Share:
What is AI?
July 8, 2019
Future of Tech
The Summary:
Artificial intelligence (AI) is a term often thrown around and has come to be seen as the arch-enemy of many of the jobs we have today. But what does it actually mean? The incredibly broad definition given to us in 1955 covers “every aspect of learning or any other feature of intelligence” that is simulated by a machine, however, in today’s world we tend to think of artificial intelligence as a combination of both complex reasoning and the ability to learn.
A good example of this is Alpha Go, an Artificial Intelligence that taught itself to play the ancient Chinese game of Go and ultimately beat the world champion in 2017. Unlike Chess, it is impossible for a computer programme to model all possible future moves, so Alpha Go needed to develop its own strategy based on the current board situation and its learned experiences - much like a human player would.
Artificial Intelligence is a technology that is already well in use but we haven’t even scratched the surface of its capability. In the future, it will almost certainly have an enormous impact on every aspect of our world as we know it.
The question we face now is how we ensure that this incredibly disruptive technology is harnessed for social good. This is something that is certainly not taken lightly. When Deepmind (the creators of AlphaGo) sold to Google the founders insisted that an ethics board be established.
“AI is one of the most important things humanity is working on. It is more profound than, I dunno, electricity or fire.” - Google CEO, Sundar Pichai
The Promise:
Virtual Assistants
Virtual assistants are already commonplace in mobile phones and home speakers and we are seeing the rise of customer service chatbots built into company websites. However, the next evolution of virtual assistance is set to take this to a whole new level. Soul Machines is a NZ-based company developing hyper-realistic digital humans, that react and respond with emotion. These AI-driven humans are creating a completely unique way of interacting with technology.
Robots
The advancement of humanoid robots is well underway and with the continued advancement of AI we will see these becoming far more versatile and useful in our everyday lives. When we look beyond humanoid robots the field is even more diverse and exciting including autonomous vehicles and search and rescue robots ideal for entering hazardous environments. Task-specific robots would also be ideal for performing dangerous tasks in fields like mining,forestry and disasters.
Expert decision making
Beyond winning board games, artificial intelligence is having impressive success in far more important roles that have always been the domain of human experts. The United Kingdom's National Health Service (NHS) already has adopted AI to assist in the diagnosis of patients e.g. by helping read scan data and they see the technology as “revolutionising outpatient services”. The financial world has also big adoption of the technology with AI helping to assess risk, detect fraud and engage in automatic stock trading. In 2018, NZ began allowing the provision of digital financial advice with KiwiWealth launching the first of these services.
The Breakdown:
When we think of usual computer programming it is highly prescriptive. “In this situation do this” or “in that situation do that” etc. In contrast, artificial intelligence algorithms don’t tell the system what action it should take but instead equip it with tools to find it’s own answer.
A good example is a robotic dog. With standard programming, the engineers would work to figure out the best sequence of motions to make the dog walk properly. In contrast, an artificial intelligence algorithm wouldn’t give the dog precise instructions but would instead give points based on how far the dog manages to walk. The dog would then perform multiple trials, performing random sequences of movement. By chance, some of these attempts will enable the dog to be somewhat successful in walking. The better the dog does the more points it will score, so movement sequences that score higher will be favoured in future attempts. This allows the dog to eventually find a sequence of movements that allow it to walk effectively.
This approach to artificial intelligence produces results often better than what could be normally conceived, particularly where the task is complex. These algorithms are known as neural networks and based on similar principles to how an organic brain functions.
Artificial intelligence is often categorised as Weak AI or Strong AI. Weak AI focused on achieving a very specific task and artificial intelligence is optimised to achieve that. Examples of Weak AI are face/object recognition and natural language processing (i.e. understanding normal conversation). Strong AI is a whole step ahead and could be considered the holy grail of AI. Strong AI is not focused on achieving any one task, it is about achieving ‘general intelligence’ capable of doing almost anything in much the same way as humans. Google’s Deep mind is one of the big players in this space and although we are still a long way off true Strong AI, they recently published a paper in Nature on a system capable of mastering a wide variety of different computer games.
As human beings, we are very much defined by our ability to adapt and learn new things. In 10-20 years our digital creations may well be doing this too.
Useful Links:
Share:
