Can Artificial Intelligence replace Human Intelligence?
Artificial Intelligence Versus Human Intelligence: Artificial intelligence (AI) has been the hype of the decade with some remarkable innovations which would seem impossible a few years ago. Right from the virtual assistants in our smartphones to sophisticated humanoid robots such as Sophia, AI seems to have grown into a reality right out of a sci-fi movie. However, this also brings us to an intriguing question: “Can artificial intelligence replace human intelligence in the future?”
To answer this question, we should first understand the true concept of intelligence and how it compares with AI.
Comparing human intelligence and artificial intelligence
Intelligence has been a rather controversial topic in the history of mankind. While some psychologists believe that intelligence is a singular ability, others claim that intelligence includes a combination of abilities and skills. While there is no standard definition for intelligence, some widely accepted interpretations seem to suggest that true intelligence encompasses the following abilities:
Logical reasoning is the idea of employing a coherent and reasonable sequence of steps in order to make important decisions and solve problems, among others.
Reasoning is defined as the process of taking bits of information and putting them together to form logical conclusions. For example, if we see a bunch of players in a cricket stadium, we can identify them as cricket players. Now if we see the same players doing a photoshoot for a brand advertisement, without a bat and a ball, we will still know that they are cricket players. So how is that possible? We are taking information that is known, combining it with the background knowledge of cricket and arriving at conclusions regarding what is unknown to us. Now, our AI would not have been able to make this inference if it was taught that a person should have a bat or a ball, or they should be wearing a cricket jersey in order to be identified as a cricket player.
Experts believe that it might be some time before machines can achieve true reasoning capabilities. However, some also believe that given the current breakneck pace of technological advancement, it is highly likely that in the decades to come, machines might just outperform humans in terms of reasoning.
Many researchers strongly believe that intelligence and problem solving are closely related. This is evident in the fact that almost every definition of human intelligence recognizes problem-solving ability as an integral component of intelligence. Moreover, it is commonly experienced that a person’s intelligence is often thought of as the very predictor of how well they can solve problems.
In Computer Science, a problem solving approach includes algorithms and heuristics to find solutions. The problems solved by AI were solved by humans at one point in time. The steps involved are:
Goal formulation: This is the most basic step in a problem-solving approach. It involves the arrangement of finite steps to formulate the end-goals. Goal formulation is mainly based on the AI agent and its performance measure.
Problem formulation: This is one of the most important steps involved in problem solving because it determines what action should be implemented in order to achieve the aforementioned formulated goal. In AI, this step is based on the software agent.
Learning from experiences:
Another vital aspect of intelligence is the ability to learn from experiences. Researchers have shown that our experiences can indeed help us crack complicated situations. Learning from our past experience can actually change our brain’s circuitry, enabling us to easily identify what we are seeing and make decisions to take suitable actions. Professor of experimental psychology and lead researcher, Dr Zoe Kourtzi from the University of Birmingham, stated:
“What we have found is that learning from past experience actually rewires our brains so that we can categories the things we are looking at, and respond appropriately to them in any context.”
Fortunately, AI enables machines to learn from their experience and perhaps even adjust to new inputs like a human would do. In fact, the very concept of deep learning involves processing huge amounts of data and recognizing hidden patterns in the same.
But what if we do not have the data required to model the problem? It would be amazing to have an algorithm to learn everything from scratch as if it were a baby. This is where reinforcement learning comes in. In simplest words, reinforcement learning can be described as an algorithm which “learns to learn”. It takes its inspiration from behavioral psychology wherein the machine learns how to “behave” based on rewards and penalties.
Creativity is often associated with something to do with arts. But this notion is wrong. In fact, creativity could be anything ranging from creating or discovering new ideas and possibilities. Creativity enables us to approach an unfamiliar situation with a novel solution.
Having said that, a lot of people believe that creativity is something that is exclusively limited to humans. They feel that the ability to be creative is something that is far beyond the reach of machines. The most common criticism against AI’s ability boils down to the common belief that anything which comes out of the algorithm is the creativity of the programmer rather than the machine itself. However, getting machines to accurately imitate humans might be a good start for a creative AI. Perhaps, the next strategy would be to get the machine to use those skills to build something original. More importantly, AI combined with creative problem-solving capabilities has the potential to open up several opportunities. Scientists are working on advanced algorithms in order to predict problems and come up with creative solutions. This, as believed by experts, has the potential to push the boundaries of creativity.
All in all, we can see that Artificial intelligence
“Certainly AI is proving to be an invaluable tool, and intelligent workflow is going to be the labor-saving norm within just a few years. But business processes involve intelligent thought and intelligent behavior. AI is great at replicating intelligent behavior, but intelligent thought is another matter. We don’t fully understand how intelligent human thoughts develop, so we’re not going to build machines that can have them anytime soon.”
Now that we have a simple understanding of how intelligence can be interpreted for the machine world, let’s have a look at a comparison of a human brain and a computer.
Biological brain versus the world’s most powerful supercomputer
|Performance||It is postulated that the human brain operates at 1000 petaFLOPs which is equivalent to a billion billion computations per second.||As of now, the world’s fastest supercomputer operates at a speed of 415.5 petaFLOPS (415.5 quadrillion computations per second)|
|Energy required||The energy required is just enough to operate an LED bulb.||The amount of energy needed to run the world’s fastest supercomputer’s computations will be sufficient to power an entire building.|
|Size and space||The average adult human brain weighs 1.3 to 1.4 kgs. It is roughly the size of two clenched fists.||The fastest supercomputer is the size of an entire stadium.|
The chart tells us what a miraculous creation the brain is. In fact, even the fastest supercomputers out there have failed terribly in trying to emulate the brain. Researchers Markus Diesmann and Abigail Morrison built an artificial neural network with 1.73 billion neurons and 10.4 trillion synapses. Although this is a remarkable achievement, this represents just a small number of the total neurons in the human brain. We have about 80 – 100 billion neurons in our brains which is just about the number of stars in an average sized galaxy like the Milky Way.
The researchers were able to simulate 1 second of biological brain processing time in a span of 40 minutes with about 83,000 processors.
However, even though the activity took an enormous amount of computing resources, we must understand that in the previous generations of computers, the fastest computers required a lot of resources and had limited applications. This brings in hope that maybe someday scientists can have AI come at par with our complex brains. But what next? We wouldn’t want an artificial brain to become more complex and powerful than our brains, would we?
So what is AI better at than humans?
Even though we have understood a gist of how AI and supercomputers are nothing in front of the complexity of a human brain, it turns out that there are certain areas where AI is outperforming humans.
Medical diagnosis has become a major focus area for AI. This is so because it is relatively difficult for humans to do an entire correct diagnosis. This can be backed by a study conducted by University Hospitals Birmingham which showed that AI could correctly identify a disease 87% of the times as compared to 86% in the case of healthcare experts.
IBM’s Watson for Health empowers hospitals, researchers and patients for streamlining the workflow and enhancing decision making. Similarly, Siemens Healthineers’ AI-powered healthcare solution can read chest CT images, take automated measurements and generate a medical report with appropriate quantifications.
In 1997, IBM’s supercomputer Deep Blue won against the then world chess champion, Garry Kasparov. Back in those times, the idea of having a machine beat humans was extraordinary. Since then, chess AI has developed to an extent that it does not require a supercomputer.
In 2016, Google DeepMind’s AlphaGo beat Lee Sedol – a highly ranked Go player – in the board game of Go which is thought to be more complicated than chess because of the large number of possible moves. Moreover, in 2017, an advanced neural network ‘AlphaGo Zero’ beat ‘AlphaGo’ 100 times in a row.
Reinforcement learning was instrumental in AlphaGo’s success. It was trained on a large number of games from where it picked play styles that resulted in a win.
Translators are tools that use advanced neural networks to translate text – written or spoken. Sometimes, these tools can also translate the meaning and sentiment of the message.
Google Translate App can translate text in 27 languages in real time. Moreover, in 2019, a man from Nigeria developed a portal that can translate a whopping 2000 African languages. This was a huge achievement in order to overcome the communication barrier in the continent.
What does the future look like?
At present, AI has some limitations which prevent the technology from reaching its full potential. We are still far from the highly ambitious goal of having machines advance to the level of human intelligence. However, as technology is getting more advanced by the day, the future of AI looks quite promising.
The ultimate goal would be to see an intelligent system thrive in the real world with little to no human intervention. Perhaps, with global giants such as Google and IBM working on AI, it is possible that the day isn’t too far. The next generation will see the coexistence of humans and humanoids. Life at such a time would be much more advanced and superior than anything we could have imagined. It won’t be long before science fiction of the past becomes a complete reality and for the good.
What is Artificial Intelligence?
Artificial Intelligence is a broad area of computer science focused on creating intelligent machines that can perform tasks that would otherwise require human intelligence. However, AI is not restricted to things that are biologically observable.
What are the prominent sub-fields of Artificial Intelligence?
The prominent branches of AI include Machine Learning, Deep Learning, Natural Language Processing (NLP), Computer Vision, Reinforcement Learning, robotics, etc.
What is deep learning?
Deep learning is a subfield of Artificial Intelligence that mimics the human brain and processes vast amounts of data to recognise patterns for decision making.