Artificial Intelligence Explained through Deep Learning

 

In our articles at AnRcloud.com we have quoted numerous personas in the technological and scientific world who pronounced themselves on the dangers of artificial intelligence and the need of it being controlled. The ethics and the philosophical frame around this question are still creating the buzz without bringing any clear answer so far.

While observing these discussions, it is crucial for everyone to understand the principles underlying artificial intelligence technology. Moreover, it is gaining more power and new features daily. Babak Hodjat explains the technologies and terminology that are used when talking about artificial intelligence in his recent article for Wired[i]. Among the main technologies Hodjat mentions Big Data, Machine Learning, i.e. “algorithms that can learn from data”[ii] and Deep Learning – a class in machine learning algorithms that allow “learning with deep architectures for signal and information processing”[iii]. Deep Learning is compared to the human brain functioning, as it uses neural networks built up in numerous layers of abstraction. It is specifically deep learning algorithms that have gone through a major breakthrough with the Moore’s law and the development of calculation technologies. They are used when analyzing big volumes of data for computer vision and linguistic technologies – speech recognition and natural language processing. It should surely have its future in machine translation, as well. While being inspired by human brain, the technology is still rather different – for instance in image recognition applications deep learning systems process the picture in layers. “They start with edges and then get more abstract with each layer, focusing on faces or perhaps whole objects”[iv]. The complexity of the process requires resources from extremely powerful computers and make it available mainly for gigantic companies like Microsoft[v], Facebook[vi] and Google[vii]. Mobile Internet and advances in processing power made it possible for smartphones now to run deep learning algorithms using remote servers and even locally. The tests have been run successfully in order to “get the phone to determine people’s emotions or identities from recordings”[viii]. The performance of deep learning algorithms is still below the one of humans, but it starts to gain power. These algorithms require as well big volumes of structured data to be analyzed and learn from. At the same time, the technology is used now in applications and as a help in decision making. There is a possibility for even greater development with faster calculation and the possibility of structuring data in a better way for finding newer patterns for intelligent and enhanced decision making.

Certain examples prone that deep learning algorithms can outperform human brain in certain specific tasks[ix]. This still remains to prove. However the potential of data analysis and pattern recognition when performing targeted marketing or financial analysis is huge.

With all this Deep Learning seems to be a major technology that could bring artificial intelligence to the next level and more common use in daily applications. However, it should not be overrated – as explained by Yann LeCun, head of the Artificial Intelligence Research Lab at Facebook in his interview for IEEE Spectrum: “It will be part of the solution… that will look like a very large and complicated neural net”[x].

As discussed in other AnRCloud articles, artificial intelligence, being progress and technology development, remains a complex phenomenon and a sum of several technologies working together for a better and a more efficient result.

 


 

[i] Myth Busting Artificial Intelligence by Babak Hodjat for WIRED, February 19, 2015, online http://www.wired.com/2015/02/myth-busting-artificial-intelligence, accessed February 19, 2015

[ii] Myth Busting Artificial Intelligence by Babak Hodjat for WIRED, February 19, 2015, online http://www.wired.com/2015/02/myth-busting-artificial-intelligence, accessed February 19, 2015

[iii] Deep Learning Methods and Applications by Li Deng and Dong Yu, 2013, online http://research.microsoft.com/pubs/209355/DeepLearning-NowPublishing-Vol7-SIG-039.pdf, accessed February 19, 2015

[iv] Ahna Girshick quoted by Derrick Harris in Why deep learning is at least inspired by biology, if not the brain for Gigaom, February 15, 2015, online https://gigaom.com/2015/02/14/why-deep-learning-is-at-least-inspired-by-biology-if-not-the-brain, accessed on February 18, 2015

[v] Microsoft researchers say their newest deep learning system beats humans — and Google by Jordan Novet for VentureBeat, February 9, 2015, online http://venturebeat.com/2015/02/09/microsoft-researchers-say-their-newest-deep-learning-system-beats-humans-and-google, accessed on February 19, 2019

[vi] Why deep learning is at least inspired by biology, if not the brain by Derrrick Harris for Gigaom, February 15, 2015, online https://gigaom.com/2015/02/14/why-deep-learning-is-at-least-inspired-by-biology-if-not-the-brain, accessed on February 18, 2015

[vii] Google shows its deep-learning systems are doing just fine, thank you very much by Jordan Novet for VentureBeat, September 6, 2014, online http://venturebeat.com/2014/09/06/google-shows-its-deep-learning-systems-are-doing-just-fine-thank-you-very-much, accessed on February 19, 2019

[viii] Deep Learning Squeezed Onto a Phone by Rachel Metz  for MIT Technology Review, February 9, 2015, online http://www.technologyreview.com/news/534736/deep-learning-squeezed-onto-a-phone, accessed on February 19, 2015

[ix] Delving Deep into Rectifiers: Surpassing Human-Level Performance on ImageNet Classification by Kaiming He, Xiangyu Zhang, Shaoqing Ren and Jian Sun for Microsoft Research, February 6, 2015, online http://arxiv.org/pdf/1502.01852v1.pdf, accessed on February 19, 2015

[x] Facebook AI Director Yann LeCun on His Quest to Unleash Deep Learning and Make Machines Smarter by Lee Gomes for IEEE Spectrum, February 18, 2015, online http://spectrum.ieee.org/automaton/robotics/artificial-intelligence/facebook-ai-director-yann-lecun-on-deep-learning#qaTopicSeven, accessed on February 19, 2015