Artificial Intelligence

« first day next day → last day (1238 days later) »

lathomas64

1:33 PM

morning gentlemen!

er ladies and gentlemen(just in case)

2 hours later…

dreeves

3:23 PM

topic: is the turing test a good test of human-level ai?

1 hour later…

lathomas64

4:31 PM

depends on the human administrating the test I guess

mbq

lathomas64

@mbq reminds me of the space station story in I, Robot

mbq

Sadly it works the other way too...

lathomas64

4:49 PM

@mbq what's your take on this?

hmm no onebox for the question?

Q: Too many general, subjective, and lightweight questions

I don't want to be too much of a downer here, but look at some of these questions from the first day of the AI beta: What are some milestones and famous projects in AI? Which are some good books and articles on Machine Learning? Good blogs about Artificial Intelligence? What is the difference b...

mbq

What can I say; it confirms my predictions that this site won't get out of beta.

There are a lot of people that would like to make AI or think they do AI, and too few that are making AI.

Whatever it means ;-)

lathomas64

well thats because "Real AI is hard"

at least according to my game AI professor when explaining why most games fake it in regards to AI

mbq

The problem is that there is no more buzzy buzzword than AI. It can mean anything, so people stick AI label wherever they can.

lathomas64

well we don't even understand natural intelligence really so it makes sense that people don't know what qualifies as emulating it

mbq

We can't do a machine that communicates fluently with humans (most humans would be satisfied by this), so we must pretend that at least we are on a good route.

Yet the routes diverged.

lathomas64

5:02 PM

do you know of anyone trying to make a language that is easier for a computer ot speak in as a sort of middle ground for people and machines communicating?

mbq

C++?

lathomas64

I was thinking something more akin to lojban

a really rigid, structured constructed language with little room for ambiguity

mbq

Is it a good way? Understanding ambiguities is quite a good symptom of intelligence.

You can make a language in which the only valid statements are X="How much is 2+2?" and Y="4"

making AI would be trivial then.

Y X

lathomas64

yes in that case. but the goal would be to construct a language that could express much of what we can in others, and without ideas quite so explicitly encoded.

anyway it'd be baby steps towards communication in human languages

mbq

5:17 PM

I still think it is not a real problem.

In every language, for basic communication people use few basic expressions.

On the other hand, encoding some more complex thoughts is also almost equally hard in most languages.

lathomas64

its really to solve the issue of context sensitivity

there is a cat in the tree is a simple statement but linguisticly ambiguous. is the cat ON the tree or is there actualyl a hole in the tree that the cat is resting inside of.

making things like that explicit would allow the focus to instead be on being able to represent those complex ideas instead of struggling with basic expressions

mbq

But maybe it shouldn't be solved... such problem will be always too hard for all those present light AIs, and the real ones should learn how to do it on their own.

Ok, got to go, have a nice day.

lathomas64

you too

2 hours later…

lathomas64

7:14 PM

one of you should attach a chatbot to the room to try and fool us.

3 hours later…

lathomas64

10:22 PM

hey, how are you all?

« first day next day → last day (1238 days later) »