Is Skynet coming?
This week I saw yet another headline ‘Skynet is near – Google scientists create Artificial Intelligence that evolves on its own” intimating that some sort of humanity destroying AI system is not only an inevitability but on the cusp of reality. Icons such as Stephen Hawking and Elon Musk have long proclaimed the end of civilization due to Artificial Intelligence (perhaps the latter has a more sinister purpose in mind for Tesla’s autopilot!) but with all due deference to the accomplishments of these two men, they are talking complete nonsense and I will explain why.
What is AI?
Ultimately, the easiest way to frame this is that AI = Maths
Mathematics on its own is not dangerous but, famously, 
As the song goes, ‘guns don’t kill people, rappers do’! There is nothing scary about AI or technology generally, when used in the right way. Automation is improving our lives every day, making it easier to work and collaborate remotely, for instance. Imagine how much worse the horrendous economic situation we currently find ourselves in would be if nobody was able to work remotely?
The term Artificial Intelligence is, generally speaking, used as a catch-all term for algorithms that are able to detect inherent mathematical relationships between data and then predict or infer outcomes based on these mathematical relationships. As a basic example, if I have datasets for daily wholesale food prices over thirty years and the corresponding weather data, I can train one or a group of algorithms to find a mathematical relationship between the two; for example maybe lower temperatures one year equals more expensive wheat the next. If I find the model to be reasonably successful at making predictions, I could then use it to calculate expected profit margins for packaged food companies based on upcoming annual weather forecasts. When training the algorithms will be generally trying different calculations out to see what, if any, transformation on the input datasets can be done to get to the necessary output such as, does twice the amount of rain lead to a 20% reduction in food prices?
Now, the important things to note here are;
- I must direct the AI to look for specific outcomes e.g. can you predict 1974 food prices from 1973 weather data?
- The AI has no idea WHAT it is doing, other than trying to get as close a match as possible to the data I have asked it to predict
The algorithms are not sentient, they don’t have any understanding of context. They don’t know why it would be useful to be able to predict food prices. If I manage to get the system to predict food prices well, it is not going to disappear into the internet and suddenly pop up on the Forbes List above Jeff Bezos after trading up a storm. It is a useful tool that we as humans can use to help us solve interesting problems and make the world more efficient.
There are some counter points here; AI can be programmed with a sort of motivation to achieve certain outcomes (e.g. get the highest possible score in a game), and it can be programmed to improve itself independently, BUT it still has to be given that motivation by a human. It has to be coded to achieve a particular outcome, within a framework (rules of a game for instance) that is also coded. These are existential boundaries for the system that can’t be broken, just like we cannot simply decide gravity does not apply to us on any given day. The system is constrained.
At AiX, where our incredible team has created the world’s first natural language AI broker, the concept of constraints on the system are paramount. AiX must execute trading actions on behalf of the client involving real assets, often with very large sums of money involved. It cannot make mistakes, and it must be fully auditable, whilst acting like a human all the while. If it makes a mistake, it could cost a fortune. If it screws up the reporting, it could result in large fines from regulators. These are real, practical constraints and in order to deal with them, our human team had to develop the necessary frameworks in which to operate, taking into account the regulatory landscape, legal issues, client experience and commercial realities, before we could even begin to code. These are all incredibly complex, inter-related problems to solve, and a machine would never be able navigate waters such as these. The combination of human creativity and ingenuity with the power of automation, however, has created a product that is absolutely amazing and can outperform any human broker in the world in speed of price discovery and execution.
It is this human creation and use of technology that has driven innovation over the centuries and it will continue to be the former using the latter that drives us forwards. Taking as an example the incredible positive externalities of the invention of the printing press and subsequent widespread dissemination of knowledge and ideas – it is not the printing press that accomplished it, but the humans making use of it. Were jobs displaced by the printing press? Have some terrible words been written since then? Some so terrible that they sparked wars and death? Of course, but the positive has far outweighed the negative, as evidenced by the incredible progress of humanity in so many areas since those times, most of which would have been impossible without the printed word.
In these dark COVID times it is more important than ever to remember our shared humanity. I for one find it heart-warming that governments around the world have rushed to do ‘whatever it takes’ to support people. It may well be too little, too late for many due to the wheels of bureaucracy turning slowly, but politicians are at least attempting to do the right thing and hopefully will eventually get there. COVID will not destroy us and AI will not destroy us – Skynet is not coming and humanity will continue to – in aggregate – use technology as a tool with which to progress, embracing change as we do so.
Author: Jos Evans, Founder and CEO Of AiX Exchange.
Follw him on LinkedIn
Further Information: https://www.aixtrade.com