Deep Learning 

Deep Learning is a new field in Artificial Intelligence, it consists of a layered set up by which computers should be able to somewhat mimic humans brains activity and although completely challenging human brains to it's full capabilities is still far from our reach massive advances have been made by companies such as Google and Facebook. I decided to focus on Google's technological improvement in image recognition. The shareholders in this case would be Google but also anyone choosing to utilize this software in the future.

Social Impacts

In it's essence this new computer vision software is designed to analyse a picture and be able to summarize it in a few words accurately, this software is still a prototype but has efficiently recognized real life pictures such as  ''A group of young people playing a game of frisbee'' and ''A herd of elephants walking across a dry grass of field'' these summaries were exactly what the pictures represented. The way it works is that two neural networks specialized in different task have been connected, the first network focuses on processing an image into mathematical representation of its content, the following network has been programmed to generate English sentences based on a automated translation software using those mathematical representations, in other words the first network will ''look'' at the image and report what it ''sees'' while the second will transform the description into a sentence. An advantage would be the fact that scene-describing software could potentially aid visually impaired people while the negative to this software or rather its limitation is that although a large data-based of hand drawn labelled images has been used to train the software there is a lack of natural scene pictures with labeling which limits the training process of the software towards natural scenario pictures.

Ethical Issues

The people responsible for the developments in Deep Learning are all of the researches working towards improvement although in this software Google is not only responsible it is also accountable for any misuses. In addition to aid those whom are visually impaired this software can make image searches much more efficient as it can order them specifically by what they contain and not only  by what their tags or titles are. The potential misleading factor in this software would be its ability to differentiate whether a photo had been Photoshop or not as it could influence the results in a search, moreover if used in daily lives by people the software would have to be perfected in order to differentiate living and non-living objects quickly and efficiently. Lastly for years the image recognition softwares have been scoring 40s and 50s out of 100 in a scenario recognition test at which humans score amongst the 70s, with these new improvements in the two neural network software Google has accomplished a score in the 60s reaching the closest any software has to mimicking the human brain.

Link:

http://www.technologyreview.com/news/532666/googles-brain-inspired-software-describes-what-it-sees-in-complex-images/