Insights, ML and AI

When Will Artificial Intelligence be as Smart as Humans?

Artificial Intelligence: machines as smart as humans

SOLVVE has already talked briefly explained in simple terms the basics and the most important ideas and concepts of artificial intelligence in the previous articles. We found out what are the goals for the machine learning experts at the moment. However, if the whole point of developing artificial intelligence is to mimic human cognition, then the question arises: when will artificial intelligence be as smart as humans? In this article, SOLVVE is going to take a look at what human intelligence exactly is and how fast can our technology catch up with our cognition.

What is Intelligence?

In our previous article, we discussed how there is no one definition of what AI is. That is because there is no definition of what human intelligence is. There are several approaches that use a different set of criteria to define, and as you will see below, they are quite controversial.

Depending on what area of knowledge you come from, you might want to look for different things before judging somethings as intelligent or otherwise. On the one hand, some scientists look at the specific abilities: to solve problems, make judgments (especially when there is not enough data), forecasting, reasoning, learning, recognition of sounds, images, and behavior, understanding of natural languages, etc.

On the other hand, the mere combination of all or some of these factors does not suffice, and the intelligent entity should also be able to “make sense” of reality, to gain experiences, to understand different contexts and freely orient itself in any situation, to find solutions not through following pre-defined algorithms but through finding new ways by thinking through the situation, etc.

How to measure intelligence: a close up of an equation

How to measure intelligence?

This is an even more trickier question, provided that we don’t even have a strictly defined set of parameters that we try to replicate. When we talk about humans there are many different ways that try to measure it, from school exams to define if someone has gained a needed amount of knowledge to extensive tests that try to measure intelligence quotient (IQ) and emotional intelligence

When we talk about artificial intelligence, none of these are applicable since the current machine learning capacities can only perform a fraction of what the human brain can. Similar to how we cannot define what human intelligence is and how to measure it, it is hard to tell what should artificial intelligence be capable of doing for us to call it equal to human cognition.

Nevertheless, at a theoretical level, scientists try to come up with an ultimate way to tell the intelligent machines from a regular one. You have probably heard about some of these tests.

The most famous probably being the Turing test – if a person who listens to a conversation of a human and a machine cannot tell who is who, we can consider the machine to qualify as intelligent. The problem is, Turing himself never gave a definition of what intelligence is. So it is a “yes or no” situation that does not really explain what are the building blocks of the artificial intellect.

There is also a coffee test developed by Steve Wozniak. To prove that a machine is intelligent is must be able to figure out how to brew a cup of coffee in an average home that it visits for the first time. It should locate the cattle or coffee machine, find ingredients, manipulate the machine, and get a cup of coffee.

There are also so-called college student test and employment test. The prior one is expecting an artificial intelligence to pass the enrollment test to a university, carry out through the program and graduate as a regular human student would do. The later prescribes machines to perform as well as humans at any job position that is economically important to society to qualify as intelligent.

Subsequently, scientists expect artificial intelligence to be able to lead a rather regular independent human life without assistance to pass as intelligent. However, since we do not know what intelligence is, the current approach is to recreate single capacities one-by-one, until one day we know what all of the elements are and how they interact with each other. 

A robot with AI-powered social intelligence holding a girl's hand

What is the current state of AI?

Today we have what scientists call a narrow AI, algorithms that can solve strictly define issues: recognize faces, monitor and interpret behavioral patterns, recommend content or friends, etc. On the contrary, what scientists are aiming for is a so-called strong AI (also referred to as full AI or artificial general intelligence (AGI)) that will be as smart as humans.

It is also worth mentioning that reaching the human level of capabilities will probably not be the end of AI development. Overtime AI will surpass human cognition and become a superintelligence – a machine smart enough to endlessly improve itself, become self-aware. This event is often mentioned together with the intelligence explosion and achievement of the singularity, which we might discuss some other time.

Today, AI can beat a human in a specific task, but on a general scale of being a well-rounded person, AI is like a 5-year old child at the best. Skills that AI acquire are not transferable, and to acquire them AI requires far more repetitions than a human. 

When will artificial intelligence be as smart as humans?

When AI raised in popularity for the first time, many scientists were very optimistic. For example, Marvin Minsky, a scientist from MIT known for his input into Artificial Intelligence, claimed back in the 90s that the issue of solving the issue of creating artificial intelligence would be solved in a decade. Yet, in 2020 we know that we are not quite there yet. 

Polls were held among the leading specialists in the industry aiming to draw estimates of a new date. As one might expect, these are not uniform at all. These polls are renewed from time to time. Still, results remain about the same. Those who are more positive name 2060 as a new milestone date, others suggest that we have to wait until 2090 or longer. Even so, the chances of achieving AGI within our lifetime are 50/50.

Nevertheless, despite scientists’ ongoing struggle with theorizing about what intelligence is, machine learning experts keep on developing solutions that are good at solving strictly defined sets of issues. If you if have any questions or suggestions about applications of AI in such industries as eLearning, eSports, business development, or security, be sure to contact us for more information and consultation. Let us make it happen!