Imagine if an alien race was living among us in the world – quite a daunting thought, isn’t it? After all, what if it turned out that none of the aliens had any sympathy towards the human race?
So far so seemingly unrealistic, until you consider replacing the stereotypical idea of an alien with chips of silicone.
Welcome to the world of artificial intelligence.
When speaking about AI there are three main classifications:
1) Artificial Narrow Intelligence (ANI): AI for one specific thing (“Google Translate”)
2) Artificial General Intelligence (AGI): AI that can perform as though it was human
3) Artificial Super Intelligence (ASI): AI on a level beyond the capabilities of humans
Until fairly recently, AI was limited to what could be achieved with programmers putting a command in a box – in other words, you got out only what you put in (ANI). Now however, a paradigm shift has taken place, today AI is all about machine learning.
Machine learning is defined as:
“a type of artificial intelligence that provides computers with the ability to learn without being explicitly programmed (AGI). A computer program is said to learn tasks from experience if its performance at the task improves with experience, according to some performance measure. Machine learning involves the study of algorithms that can extract information automatically without human guidance.”
This type of AI is growing at an exponential rate and has been dominating the headlines this year. You’d have to have been living under a rock to have missed the Microsoft Tay debacle – a teen girl AI on Twitter who had to be shut down after it turned out that what she was actually ‘learning’ was more than slightly inappropriate.
Google offer some less controversial examples from where they sit at the forefront of artificial general intelligence. For example, a computer called “AlphaGo” developed by Google DeepMind beat a master of the ancient Chinese game “Go”. This is even more impressive than it sounds, considering that the game has so many more possible moves than chess that there’s no possible way of calculating every move on the board. This offered the first demonstration that a machine can learn and think in a human way and it happened a decade ahead of schedule.
Another of Google’s products is RankBrain. In October 2015, Google injected AI into the Hummingbird search algorithm in the form of the RankBrain algorithm. Since then, a “very large fraction” of the millions of queries a second that people type into the search engine have been interpreted by an AI system. RankBrain uses machine learning to better understand language patterns – it recognises more natural ways of expressing queries than the way we traditionally search and therefore understands the user’s question better and is more likely to give relevant search results.
A test of RankBrain was recently performed by “Stone Temple Consulting”, who maintain a database of results of millions of queries within Google. When they looked at the results to queries that were produced pre-RankBrain it was clear that there were massive gaps in understanding where the result did not match the intended meaning of the query.
For example, for one query “Why are PDF’s so weak?” the number one result that Google produced pre-RankBrain was a PDF file explaining why “the Iraqi resistance to the coalition invasion was so weak” which clearly was actually nothing to do with the user query. If you were to search this query today, post-RankBrain, you’d find that the number one search query is a page that talks about the “weakness of encryption and security in a PDF file” which is directly related to what the user was looking for.
What makes the progress of AI so exciting yet a little bit scary is that it’s currently not even at 1% of what it will be. There is a prediction that by 2045, less than 20 years into the future, computers will be more intelligent than all humans combined (ASI).
If that is the case, artificial intelligence could be man’s greatest discovery, fixing fundamental problems on our planet and allowing us to have technology and infrastructure that could have taken us centuries if we tried to figure it out ourselves. Intelligent robots could be used to cure disease, explore space, reach Earth’s nadirs and dig for fuels.
On the other hand, there is a chance that we could be creating something that may destroy us all and end society and the world as we know it. We just don’t know yet.