Should we let AI take over?

Artificial Intelligence – “The capability of a machine to imitate intelligent human behaviour”

AI has been around for decades, certainly the theory has anyway. The field of artificial intelligence was founded back in 1956 but it has always been seen as something “for the future” and an exciting movie storyline. However, over the last few years’ things have changed and AI and Machine Learning (ML) have gained some serious traction. Many industries and enterprises are investing heavily in AI research and many are actively using the technology now.

Looking back a few years, you may remember in 1997 when IBMs’ Deep Blue computer was able to beat the then current chess world champion Gary Kasparov. This was seen as AI finally showing its ability to outsmart a human. However, the Deep Blue was really a computer that was programmed to play chess and understand all the rules and moves. It was not an example of Machine Learning and as it was pre-programmed, you could argue that there was no learning involved or indeed “intelligence”. Not to do the achievement a dis-service, it was still very impressive. A better example of Machine Learning and AI came in 2015 when Googles’ DeepMind AlphaGo defeated a European champion at the board game Go. AlphaGo actually won 5:0. This was the first time that a machine had defeated a human player in a professional game. Then in 2016 it repeated the feat by defeating the best player in the world.

The difference between this and IBMs’ achievement was the fact that Deep Blue had used a hard-coded function written by a human and AlphaGo used self-learnt strategy to play the game. This showed that AI and ML have come on leaps and bounds in just 20 years.

Today modern day AI uses what’s called a Neural network which is essentially a computer system that has been modelled on the human brain. The following are current examples of AI being used today:

  • Banking – RBS and WorldPay have been using AI for years to detect fraudulent transaction patterns. Many believe that AI could have prevented the highly publicised Target breach.
  • Social Media – Facebook is using AI to try and identify potential users who may be at risk of self-harm or even suicide. I’m pretty sure they use it for all sorts of other reasons but this is one of the ones that has been publicised.
  • Telecoms – Level 3 collect threat information on 50 billion events a day, tracking jitter, latency, loss and all of this is being analysed to monitor health of their network.
  • Cybersecurity – Fortinet use advanced algorithms to detect and predict nature of malicious code.
  • UK Justice system – By using AI in the prison system for cellmate matching, reoffending can be prevented.
  • Public Transport – Trials of adaptive scheduling system have shown efficiency gains of 38%.
  • NHS – Trials being used for theatre resource booking improving efficiency of the theatre resources.
  • Media – Netflix are using adaptive video encoding for mobile users, improving the mobile video experience for customers.

How Can Artificial Intelligence Affect the Telecommunications Industry?

AI has a huge amount of potential for the Telecoms industry. It is already being used, to a limited degree, by the likes of Level3 and Fortinet as mentioned above, but where else could it be used? The possibilities are endless but here are a few areas where improvements could be made with the use of AI and Machine Learning:

  • Network Management Systems – Fault identification based on network patterns and performance
  • Software Defined Networking – Machine Learning and intelligence can become part of the network rather than an add on. SDN enables this integration and identification of patterns and anomalies. The network becomes more intelligent.
  • Proactive Network – Machines can observe patterns in data and send instructions to the network on how to respond.
  • Network Sizing – AI can analyse data and patterns and make predictions on network requirements such as bandwidth, QoS profiles, compute and storage requirements
  • Warehouses already use forms of AI to replace faulty hardware – is it unrealistic to think that AI could one day be used to replace a failed blade server or faulty router in a data centre?

Artificial Intelligence is no longer “one for the future” it is happening right now. Although it may still be in its infancy the potential benefits are massive when delivered correctly. There will of course be bumps along the way and we’ve already witnessed some epic fails but don’t let that put you off – it’s time to embrace the machines!

Innovation Hub

Check out some other blog