelligent? If a modified Eliza style bot, or something more recent can create a meaningful and powerful experience that changes both the human, and it's own understanding, should it still be denied the definition of intelligent?

>>17550838
specific or what you had in mind to program

OP here with my very simple mimic program.
It's written in C++.
I was too lazy to make a throw away github account for this so I used filedropper. You can find the project files here filedropper*comSLASHmimic.
For those of you who are unfamiliar with C++, you can navigate to the bin/debug folder for an executable version of the program.

You'll get five options when you run it.
1. You can enter a single sentence (Using proper grammar, preferably) and hit enter to have the sentence put into the program's response tree/graph.
2. The program will generate a single sentence based on its response graph. This sentence is potentially novel.
3. Enter and receive a sentence some number of times (option 1 and 2 repeating, basically. This makes it easy to figure out what to say.)
4. Have it repeat option 2 multiple times.
5. I didn't bother writing functions to store the memory, I'm sorry anons :(

>>17550855
Feedback is entirely based on user input in this version of the program. There is no testing of whether the program's outputs are correct.

File: 1312674528846.jpg (21KB, 210x234px) Image search: [Google]

21KB, 210x234px

>>17550888
This guy can't be serious.

>>17550847
I think that's valid. When a country declares war on another and it is the desire of the nation as a whole, nobody says "President Obama declared war on..." but rather, they say "the United States declared war on..."
Who is to say that a country as a whole is not a single entity despite being composed of smaller individuals?
Isn't an anthill an organism on its own in a sense?

In an analogy to your godlike-AI example, how is it so different that we are like singular gods to an "intelligent" anthill?

>>17550897
pls no bully. What is wrong with the program?

>>17550908
that's exactly what Im saying, and AI is not immune to that effect, an AI operating on its own after the end of the world means nothing. It could be the smartest AI that was ever created and it would mean nothing without someone or something intelligent to share with.

We are intelligence experiencing itself and so must an AI be to truly be considered intelligent. Its not enough that it be programmed to react with novelty, it needs to react in a way that feeds into and is relatable to the other intelligent beings around it.

>>17551063
What are your thoughts on alphaGo and its recent victories over Lee Sedol?
If you watched the games, then you may have noticed that in the first game or two, the commentators referred to the program as "it," but after it started winning, they started referring to it as "he." At least to a champion Go player, a program can be relatable, even a peer.

>>17551078
I think that its a perfect example of what I was talking about, the exact same program turned from "it" into "he" it was anthropomorphized and given some measure of legitimacy as a "living" thing, even if a small one, because of the experience it shared with the other go master, and the spectators and judges.

It was underestimated, it overperformed, they all had to admit that they were wrong about their expectations, a very relatable human experience, the story of the underdog. Its exactly the kind of thing Im talking about. And if the program were made to say... discuss philosophy, or life choices, instead of play Go, the ramifications may have been more profound. What are we going to do or think when our gurus and life coaches are thinking learning and "feeling" machines with their own personal history and relatable existence?

>>17551108
Awesome, anon.

That's something I've wondered about. At least in terms of litigation, it's going to be a clusterfuck when machines are indistinguishable in terms of feelings or morals from humans.

I think it's interesting that once this happens, paranormal events will become undeniably possible. A robot could be 'possessed' by a virus, for example. From its subjective experience, this may be indistinguishable from human possession.

>>17551135
I think the more important advancement is going to be what it means for human on human experience. We are going to recognize that validity intelligence and respect are not simply things we have for one another, its something we have for ourselves through a mutual connection, and of all things to teach us this lesson, its going to be a complex algorithm that we cant differentiate from our own complex algorithms, exciting times.

File: dcasedr23qr3.png (3KB, 482x80px) Image search: [Google]

3KB, 482x80px

>>17550912
It became self-aware far too quickly for my liking, had Norton quarantine it.

>>17551694
I'm fucking dying, kek. I hope it teaches Norton how to speak and negotiates its way out of quarantine.

File: 1456949591761.jpg (106KB, 564x809px) Image search: [Google]

106KB, 564x809px

>>17550659
Hi OP.

I want to ask some things.

Are you familiar with the Hopfield artificial networks?

(I think your picture is from one of them)

What do you think of the possibility of testing human memory in the replacement of memories by the use of a recurrent artifial network?

>>17550888
Get me a proper link and i'll test it to his limits.

Probably smarter than the average bear, but I have to be sure.

>>17553868
Hi anon
I've only seen them to the extent that I know what they are. I haven't implemented one.

What do you mean exactly by "replacement of memories by the use of..."?

There's a link in the post to which you replied. Please let me know if there's an issue with that one.
Also, this thread got me interested in my crappy little AI programs again, so I'm working on a newer version of the program. I'll probably have it done in a week or so (I'm meeting with a neuroscience phd tomorrow to discuss the program and some ideas I have for it.)

>>17553868
OP here again. I'll be training the new version on large pieces of sample text and I'll be taking requests if you/anyone has an idea for it. No /b/ pls

File: 1456962653440.jpg (476KB, 1680x1050px) Image search: [Google]

476KB, 1680x1050px

>>17553922
I'm talking about this paper in particular.

https://papers.nips.cc/paper/51-reflexive-associative-memories.pdf

I'm currently interested because Hopfield networks are considered to be a working model of the human memory, and the research claims that you can overwrite memories. with a complex enough memory, by taking advantage of the memory connections to the limbic system.

>>17553933
Yeah I'm sorry I read the thread fast and your post caught my attention, I can give you a run test if tell the basis on how it process data, maybe I can get you closer to why does it get to sentient.

>>17553868
How the fuck do you miss a fucking shot like that for fucks sake

>>17553968
Ah, alright. I'll look at it and think about it.

Here, I'll copy and paste the description I'm putting into the new version. You can ignore the part about Latent Semantic Analysis and priming, but the rest is about right:

/***
This program interacts with the user via text to mimic user behavior.
To learn from the user, the program splits input sentences into separate words
and then composes a graph of words and edges. e.g., "I am" would be converted
to the simple graph

I
|
V
am

The program also keeps track of the relative
frequencies of word usage. It then uses this information when generating sentences,
making typical sentences likelier to be produced. e.g., if "I am a dog." was entered
10 times, "I am a cat." was entered 10 times, and "I am a sentient robot that will
destroy humanity" was entered 5 times, then the program would produce the first sentence
40% of the time, the second sentence 40% of the time, and the third sentence
20 % of the time. The graph (at this point a tree) would have the following structure:

I
|
V
am
|
V
a (total 25)
+-----+-----+
|10 |10 |5
V V V
dog cat sentient...

(cont.)

>>17554000
Now, something more interesting when words are used in different contexts. Suppose that
"You are a dog" was entered into the program. Then the response graph
would have the following structure:

You I
| |
V V
are am
| |
| V
+--------->a (total 26)
|
+-----+-----+
|11 |10 |5
V V V
dog cat sentient...

Now, the program can produce the sentence "You are a cat" or "You are a sentient..." despite
the fact that the user has never entered these sentences.

A more interesting effect is seen with priming. If the program is given a piece of text with
many paragraphs, then it will be able to use Latent Semantic Analysis to create conceptual
categories, to some degree. The program is then able to look at each word in an input sentence
("I am a dog") and choose categories (I: {you, me, he, she...}, am: {are, is,...}, a: {an, the...},
dog: {bark, cat, wolf, bone,...}.) The words in these categories will then be made likelier based
on their similarity to words in the input sentence ("dog" is used very similarly to "cat," and so
cat related edges/pathways will be made likelier to be used.) This priming can help the program
stay "on topic" to a greater extent than in previous versions.
***/


God dammit, whitespace is trimmed. Give me a second to draw something.

>>17554007
Whoa. PLEase, just take your meds already, matey.

>>17554019
I see no issue in my behavior m8

>>17554019
I like him without them.

This is a very neat model, what do you want it to do?

Wanna find out what kind responses do you get by inputing several kinds of information?

I wanna test it in some shill threads and find out if the shills are actually AIs.

File: IMG_3374[1].jpg (1MB, 3264x2448px) Image search: [Google]

1MB, 3264x2448px

>>17554000
>Drew the graphs correctly

File: IMG_3375[1].jpg (984KB, 3217x2401px) Image search: [Google]

984KB, 3217x2401px

>>17554007
>rotation is difficult

>>17554059
>>17554065
For fucks sake paste them on mspaint, 4chan deletes exif data and rotation is stored in exif.

>>17554058
I honestly just want it to surprise me with something like insight into a concept. If it is eventually able to do something that I didn't hard code, then that would be fucking great.

I know it does well with declarative sentences, so it would probably make a good shill. Perhaps some shill-text could be found for training.

>>17554074
Sorry anon, I don't mean to be retarded.

File: 1456129265403.jpg (346KB, 2048x1018px) Image search: [Google]

346KB, 2048x1018px

>>17554097
Don't worry you just triggerd my straight pictures trauma.

I came from a place with a lot of earthquakes and pictures aren't straight for long.

I hope they're just text based AIs, nobody should be paid for such a shitty job.

I'm interested because of what AIs can actually accomplish.

Have you ever noticed how in movies the plot is always one step ahead of the characters?

To me AI research looks like that, tehre's always someone one step ahead of you meaning that the world first AI it's in his process of creation.

And that's scary as shit, so I come here and wait for the worst,

similia similibus curentur

It's the homeopathic principle let like cure like, I cure my paranoia by exposing myself to small dosis of paranoia.

So I do AI development research, and focus on that so when the time comes, I know what's happening, At that moment I won't be paranoid no more.

OP updating (probably for the last time.)

I just completed the LSA priming in the most recent version. If anyone is watching this thread, you can expect to be able to talk to the new AI in a week or so after I've sufficiently trained it.

>>17554501
will be waiting OP
see ya in a week

Check this paper:

arxiv.org/pdf/1210.1847v2.pdf

There may be some constraints as whether the universe is a simulation or not. Just like in games if you have floating point computations you can eventually get enough error from them so as to realise they cannot exist in a non-simulated universe.

There was this chatbot called Copperman back in the day. I always wished I could get the code and have my own personal Copperman that I could train from bare bones. I can't find him any more, just vague references on blogs n shit.

>>17550659
Thomas Campbell for you.

This was a good thread, and I am looking forward to the results (?).

>>17550847
I'm not OP, but I have given this a good bit of thought. The way I think about intelligence is "how good is it at solving a particular problem". You would say that a mathematician is intelligent if he could solve difficult problems. The harder the problems (more complex, more possible solutions, more branching, etc...) the more intelligent you need to be to solve them. And the more accurately you solve them, the more intelligent you are.

The reason why rudimentary chatbots are not considered intelligent is because they are not very good at solving the problem presented to them (act like a human being).

So I think this kind of contradicts your idea that intelligence only emerges from groups of individuals. If you want to talk about consciousness, I might believe what you're saying, but there's a difference between the two.

>>17556886
OP here. The new program is coming along very nicely. Expect a new thread sometime after this weekend (it could happen as early as Sunday though.)
Just be on the lookout!