AI the buzzword !
A buzzworld ?
I regularly read job offers or I often receive calls for jobs for different positions and they do have in common the following : “we work in AI” or “we use AI”.
In place of AI, jobs offer use Machine Learning (ML) 95% of the time, Deep Learning (DL) 4% of the time and or Reinforcement Learning (RL). As these names suggest, these are Learning methods. They represent the ability of a machine to learn from the data. So, these are subfields of the artificial intelligence field, relative to mathematics and statistics.
When I ask recruiters if it is AI (multi-agents) or learning methods, they always answer “we use machine learning” or “we use deep learning”. “why don’t you just call it learning methods instead of artificial intelligence?” Their responses are “AI is a buzzword, we need to sell dreams!”. If your enterprise works on a part of machine learning, deep learning or even in AI (research), you do not need to sell dreams because it’s interesting enough to bring people to you.
Let’s make a quick explanation of what is AI.
What is Artificial Intelligence (AI) ?
AI was born in 1956, during The Dartmouth Conference by John McCarthy, an American computer scientist. It all started in these terms:
We propose that a 2 month, 10 man study of artificial intelligence be carried out during the summer of 1956 at Dartmouth College in Hanover, New Hampshire.
The study is to proceed on the basis of the conjecture that every aspect of learning or any other feature of intelligence can in principle be so precisely described that a machine can be made to simulate it. An attempt will be made to find how to make machines use language, form abstractions and concepts, solve kinds of problems now reserved for humans, and improve themselves.
- Dartmouth AI Project Proposal; J. McCarthy et al.; Aug. 31, 1955.
60 years ago, McCarthy and his team made the assumption that in just two months they could do a General Problem Solver. 60 years later, no General AI has been deployed.
Dictionary :
artificial intelligence
noun
The theory and development of computer systems able to perform tasks normally requiring human intelligence, such as visual perception, speech recognition, decision-making, and translation between languages.
There are actually three historical categories of artificial intelligence.
Weak AI:
Generally, this type of AI is trained for a particular task (ML, DL) like a personal assistant (Siri, Amazon Alexa, Google Assistant…).
General AI:
Or Strong AI is comparable to Human level intelligence. The machine has the intelligence to find a solution to an unknown problem. This kind of AI can understand and reason its environment as a human.
Artificial Super Intelligence:
This kind of AI is particularly hard to describe. If we consider the description of the University of Oxford and AI expert Nick Bostrom (Superintelligence: Paths, Dangers, Strategies) superintellingence is reached when an artificial intelligence becomes better in any fields compared to human intelligence.
The bias is that we always compare with the known “human intelligence” so it is possible to miss other forms of intelligences. The Turing Test is basically designed to compare AI and human perfectly, more open versions are now used.
But, AI can be classified into four other types (Arend Hintze classification):
Type I: Reactive Machines
Basic of AI, this type of AI has no memory, no past, it cannot use experiences to determine the next decision. IBM Deep Blue (AI expert in chess, defeated Garry Kasparov in 1997 — book) is the best representation of this type, Deep Blue sees the movement and reacts for the next move. It analyzes possible moves (its own and its opponent’s) and chooses the most strategic move. Deep Blue and now Google’s AlphaGO were designed for a precise goal and cannot easily be applied to another situation. They are specialised in one unique task.
Type II: Limited memory
AI uses past experiences to predict future ones (autonomous driving). Memory is based on observations and conserved during a time-lapse to generate actions in front of a problem.
Type III: Theory of mind
This type of AI needs a different kind of of computer to perform on (quantum computer ?). It’s the ability to understand (thoughts, feelings and emotions) all the entities and agents in the world around the machine. It’s the basics of social interactions.
We haven’t yet reach this step but, several studies like a SyNAPSE chip bring interesting results or the combination between neural network and quantum work.
Type IV: Self-awareness
It’s the evolution of type III, the machine has a conscience. The machine is also aware of herself, of her condition, of her environment and the feelings of other species. The machine can detect the feelings of someone in relation to its environment. The machine is able to make choices, measure the implications of its choices and determine the consequences from past experiences and future predictions.
Actually, a new class of calculus was developed and described here. The author of this new method — Daniel J. Buehrer — has proposed a new theory around the idea of a computer that can ‘feel’ and ‘think’, if it’s correct and applicable this will be a major step in the field of AI.
