Home > Meeting Topics and Material, News Stories > Weekly QUEST Discussion Topics and News, 14 Feb

Weekly QUEST Discussion Topics and News, 14 Feb

QuEST ‘Valentine’s Day’ Edition Feb 14, 2014
1.) Presentations / topics associated with Alex Wissner Gross – Equation for Intelligence. “Is there an underlying mechanism for intelligence? Yes, intelligence consistently tries to maximize diversity of future options,” says Alex Wissner-Gross, PhD. What is the most intelligent way to behave? Wissner-Gross explains how the latest research findings in physics, computer science, and animal behavior suggest that the smartest actions, from the dawn of human tool use all the way up to modern business and financial strategy, are all driven by the single fundamental principle of keeping future options as open as possible. Consequently, he argues, intelligence itself may be viewed as an engine for maximizing future freedom of action. With broad implications for fields ranging from management and investing to artificial intelligence, Wissner-Gross’s message reveals a profound new connection between intelligence and freedom. Wissner-Gross is a scientist, inventor, and entrepreneur. He serves as an institute fellow at the Harvard University Institute for applied computational science and as a research affiliate at the MIT Media Laboratory. In 2007, he completed his PhD in physics at Harvard, where he researched programmable matter, ubiquitous computing, and machine learning. http://www.kurzweilai.net/ted-an-equation-for-intelligence-by-alex-wissner-gross The researchers developed asoftware engine, called Entropica, and gave it models of a number of situations in which it could demonstrate behaviors that greatly resemble intelligence. They patterned many of these exercises after classic animal intelligence tests.
2.) Georgia Tech Talk – by Luis von Ahn, CMU – Notes from watching below: essence of the discussion is the insights into the differences in humans/machines provided by the need for Captcha’s – captcha – annoying – his thesis – are you a human – reading distorted characters is easier for human – captcha is a program that generates a test that computers can’t pass – picks a random string of letters – renders into a random distorted image – can’t trust online polls – applications – free email services – spammers flood these services – so they can send millions of emails a day versus the limitations of hundreds – there are captcha sweat shops – 2.50 hour for each human – 720 captchas per hour per human – generates jobs – 1/3 cent per account – at least cost the spammer something – another hack – write a program fills out form registration can’t solve the captcha – so what the porn company does is sends that to some human looking at porn – and labels if you want to see the next image/video you have to solve this captcha – they immediately do that and the porn company gets the account done with the captcha solved for them – very neat idea – human computation – lots of things people can easily do that computers can’t do easily now – 9 billion human hours of solitaire played in 2003 – empire state building took 7 million human-hours (6.8 human solitare hours) 20 million human hours to build panama canal – wasted human computations – humanity is extremely large scale diverse set of elements – humancomputer symbiosis – making better use of human processing – image processing – labeling – still an open problem – Martha stewart – google images works by using file names and html searches – doesn’t always works – rabbits etc – don’t understand the query for ‘dog’ – accessibility – most of web is not fully accessible to visually impaired – screen images only have labels that can be read the computer can’t know what the image is so the visually impaired can’t use the web – how do we label all the images on the web – how to use humans cleverly – get people to want to label images – so they even want to pay me to label the images – esp game – as a side effect they label the images – they do it fast – could label all images on google image search in a couple of weeks – partners you don’t know and you can’t communicate with them – goal is for you and your partner need to label the images the same – best strategy type lots of words for an image – when you agree with the partner you both get points – two independent sources so good labels – told type what the other guy types – car / boy / hat / kid / can’t see the other players guesses – actual game looks like ******* link game ***** —- there are taboo words – from the game itself – two words others have agreed upon for this image – so make game more difficult and get more labels – could label all the yahoo images by 5k players in a couple of months – there is a single player version of the game – can pair up a single player with a prerecorded set of moves – so playing with someone else just not at the same time – can pair zero players – play pre-recorded with pre-recorded – can we cheat – two people log on hoping to agree on labels – test images to find if they are human – only store a label after n pairs have agreed upon it – quality of labels has been high – dog search example much better – examples are impressive – complete list also for beach image – people start to feel connection with partners – meet your soul mate through the game – why do people like the esp game – ‘anonymous intimacy’ – games with a purpose – esp game is an algorithm – input image – output set of keywords – games with a purpose – running a computation in people’s brains instead of silicon – other examples – finding objects in images another example – which pixels are the man which are the plant … – could be use for training computer vision – peekaboom – two player game – can’t communicate with partner – beginning of round image and label come from esp game – goal is for boom to get peek to say butterfly – guess what word boom trying to get to guess – then switch roles – by watching where boom clicks – there are hints – noun verb text etc – can use pings to point to particular part of image – getting lots of images where players agree combined with image segmentation algorithms get about 50% objects works very well – man/dog can find images that have both and show where they are in the image – verbosity another game – common sense fact – water quenches thirst / cars usually have four wheels – computers don’t have them – collect them and put into a computer – lots of efforts trying to do – failed – put into a game – milk facts are white / liquid cereal eaten with it has lactose — two player word guessing game – narrator and guesser – narrator gets a word and trying to get the guesser to say that word – common sense facts are sent to player two and guess the word – player 2 is verifying the facts – asymmetric verification game – input to player 1 output to player 2 then player 2 has to guess the input to player 1 — both player doing something slightly different – symmetric verification is the esp game – symmetric games constraint is number of outputs per input – asymmetric constrain is number of inputs that yield the same – lots of power in clever way to use human computational cycles – matrix – instead of using humans as source of power – keep us around to solve problems computers can’t solve
3.) Cognitive Psychology 63 (2011) 107–140 – Intuition reason and metacognition – by Thompson, Turner and Pennycook – Dual Process Theories (DPT) of reasoning posit that judgments are mediated by both fast, automatic processes and more deliberate, analytic ones. A critical, but unanswered question concerns the issue of monitoring and control: When do reasoners rely on the first, intuitive output and when do they engage more effortful thinking? We hypothesised that initial, intuitive answers are accompanied by a metacognitive experience, called the Feeling of Rightness (FOR), which can signal when additional analysis is needed. In separate experiments, reasoners completed one of four tasks: conditional reasoning (N = 60), a three-term variant of conditional reasoning (N = 48), problems used to measure base rate neglect (N = 128), or a syllogistic reasoning task (N = 64). For each task, participants were instructed to provide an initial, intuitive response to the problem along with an assessment of the rightness of that answer (FOR). They were then allowed as much time as needed to reconsider their initial answer and provide a final answer. In each experiment, we observed a robust relationship between the FOR and two measures of analytic thinking: low FOR was associated with longer rethinking times and an increased probability of answer change. In turn, FOR judgments were consistently predicted by the fluency with which the initial answer was produced, providing a link to the wider literature on metamemory. These data support a model in which a metacognitive judgment about a first, initial model determines the extent of analytic engagement.



  1. No comments yet.
  1. No trackbacks yet.

Leave a Reply

Fill in your details below or click an icon to log in:

WordPress.com Logo

You are commenting using your WordPress.com account. Log Out /  Change )

Google+ photo

You are commenting using your Google+ account. Log Out /  Change )

Twitter picture

You are commenting using your Twitter account. Log Out /  Change )

Facebook photo

You are commenting using your Facebook account. Log Out /  Change )


Connecting to %s

%d bloggers like this: