Artificial Intelligence and Machine Learning are two industry buzz words that we’re hearing more and more of at the moment. Even at my work there are faint rumblings of interest in the topics. This of course has both grabbed my attention and poked my ‘sceptics bone’…
As a result I’ve begun researching and learning about the subjects and I have also agreed to give two talks on them. To help this process along, and share my thoughts and research more widely I’m going to write a series of blog posts.
As an introduction I’ll state my starting position and what I’ll cover during these posts. Maybe as these posts progress my initial view will change, we’ll have to see.
Overall I’m an optimistic sceptic, while I believe machine learning will revolutionise the way we live and work I don’t believe a ‘Skynet’ AI revolution will occur any time soon, if at all. As AI expert @Paul_Scharre recently said in a Channel 4 interview about ‘Killer Robots’…
I think a Terminator scenario like we see in science fiction movies, that’s pretty far fetched, that’s a long ways off and I think we can sleep safely at night not worrying about those kind of scenarios. But the basic technology to put together a weapon that could go out on its own, that could find its own targets that exists today.
Paul Scharre highlights an important point, that there is a significant difference between Machine Learning and sentient AI. Machine Learning which is what Facebook, Google, Amazon, the MoD, etc are all interested in is essentially highly advanced data processing that enables highly capable automation, like killer drones. It is not though ‘intelligent’ in the same way a human is.
Machine learning merely scratches the surface of AI and there are a huge array of issues to overcome before we achieve what is really meant by Artificial Intelligence. Conciousness being a very important one. We may also find Machine Learning so useful and revolutionary we don’t need or want to bother with AI.
The media are also having great fun with the topic, which is both stoking interest and fears. The recent Facebook AI Bot shutdown was described in the Evening Standard thusly…
A London-based artificial inteligence expert has played down fears of a runaway AI “consuming everything in its path” in scenes similar to Terminator 2 after Facebook shut down two chat robots that developed their own language.
The ‘Skynet’ idea is more clickbait than truth. In this case Facebook simply turned off an experiment that didn’t work out. Overall the likelihood of humanity achieving the AI ‘singularity’ in the next 50 years is low.
So if ‘Aritificial Intelligence’ is highly unlikely in the near future what do I plan to cover in these posts and my talks? A good place to start is with the history. One of the breakthroughs that Machine Learning will achieve in the near future is highly advanced automation. The fear is that this may have a negative effect on our economy and put a lot of people out of work. Automation though is not a new thing, we’ve been dealing with its consequences for nearly 400 years. So my first post will be on the history of automation.
Then I will look at what Machine Learning is; what challenges we will face if we want to achieve AI; how this will effect our economy and lives; what the billionaires investing in this stuff think; and I also hope to produce an example of a simple, functioning neural network with TensorFlow. We’ll have to see how that last one goes though…
Finally I have to say a little thank you to a new friend of mine @jonesc55 who knows far more about these topics than I and has been providing some very useful insight and guidance.