One of the goal of AI is to improve predictions…
This time, after reading articles and mostly listening talks about AI, I want to write about:
(1.) the main role of AI : making better predictions, as AI is the best predictor ever
(2.) the risk of it, since increasing predictions implies that this AI system gets implemented with a lot of data, our data, eventually, which directly leads to the fair question (and issue) of AI limitations:
(3.) (our) data protection.
I will first start by writing about the advantages of AI in business, how it will help improve business by improving predictions’ accuracy, then talk about its risks and limitations.
Predictions and business
Predictions start to be better predicted by AI because it relies on more data, i.e. the output can be better predicted because the model can learn better from so many past cases implemented into its system. Remember the previous section about supervised learning or machine learning when I introduced the idea of training set / testing set, input versus known output. Once the computer is fully trained it can now predict an output based on new inputs, i.e. new data you feed it with.
Let’s say you want to know what the weather would be in a week because of a big event approaching (and you think the weather might affect the venue of people). By definition the weather is unpredictable, as the stock market, because this is a non-linear function but models try to approximate it as much as they can.
With AI, predictions become faster, more accurate, easier, and therefore cheaper. The only “problem” is to integrate such predictions in business models so that people can really make their decision and consequently the future of their company in the hand of AI, i.e. of predictions made by AI. New companies, as “house of bots“, try to teach so.
Google, for example, is developing a new AI supercomputer with intermediate states – between 0 and 1 binary (basic) states, using quantum bits – which allows to improve even better predictions (i.e. increase prediction accuracy). You can think of basic states as an ON and OFF switch (if you think of the light, for example) and quantum bits as intermediate states (between ON and OFF) as “50 shades of grey”. These quantum bits are a superposition (a linear combination) of these basic states (0 and 1).
So an easy way to get started with AI applications onto your business is to define: Which decisions would you like to be relying on less uncertainty and be more accurately predicted? With this question, you will find the answer which will allow you to identify what you need AI for. From there, you might wonder how you will use and trust AI – its predictions – to impact on your business/organization/… decisions and, therefore, actions? Fair question to wonder how much you could rely on it? How would you value the impact of AI outcome? And, last but not least, and maybe rather to start with, what type of data would you need to improve the AI system you got, i.e. its predictions accuracy to make your business/organization/… perform even better.
Indeed, based on inaccurate predictions, a lot of false positive, for example, you may loose a lot of money. Therefore, predictions which become more and more accurate, in other words predictions that tend to be 100% correct without ever reaching this level might profit a lot your business model. For this, you just need to choose the best AI system corresponding to your needs. This is key. Then you are good to go and try out, keeping in mind that the AI system can always improve, i.e. perform better.
For example, the more data – input/output case scenarios – and the more diverse dataset you implement your AI system with, the more it would be able to make accurate and appropriate predictions as more and more diverse data would allow the AI system to learn more, faster and better.
For instance, if you buy an alarm system for your company / organization, it may be improved so much that you could even be able to predict the intention of someone to commit a crime (break in) at the door or in the vicinity of your company before she even thinks or even knows she will make one.
This example was inspired from a great movie with Tom Cruise called “Minority Report” showing that in the future – around the years 2050 – it would be possible to predict who is going to commit a crime, thus prevent it from happening. In theory, sounds great. But then comes worst.
A recent serie on Netflix called “Person of Interest” tells the story of millions of people being screened in public spaces from cameras equipped with face recognition softwares. This becomes more scary. Besides, this is no science fiction anymore. Such practice is currently applied in China where good points and bad points get attributed to citizens based on their behavior. The latter gets screened by millions of cameras connected to a big database and to the Internet, therefore easily accessible to their employers, to the bank to which they wish to ask a credit for their houses, etc. Sounds unbelievable but this is already happening…
Imagine all the data about you one could find, for example, if you pay your bills on time, how many fanes did you receive the past six months, why, etc. – all translated into points (good or bad) which would have an impact into your life (access to credit, etc.). As if someone is always watching you and accordingly, give you reward or punishment… Doesn’t sound very appealing, does it? More importantly, would you still feel free?
You have reached the end of this post. The post would continue on a new page.