I am back again on a topic which is the buzzword for 2018 and 2019, again the companies and less educated people make marketing statements on technologies they do not understand. That's right AI (Artificial Intelligence) and ML (Machine Learning) included. Based on the buzz and fuzz on trends websites we have seen ML and AI rising like mushrooms after rain, we now have "experts" in this field, uneducated (in technical fields) people talking about how they already use AI. Also there was no conference (big or small) without AI or/and ML as main topics in 2018, and 2019 looks the same. But let's look a bit more and analyze some things.
This is the first article of a long series on the topic of AI and ML and for each question raised in this article I will write some mini series of articles based on my personal work experience in this field.
What is the definition of intelligence?
First of all even the most intelligent people on the planet could not get to an agreement on what intelligence is. How do we define it? How do we agree on it in a non biased way?
Do we define intelligence from a philosophical point of view? From a technical perspective? From own interpretation? There are so many theories and biases regarding what is intelligent and what not, the worst part is that we define these in a very biased way and whatever suits our needs at the moment. A dog is intelligent but only when compared to a frog, if I dare to say that a 2 year old German Shepherd is as smart or even smarter as your 2 weeks old child we will never be friends again. More on this topic in another article.
I would dare to say that whatever is presented right now is a more than idealized image of intelligence, and should be almost superhuman, if I dare to present or talk about a simple algorithm to learn something or a simple algorithm to provide a solution, this is not intelligent this is just a dumb algorithm. Although our brains work also based on small algorithms.
Where should AI be implemented?
According to the large masses it should be everywhere and this raises two radical sides - the doomsday believers and the AI groupies. This is also an artificial problem developed by marketers who try now to sell everything as intelligent and produce some crazy stuff like this one:
Do I think is really the best use?
No!
Do I think is creepy?
A bit.
Do I think we need more of this?
Nope.
Is it really AI?
Not quite, depends how you define AI but is for sure marketed as such.
Is Alexa & Co. AI?
Pardon me Amazon, but no, is still dumb as a rock, whoever tried to develop Alexa skills had also to work past the vocal commands and answers for them. So ... nope I can program 5 jokes as answer to "tell me a joke" and answer with a random joke, I can program it to get whatever joke updates from a website so it looks like is smarter BUT it can't invent jokes (well, you can still do a small program to do that but it has to have first a database of jokes to start with, can't understand the concept of joke or can't create one based only on dictionary as base for words.
Don't rub your fingers Google or Apple or Microsoft, your Siri, Google Assistant or Cortana are not brighter. Maybe all are good enough that you guys steal data from us to gather enough training data for your next chatbot ... but at the moment they are all dumb and pre-programmed to do things, and of course we have some bugs in code so that Alexa laughs creepy. But there is no sign of intelligence in that. To see how dumb they are actually:
From 7th minute comes also the interesting part on the voice training, but there is no intelligence in that, just a pattern of learning without any context.
Ethics and AI?
I have seen people who do not display any ethics in their job position who dare to talk about ethics and are members of groups which will try standardize the ethics ... there is an old saying ... like father like son.
BIG DATA - "L1ZY" from GHOST+COW FILMS on Vimeo.
The above video can be also used as a base for wrong ethics, and what can happen when the wrong people are empowered to decide on the ethics for a powerful technology.
AI and ML to empower experts?
This is what we keep hearing, that AI will empower experts and automate boring tasks. Let's ask ourselves how many of the automations are actually in need for AI or just by simplifying the business processes we could reach this automation with lower costs? How much easy will be our business life without overcomplicated processes? Do I think that AI will be able to adapt to stupid decisions which make a simple approval process to a very complex one involving at least 20 emails with different back-office representatives? NO! But at the time the AI is able to adapt like a human just for a simple process then I would have invested huge amounts of money for nothing, as CEO I could just have fired the person who builds such complicated processes and hire a smarter one, still costs less than developing a monster code to cope with some crap.
Experts? Today we have experts but how can we be sure that tomorrow we will have experts and the software written today will have some experts to serve?
Creativity? Feelings? Is AI able of such things? How about empathy?
Well depend who you ask and how creativity is defined. This would be an interesting topic to write on. But for the moment I want to let you with a question: Generative art which is human coded is not considered AI, but if I train a code to draw a flower is AI, which one is creative? Both? None? The human who thought about how the machine should draw something?
For the moment we are far away from AI as it is presented and faked in many situations, maybe in few years we will be further or the trend will die in the shadow of another buzzword like it happened with IoT and Smart movement, all the conferences with Smart Cities, Smart Whatever are long gone, now is AI time.