Implications of an Improved Lip-Reading Software

Researchers at Oxford University have recently developed a new program called LipNet which uses video frames to put together different movements of the mouth to make sentences which it then outputs as words. While this seems like a fairly simple way of doing lip reading, it has actually been amazingly successful. This new system of reading lips has reached an astonishing 93% accuracy
which trumps the previous best computer rating of 79% and smashes the human expert level of 52%.

The importance of this new software has implications way beyond just another creative example of a computer taking an input and returning an output. The social opportunities are massive for this type of software. From a human welfare perspective, over 360 million people worldwide struggle with hearing problems and a program such as this could drastically improve their lives. Additionally, commercially this software could be incredibly helpful anywhere from reading opposing teams lips in sports, reading the lips of people speaking far away without a microphone, or reading lips in a noisy environment. There could also be national defense implications with being able to read peoples lips through video footage without sound or by spying on people's conversations from a distance with binoculars. Regardless of how this could eventually be used, the possibilities are endless.

The biggest concern however is privacy. We live in a world where almost everything you do is being recorded whether by phones, street cameras, or even satellite surveillance. If this lip-reading software is used by the wrong groups it could be a massive breach of privacy to people whose conversations could be intruded on by outside sources. This will most likely become one of the developing hot topics, like drone use and cybersecurity, that will require new laws and regulations to keep under control.


Resources:
1. http://www.digitaltrends.com/cool-tech/lipreading-artificial-intelligence/
2. http://www.signlanguagenyc.com/wp-content/uploads/2016/03/lip-reading-asl-cart-services-nyc-02-219x300.jpg
3. http://www.notbored.org/watching-all-day.jpg
4. http://wwwhatsnew.com/wp-content/uploads/2016/11/LipNet-730x480.jpg

MogIA Accurately Predicts Election

In light of current events in the United States, I thought that I would take this week to write about what is on everybody's minds-- the Election. I don't want to get political in what is a very sensitive time so I am going to try and be as unbiased as possible throughout this post.

Leading up to the election Hillary Clinton seemed poised to win the Presidency according to many polls and news sources. However, to the surprise of many, Donald Trump was able to win several key swing states in order to become the next President of the United States. Who could have ever predicted this? Apparently an AI system called MogIA.

Business Insider posted an article leading up to the election about MogIA and how it has successfully predicted the the last three Presidential elections as well as both of this year's primaries. How is this possible you may ask? MogIA uses data from sources such as Google, YouTube, and Twitter to analyze how well a candidate is doing. It then updates itself over time in order time by learning from the online environment.

While MogIA is definitely not perfect it has yet to be wrong when predicting presidential elections. MogIA is an example of yet another incredible algorithm that can analyze an event and both unbiasedly and accurately predict outcomes better than humans.

Sources:
1. http://www.businessinsider.com.au/artificial-intelligence-trump-win-2016-10
2. http://zdnet1.cbsistatic.com/hub/i/r/2016/11/09/25fd2587-0c5c-4afc-bcdf-96539d83d420/thumbnail/770x578/feb141cb6013a9120dad232d0b7d513b/trump-cyber-crisis.jpg
3. http://realiran.org/wp-content/uploads/2016/08/US-elections-2016-638871.jpg

Sorting Data

Have you ever wondered what the most efficient way to sort data is? Say you have 10 randomly generated numbers and want to sort them from highest to lowest. I think we could all figure out a way to do that in java without it taking too long but it becomes an issue when dealing with big data. As a result there are over 15 different methods for sorting data. This video below of data being sorted is actually oddly satisfying.

Additionally a new sorting algorithm was created to solve the "cake-cutting" problem. This is a problem in which you are determining how to cut a cake fairly amongst multiple people given their different preferences. Think about a holiday party with a fruit cake. Every member of the family has a different preference for how big of a slice and which fruits they want in their slice. This problem tackles the problem of fairness which was previously viewed as unsolvable mathematically. However, recently a researcher was able to solve this problem for a group of anywhere between 3 to 203 people. 

The most notable part of this solution is that it shows that previously "unsolvable" problems can actually be solvable. Mathematicians should now be inspired to give those impossible problems a look again. Also, if you haven't already, please watch the data sorting video above it's actually really cool.  

Resources:

1. http://www.digitaltrends.com/cool-tech/sorting-algorithms-video/

2. http://www.digitaltrends.com/computing/cake-cutting-algorithm/

3. http://d2gk7xgygi98cy.cloudfront.net/1296-3-large.jpg