ShanghAI Lectures 2012

ShanghAI Lectures 2012

Interest Group Board Entries

Daniel Rojas's picture
4 years 10 months ago
 

Reading chapter 5.6 of HTB (Matching Brain and Body Dynamics) I stumbled across the name Daniel Wolpert, and thought it was familiar! Saw him on TED.com a while ago...
It's a great talk, I highly recommend watching it!
http://www.ted.com/talks/daniel_wolpert_the_real_reason_for_brains.html


Nathan Labhart's picture
4 years 10 months ago
 

Thank you for the link – it's a great TED talk, indeed!


Jonathan Nagel's picture
4 years 10 months ago
 

Hello
I don't know where else to ask, so I do it here.
I have to reinstall the webot-application to be able to solve the exercise 2, but I can't open the webpage cyberbotics.com. Don't know if it is offline, but it doesn't wok on different computers and browsers. Has anyone an idea?


Yiwen Shen's picture
4 years 10 months ago
 

I also can't open the webpage, but my webots works well. why do you need reinstall the webots? because of the license?


Nadia Figueroa's picture
4 years 10 months ago
 

I am having the same problem as Jonathan.


Marco Unternährer's picture
4 years 10 months ago
 

Hey Jonathan,
I still have the the windows installer, you find it here:
https://docs.google.com/open?id=0B6t56IB_eb6hNzFsbVdhT0VzbVU

cheers, marco


Nadia Figueroa's picture
4 years 10 months ago
 

thanks!


Jonathan Nagel's picture
4 years 10 months ago
 

Need the installer for Mac, because on my Windows-PC it is not working properly because of my Graphic-Card... but thanks anyway for the link


Ahmed Naguib's picture
4 years 10 months ago
 

Hello,

I was reading the text book "how the body shapes the way we think" and i came across this sentence in chapter 2:
"And to our knowledge, there are currently no artificial systems capable of dealing
with this kind of knowledge in a flexible and adaptive way, because it is
not propositional and thus hard to formalize in a symbolic system." (page 33).

Prof. Rolf was arguing in this part that the classical approach fails to describe or understand intelligence and gave an example by how hard (impossible) for an abstract algorithm to understand natural language and extract statements/info out of it.

Well, i kindly disagree, IBM started a project called "DeepQA" in 2007. the purpose of that project was to create an algorithm that can understand natural language as it is (with no cheating nor exploits).
To evaluate their results, they called their product "Watson" and aimed for watson to participate, compete and become a champion in Jeopardy, a popular open-field Q/A game in the US.
Not only they were able to succeed, they actually were able to compete and win against national champions using only their algorithm and a huge database of textbooks & wiki articles stored offline in an 8 refrigerator-sized database.

Please, kindly have a look at this presentation:
[see 1]

And so, i though that this argument is not valid anymore


Nathan Labhart's picture
4 years 10 months ago
 

Hi, this is a very good topic for discussion! How can we know whether Watson actually "understands" natural language? We get into "Chinese Room" territory here :-) I agree that answer extraction is possible to a certain extent, but what about "information" that is not written in its database, such as feelings?
I will forward your comment to Rolf Pfeifer and ask him to discuss this!


Ahmed Naguib's picture
4 years 10 months ago
 

Hi Prof. Labhart.

I know it's a gray area, however, in order to prove or disprove that watson understand natural language, we have to have a clear definition of what "Understanding natural language" really means!

i have though about it a little. truth is, our language is far from being perfect. i can for example, say: "small pizza italy". to someone who is hungry, he'll understand that i had eaten an italian pizza. for someone from italy, he might understand that i'm saying that italian pizza is too small.
what i wanted to say is that our understanding of natural language is composed of 2 components, first, we try to extract keywords from a sentence, and then, we try to link these keywords in a meaningful context. this linking rely first on grammar. however, when grammar, or even keywords fails to deliver, we start to use our own history and background to make as much sense as possible using these keywords.

in that sense, Watson is actually understanding natural language. as he extract keywords and try to make sense of them by sentence grammar & a huge database.

This is, in fact, how we came to understand Hieroglyphic. we started by extracting some symbols and matching them with other contexts, until eventually, we were able to associate a particular meaning to each symbol.

even in our feelings, i don't think that newly born babies can understand complex feelings, like depression, anger, hatred, ... etc. it's accumulated from incoming knowledge, media & environmental interaction.

However, i still admit that some Instinctive feelings, like fear & discomfort, spawn in new beings without prior knowledge or training.

As you kindly said, it's a gray area, since we can not take a bird view.

PS: similar topic can be discussed regarding SIRI, Apple voice synthesizer, who can, arguably synthesize voice and "understand" it.


Creative Commons License