View Topic

Rats, I was going to use the "won't" example, but changed my mind. I should've gone with my first thought! XD

Another fun fact: "ghost" didn't always have an "h" in it. Then the printing press was invented and (due to how the printers of the bible wrote "ghost" with a "gh-") the idea of "ghost" with an "h" was popularized.

(I don't have any sources right now but I'll go look it up. I'm probably remembering things wrong. Who knows, this might be an untrue myth about printers' influence).

Yay. Language is fun.

I'm kinda curious -- have you taken linguistics papers or do you just like language? Also, how do you do the "@Person person" thing?

You just type @Elbbsas or rather @<instert usernamename here> like right there you accidentally tagged someone called Person, who it turns out is a user that exists  http://www.wintreath.com/index.php?part=profile&id=1134

If you want the BB command it's:

Code: [Select]

[member=<member number>]username[/member] though you shouldn't need that.

Also the Printing-Press thing is the same reason that we have "the" and "double-u" W are a thing:
it's not a valid source but here is a fun video about it:-

Video (from a really good series)

And I have read some linguistics papers but not very many, I am interested in it but I am not studying it thoroughly... though I got a C for my High School (A-Level) English. :P
Have you yourself?

Oh my gosh. Person, if you are reading this because you got a notification or some such thing, I apologize for bothering you.

I'm currently taking a linguistics paper/course. "Introduction to Linguistics," which has been a lot of fun, and more informative than any of my English papers thus far. Did you know that a proper, thinking A.I. is impossible until linguists figure out language? Huzzah, no robotic revolution for many decades to come.

I'll have to watch the video later tonight! It looks interesting.

Wat

Where.

Who?

ur mum

Wut? 

When!

Why‽

Edit:

Did you know that a proper, thinking A.I. is impossible until linguists figure out language? Huzzah, no robotic revolution for many decades to come.

Unless the AGI is completely unlike humans and doesn't understand language... Or if the AI figures out language itself without direct help from humans, which is the whole point of a self-improving superintelligence.

Unless the AGI is completely unlike humans and doesn't understand language... Or if the AI figures out language itself without direct help from humans, which is the whole point of a self-improving superintelligence.

And if a learning AI figures out language, then more AI can be made using that knowledge!! That would be so cool! It probably wouldn't happen in English though. Maybe Sanskrit, or Japanese, or literally any language other than English. English is a messy magpie of a language. =D

Unless the AGI is completely unlike humans and doesn't understand language... Or if the AI figures out language itself without direct help from humans, which is the whole point of a self-improving superintelligence.

And if a learning AI figures out language, then more AI can be made using that knowledge!! That would be so cool! It probably wouldn't happen in English though. Maybe Sanskrit, or Japanese, or literally any language other than English. English is a messy magpie of a language. =D

If an AI could figure out one language, it would probably be smart enough to decipher all of them. Not to mention that a recursively improving AI would end up way smarter than any human. See this graphic:

Spoiler

And while, as you said, this would be cool, it'd also be very scary. Here is a good article/blog series on the topic, if you're interested.

And while I agree that English is a messy language, I'm not sure other languages (aside from some conlangs, perhaps) are actually much better. Most languages have their own quirks, rule exceptions, etc. that complicate learning.

If an AI could figure out one language, it would probably be smart enough to decipher all of them. Not to mention that a recursively improving AI would end up way smarter than any human. See this graphic:

Spoiler

And while, as you said, this would be cool, it'd also be very scary. Here is a good article/blog series on the topic, if you're interested.

And while I agree that English is a messy language, I'm not sure other languages (aside from some conlangs, perhaps) are actually much better. Most languages have their own quirks, rule exceptions, etc. that complicate learning.

(N: I am deeply sorry about this wall of text, I got carried away. It rambles. A lot. I think I'll put it in spoiler tags.)

Stupid "it." Why didn't I use other pronouns in this.

That is true. Especially since once it works out one language, it could use a multitude of different sources written in that one language. Thinking about it, English would be easy for an A.I. to work out, considering that it (the A.I) could learn all of the languages it (English) was composed of.

It would need to be a very, very smart A.I to make the step between "mimicking" language and "learning/knowing/understanding" it, however. It's like Google Translate. That works via massive processing power and matching translated articles, so good luck getting all the implications of "shut up" with it without someone actually programming it in.

Of course, it would grow downright impossible for humans to work out if an A.I is mimicking or understanding as A.I technology improves, be that to "general" or "super" levels.

And about that article... that was a strange thing to read. It seemed to waffle between the notion of "it will become smarter than us by reprogramming itself" and "humanity will thus screw itself over by not programming it right." It was a fun read though.

The prose "example" was the only real thing that annoyed me, however. For one, it reminded me far too much of Plato's cave analogy that, while it explained Plato's point, it doesn't do buggery in proving it (the point, that is). Plus, it seems to set up the (hypothetical) company as idiots. You'd think by the time learning A.I are so common that a "fifteen-person start up company" are able to screw over the whole planet via poor programming, there would (A) be regulations in place to prevent said screwing over or (B) the planet would already be dead.

On the bright side it was the same plot used in a fanfic I read in 2012, so it provided a good few chuckles.

But hey, the article has got a heck of a lot more sources than I've looked at, so they probably know what they are talking about and I should not question anything they say, all hail our source-filled overlords (/joking). But in all seriousness, I'm slightly disapproving of the emotive language used. In fact, a lot of stuff about A.I these days tends to use emotive language. At least, the mainstream stuff. I haven't been looking into the scientific articles and actual science-y science.

~

Overall, I stand by what I said. If an A.I gets smart enough to learn language and thereby manipulate/overthrow/oh noes, bad things for the human race... it'll probably kill off humanity very quickly to the point where nobody will notice, or it will manipulate all of humanity into working for it and thereby "satisfying all our needs with friendship," or whatever hard-line moral code its programmers gave it before letting it run free. Regardless, things will be very cool, whether in the short time before our demise or not.

Arg. Looking at all that, I sound like an optimistic conspiracy theorist. Dammit, I don't want to overthink things, I just want cool science, writing opportunities, and fun conversations. Wait. The latter requires overthinking. Ack. My brain. It is too small. Why do I find this type of thing fun. I should probably sleep instead of trying to type coherently.

The prose "example" was the only real thing that annoyed me, however. For one, it reminded me far too much of Plato's cave analogy that, while it explained Plato's point, it doesn't do buggery in proving it (the point, that is). Plus, it seems to set up the (hypothetical) company as idiots. You'd think by the time learning A.I are so common that a "fifteen-person start up company" are able to screw over the whole planet via poor programming, there would (A) be regulations in place to prevent said screwing over or (B) the planet would already be dead.

The example Tim Urban provides is actually a ripoff of the paperclip maximizer thought experiment. I think this LessWrong wiki article is better at explaining its implications; it's less of an illustration of what would actually happen, and more of an analogy for what might happen.

Of course, it would grow downright impossible for humans to work out if an A.I is mimicking or understanding as A.I technology improves, be that to "general" or "super" levels.

At that point, is there even a difference? Despite having no proof, I take for granted that everyone around me is a conscious, sapient being.

And about that article... that was a strange thing to read. It seemed to waffle between the notion of "it will become smarter than us by reprogramming itself" and "humanity will thus screw itself over by not programming it right." It was a fun read though.
...
Overall, I stand by what I said. If an A.I gets smart enough to learn language and thereby manipulate/overthrow/oh noes, bad things for the human race... it'll probably kill off humanity very quickly to the point where nobody will notice, or it will manipulate all of humanity into working for it and thereby "satisfying all our needs with friendship," or whatever hard-line moral code its programmers gave it before letting it run free. Regardless, things will be very cool, whether in the short time before our demise or not.

I feel like the point is that we should at least try to build restrictions into whatever AI may develop so that it may be Friendly.

I feel like the point is that we should at least try to build restrictions into whatever AI may develop so that it may be Friendly.

Agreed. Hopefully by that time, philosophers have ironed out a nice set of moral instructions to give it. :/

things and stuff

  I don't really know what you're all talking about, and that's a lot of stuff to read, and I don't really feel like doing that right now, so I'll just talk about something like math:

0.999... = 1

I know some people may disagree immediately, but there are many proofs out there that prove this, one of which is as follows:

x = 0.999...
10x = 9.999...
9x = 9.999... - x
9x = 9 (since x = 0.999... you can substitute 0.999... for x)
x = 1

And since x = 0.999... and x = 1, 0.999... = 1.

I think I'll go more in depth about this in a thread in the Hall of Great Discussions. If I make a thread there talking about this, I'll add the link in here: http://wintreath.com/forums/index.php?topic=4067.0