AI: between imagination and reality

Between imagination and reality

When we talk about Artificial Intelligence we use to think of movies – such as I, Robot, Matrix, Terminator, or we tend to imagine super-power like machine as Deep Blue the IBM computer which defeated World Chess Champion Garry Kasparov during a match i 1996.

We analysed the public perception on this matter in our latest survey AI for the People, from which resulted that:

  • 42% of people interviewed immediately associate AI with robots;
  • 25% of people interviewed thinks about AI as of human-like thinking computers.

So it is quite evident that books and movies have influenced our perception of AI, making us thinking about it as machine’s ability to think for themselves. As a consequence, these narratives have detached our perception from the true potential of AI in solving different issues on a a wide scale.

Artificial Intelligence vs Human Intelligence

From a certain point of view, AI could be seen as a research process aiming at emulate human brain. The goal is to create an “artificial brain” able to re-enact those activities we usually assign only to human beings abilities.

If we define Artificial Intelligence as the “theory and development of computer systems capable of performing tasks that normally require human intelligence”, how can we, by contrast, define human intelligence?

“Modalities” of Human Intelligence

According to the American developmental psychologist Howard Gardner’s “theory of multiples intelligences”, human intelligence can be classified into specific and independent ‘modalities’ deriving from different parts of the human brain:

  • Verbal-linguistic
  • Logical-mathematical
  • Visual-spatial
  • Bodily-kinesthetic
  • Musical-rhythmic and harmonic
  • Interpersonal 
  • Intrapersonal 
  • Naturalistic
  • Existential

From this classification we can desume that is really difficult to emulate all the different aspects of human intelligence. So, the right question to ask could be: on which type of intelligence should we develop an Artificial Intelligence system?

AI’s applications

AI is often associated with something intangible. In fact, 40% of respondents to our survey stated that they have a very limited knowledge on the matter, moreover they do not really trust informations found on newspapers or the Internet.

However, AI’s application are more common than we may think and they include:

  • Virtual Assistants, which can plan meeting or find attachments in our emails.
  • Handwriting-recognition systems: for a computer is really hard to recognise human calligraphy; in order to do so they need to detect all possible different ways in which people write the same thing.
  • Facial-recognition systems: they can identify people by means of pictures or video-frames. This application is quite scary considering that there are cities in China using it to control citizens, but it is a more common application than we believe (e.g. on web platforms like Facebook).
  • Voice-recognition systems: have you ever asked something to Siri, Alexa or Google Home? Well, these systems are all based on AI.
  • Automatic translation: thanks to applications such as Google Translate now we do not have to worry about moving around the world without understanding what people or signals say.
  • Recommendation Engines: they use AI to identify internet users’ habits and provide suggestions on what to buy, watch or listen to online. The majority of platforms use this application (Netflix, Spotify, Amazon, etc.).

The role of algorithms.

When talking about AI we have to take into account algorithms. We can define them as ” a sequence of procedures which allow to perform a defined task”.

Each algorithm accepts defined inputs (the things you act on) and has the goal of producing defined outputs (the desired results) and there may be different algorithms to perform the same task.

One of the most popular algorithm is Bubble Sort, a simple sorting algorithm that repeatedly steps through the list, compares adjacent elements and swaps them if they are in the wrong order. Thus the algorithm process a random sequence of number (defined input) and sorts it in the right order (defined output).

But algorithm can be applied in other contexts as well. For example: you are invited to an exclusive party and you can bring the five most interesting people you know, choosing on the data provided by social networks like Facebook, Instagram or LinkedIn. So, who would you invite?

However, in this scenery the true question is: how can we choose those five people using social media data? Which parameters do we take into consideration? We could, for example, count the numbers of like, check the hashtag this persone uses or the place where its pictures have been geotagged.  

It’s not easy to emulate human complexity

In fact, each one of us has a different notion of what is fun or interesting. How could we even set criteria or data to establish if someone is entertaining or worth of our interest?

It is in a situation like that imagined above that we can grasp human complexity in contrast to artificial intelligence and that is is inevitable that to from subjective inputs derive subjectives outcome. There are always real people behind algorithms.

In conclusion, AI – which seems to have a solution for all our problem – has still more flaws and complexities than we know.