I was talking to a bot. And I didn’t realise it. I thought it was a real person.

We all experience these defining moments, where there was a “before” and there is an “after”. This was such a moment for me, when I realised that AI has gone very far indeed.

Of course, we all realise that AI has taken more space in our lives. But 10 years ago, I would not have imagined that I could mistake a robot for a real person. Artificial Intelligence is not just growing rapidly, it is actually exploding.

 

 

I wanted to write about the reasons and the consequences of this explosion, but then read “The 3 breakthroughs that have finally unleashed AI on the world“, written by Kevin Kelly in Wired. This is an excellent, insightful article which you should read and I have summarised (to the extent that I can do justice to the writer) his 3 reasons why AI is growing exponentially:

1) Cheap computing power. Not only computers keep getting cheaper and more powerful thanks to Moore’s law, but parallel GPUs and the cloud are making computing power extremely powerful and cheap.
2) Digitalisation of all contents. AI needs content to learn, and this content is now immediately available in digital format and in massive quantity
3) Better algorithms and methodologies. Just like any scientific field, AI is reaching a maturity point thanks to the results of all previous research.

I totally share the view of Kevin Kelly on these 3 points, and would add 3 other ones which I think are also very important:

4) “Minimum Viable Product” has been achieved in multiple areas of AI, and can be used as building blocks for other AI applications. Remember 10 years go, when Nuance and Dragon were at the forefront of speech recognition of that time, you had to repeat 10 times the same sentence to get a result. Today, speech recognition is almost a commodity, and is a building block for Natural Language Processing, machine translation, etc.

5) Programming is much easier. Whereas 10 years ago, you had to be a hardcore computer scientist to develop AI, today an average programmer can build Artificial Intelligence applications thanks to all the available libraries that simplify “difficult” functions. 30 years ago you had to be a geek to make your Apple IIe play the 9th Symphony in 8-bits, today a 9-year old can program the same – and get much better results – on Scratch. This is what’s happening for programming in AI.

6) And finally, access to knowledge is widespread, thanks to the Internet, MOOCs, forums, blogs, etc. 20 years ago, if you were a PhD in computer science at a top university, you had access to the latest developments in AI. If not, tough luck. In 2013, 160,000 people registered to the Introduction to AI from Stanford….

 

What are the consequences of this explosion of AI ? 

– One after the other, skills that were mastered by humans are being mastered by machines : shape recognition, recognition of written documents, recognition of oral speech, understanding of the meaning behind the words, etc. What is mastered can not be un-mastered, and is the stepping stone for the next skill to be learnt by machines.

– Almost everyone can now develop AI, access the latest source of knowledge, use standardised frameworks, with cheaper computing power. We could summarise it as:

Growth of AI = More People x Accessible Knowledge x Standardisation of Programming  x Computing Power

– All these individual components are each growing very quickly. Much more people are involved in AI, knowledge is being disseminated faster and faster, programming is getting easier, computing power cheaper and faster.

– Which means that the growth of AI is nowhere near slowing down – on the contrary, it should grow even faster as these 4 factors are self-reinforcing. 

Conclusion: 

– Artificial Intelligence is set to grow even faster than what we have experienced so far. How to make it a good thing is definitely the question for our times. 

Follow me on Twitter here