Deep Learning: What Happens When Computers Can Learn Faster Than Humans?

Deep Learning: What Happens When Computers Can Learn Like Humans?

Deep Learning: What Happens When Computers Can Learn Like Humans?

Deep Learning is an unimaginable innovation that might be soon disrupting various industries. Deep learning is a technique by which machines are able to learn, and in particular, recognise voice and images more effectively than they do at the current time. According to the website www.deaplearning.net, deep learning is a new area of Machine Learning research, which has been introduced with the objective of moving Machine Learning closer to one of its original goals: Artificial Intelligence.

This provides both good news and bad. On the one hand the technology offers significant opportunities for advancement, but on the other hand it is possible that jobs will be lost as a result. Deep Learning is starting to progress into industries as varied as fashion and finance and it promises to disrupt established industries. Will Knight (2015) wrote recently for MIT Technology Review about the amazing developments of deep learning:

“(Deep Learning) involves applying layers of calculation to data, such as sounds or images, to recognise key features and similarities.”

Examples of this are given to be when a person’s face is observed from different angles, and being able to recognise that it is still the same face, or alternatively understanding when a word is the same, even if it is being spoken in different accents.

Deep learning is already out there in the market. For example, Google has utilised deep learning for the purposes of voice recognition on telephones. And who hasn’t experienced the identification of friends on Facebook in photos that are uploaded? The technology does not always get it right, but it is pretty good and extremely fast. These are not the only practical applications of deep learning that are being put to use by industry. Another example at eBay is an attempt to put products into categories automatically based on the types of photos that users post of the items that they want to sell. However, this is hard to do and some products have not been tagged effectively as a result. Deep Learning promises to revolutionise everything, which can be both wonderful and terrifying, as Technologist Jeremy Howard shares in this TED talk, that narrates some of the most fascinating new developments in the fast-moving field of deep learning.

As seen in the talk, deep learning is being put to meaningful uses as well, such as health. One example cited by Will Knight is that of Emmanuel Rios Velazquez who is researching in Boston at the Dana Farber Cancer Institute. In this case, Velazquez is looking into the extent to which deep learning could be used to understand possible patient outcomes based on images that have been taken of his/her cancer. Another potential use is being tested out by Olexandr Isayev, a research scientist that is based at the University of North Caroline at Chapel Hill. In this case computers and algorithms are being used to recognise the potential of drug molecules to be helpful, scanning through millions and millions of possibilities in the process. This has not been as successful as hoped to date, but it nonetheless illustrates what could be possible with greater optimisation and development in the future.

Worthy uses aside, deep learning is already starting to be applied in many other areas. A start-up known as MetaMind is one such company that is focused on the use of deep learning. The company was set up by Richard Socher. Socher studied for his PhD at Stanford and worked with Andrew Ng and Chris Manning. He has received awards and accolades for his research and work into deep learning and its potential. The company helps organisations or people to do activities such as sort datasets in an automated manner based on a particular classifier. MetaMind is helping people working in a wide range of different industries to classify products. Examples include those working in fashion, cars, houses, satellite images and many others.

Other examples of uses of deep learning were discussed and analysed at a recent deep learning event that was held in Boston. At this event experts in the field and others including entrepreneurs, researchers and engineers all came together to review progress to date in this technology. They also reviewed the varied possibility for applications such as the use of deep learning in advertising, the finance industry and the medical world. One enterprising individual reportedly had used deep learning on hedge funds to try to understand when major shifts were likely to happen, such as a big drop in the value of a country’s currency. Another fascinating application of this technology was reported to be the use of deep learning by insurance companies, and one such company based in the USA was trying to see how it could be used to pick out fraudulent claims when reviewing all claims. The bottom line is that deep learning isn’t just all about telling your phone what to do or tagging your buddies on Facebook. It has the potential to bring about real societal change.

The bottom line is that deep learning isn’t just all about telling your phone what to do or tagging your buddies on Facebook. It has the potential to bring about real societal change. The downside of deep learning though, is a radical transformation of the labour market, which can bring social unrest. Are you ready for the machine learning revolution ?