Incoming Transmission

Say hello to other friends of the Royal Navy Airship Service!

Incoming Transmission

Postby ASoriano » Wed Oct 27, 2010 7:47 pm

Greetings all,

My name's Andrés. I've been reading the Flying Cloud, R-505 for a while now. I decided to join the forums to vote for the good Captain in the most popular character poll. There's not much to know about me, I was a graduate student in quantitative psychology and now I work as a financial advisor/consultant. My passions are mathematics, metallurgy, small arms (handguns & swords mostly), Scotch whiskeys, rum, role-playing games, computers, and my significant other. Feel free to ask any other questions.
ASoriano
 
Posts: 24
Joined: Wed Oct 27, 2010 4:48 am

Re: Incoming Transmission

Postby Kona » Thu Oct 28, 2010 1:03 am

Copy that TX, Andrés, and welcome aboard! I beg to differ; there seems to be a LOT to know about you! To begin, is “quantitative psychology” a study in no-brainers? Hmm… small arms, whiskey and rum… a case for the Bureau of Alcohol, Tobacco and Firearms? Actually, what are they all doing together? I called them up one time and asked, “So what wine goes with an AK-47?” Hey, they were helpful! They asked “What are you smoking?”

Your psych studies will come in handy here, Andrés! Grab a berth and join us on the mess deck for coffee and scuttlebutt!
The tide is out; please leave a message.
Kona
 
Posts: 264
Joined: Wed Jan 06, 2010 5:54 pm
Location: Enron-by-the-Sea

Re: Incoming Transmission

Postby ASoriano » Thu Oct 28, 2010 7:35 pm

Well quantitative psychology is a little known field in psychology. Basically it breaks down into two subfields, psychometrics and my own specialty which is mathematical psychology. Psychometrics is focused on developing new statistical procedures for psychological problems, most commonly scaling and testing. A very popular theory running around right now is item response theory, though my school had an intense focus on factor analysis as well. Then mathematical psychology is the attempt to fit a mathematical model to cognitive processes. This runs the gamut from the diffusion model of two-choice decision making to two-choice Weiner models of confidence to neural networks. I had an intense interest in neural networks, but I lacked the programming skill to implement it so I stuck with playing around with Bayesian hierarchical models of two-choice decision tasks.

Note: I do not handle guns while drinking or drink while handling guns. In fact, nowadays I hardly ever do either.
ASoriano
 
Posts: 24
Joined: Wed Oct 27, 2010 4:48 am

Re: Incoming Transmission

Postby peterh » Thu Oct 28, 2010 8:54 pm

OK. Now I feel stupid.

:mrgreen:

I wonder where quantitative psychology and AI touch.
Don't panic... we're on the Titanic.
peterh
 
Posts: 106
Joined: Fri Jan 08, 2010 10:38 am
Location: down here

Re: Incoming Transmission

Postby ASoriano » Thu Oct 28, 2010 9:29 pm

peterh wrote:I wonder where quantitative psychology and AI touch.


That's actually a very interesting philosophical question. AI technically falls under mathematical psychology as you're applying a mathematical model (neural networks/AI) to a cognitive problem (voice/letter recognition, NPC behaviour in video games, etc.). When I started applying neural network models to letter recognition I noticed one thing. While I could make my program learn to recognize letters and eventually words, I had no clue how it was doing it except on a very abstract sense. Essentially I had a black box of nodes that through a series of rules and probabilities learned, but I couldn't express it as a holistic model of the human mind doing the same task.

Now, if quantitative psychology strives to apply the most accurate mathematical model to a cognitive process while still being able to explain what each variable means in a meaningful way, then is AI really doing that? If my model has a huge black box that I can't explain, then is it really furthering the science? Do we really need to explain it down to the last detail or just make it able to duplicate the human process?

In the end you open up a huge can of worms. Some people argue that AI isn't truly trying to model the human mind so much as imitate it in the same sense as a marble statue isn't a human even if the David looks like one. Other people say that if the end result is similar enough, then the process is irrelevant. It all depends on how you look at things.
ASoriano
 
Posts: 24
Joined: Wed Oct 27, 2010 4:48 am

Re: Incoming Transmission

Postby peterh » Thu Oct 28, 2010 10:14 pm

Funnily enough, I was *specifically* thinking of OCR (Optical Character Recognition) when I posited the question.

When we read, we do not actually recognise characters - we recognise patterns. If smoeone treis to trohw us a cuvreball, we can usually figure out pretty well what it is supposed to say. As soon as we get the curveball, we also recognise the word "throw", because they belong together in that expression.
Really curveball I want if a would you throw to (if I would want to really throw you a curveball), I would have to upset the pattern, not the individual characters.

We also use the context to help us read text. For instance, if we are discussing a particular web comic, in which teh size of a ladies' bosom plays an integral role in the storyline (in a much saner way than you would imagine, by the way), I could use the phrase "Faye's physical characteristits". Most anyone reading my post would recognize the typo for what it is. OCR would correct it.
And people who know which comic I am on about, would also recognise the other typo in the above paragraph as intentional.

So, yes, we use our sense of humour as well. We have a sense of humour because we like that. AI has no reason to have a sense of humour.
We can associate, using a massive set of a priori ideas and mental images, and we can do interpretive reading, because we understand the underlying idea.
Don't panic... we're on the Titanic.
peterh
 
Posts: 106
Joined: Fri Jan 08, 2010 10:38 am
Location: down here

Re: Incoming Transmission

Postby Mutabilis » Thu Oct 28, 2010 11:52 pm

i want a pintsize. who, incidentally, makes questionable content out of your post.
Mutabilis
 
Posts: 189
Joined: Wed Aug 25, 2010 12:35 pm

Re: Incoming Transmission

Postby peterh » Fri Oct 29, 2010 5:17 am

That would be light work for him. Pintsize makes questionable content out of anything.
The Singularity will have to wait, though.
But we digress.
Don't panic... we're on the Titanic.
peterh
 
Posts: 106
Joined: Fri Jan 08, 2010 10:38 am
Location: down here


Return to Introductions

Who is online

Users browsing this forum: No registered users and 1 guest

cron