What is AI? Artificial Intelligence in simple words

“AI is likely to be either the best or the worst thing to happen to humanity.”

– Stephen Hawking

With the advent of ChatGPT, self-driving cars, facial recognition, and Snapchat filters, to name but a few examples, it seems that Artificial Intelligence (AI) is now everywhere. In fact, AI has become so central to our way of living, that AI was declared 2023’s “word of the year” by Collins Dictionary. Even though AI seems to be relatively new, the phrase was actually created in 1956 at a conference that is widely accepted to be the birthplace of AI – the Dartmouth Summer Research Project on Artificial Intelligence. With many possessing only a limited understanding of what AI can do for them in a general sense, the complexities of how AI operates are vast and the possibilities of what AI could achieve in the future seem endless. Continue reading to find out: what AI is; how AI works; the different types of AI; the benefits and applications of AI; the limitations of AI; ethical and safety concerns of AI; and what the future holds for AI.

What is artificial intelligence (AI)?

Artificial Intelligence (AI) is actually a branch of computer science that aims to allow a computer or robot (controlled by a computer) to perform tasks that would normally require human intelligence (e.g. learning from experience, adapting to new situations, problem-solving, reasoning, comprehension, making judgments and much more). While it may sound like AI allows computers to ‘think’ independently like humans, this is not the case. Many people assume AI systems, like Alexa or Siri, can understand text or speech in the same way that humans do; ultimately, AI is only a series of algorithms that can recognize patterns from information inputs. AI systems, unlike humans, also lack comprehension. 

As a way to determine whether an AI model has reached the capability of mimicking human thought processes, they are often assessed against the ‘Turing Test’. As yet, there has been no AI model that has officially passed the Turing Test – ChatGPT creators claimed their model passed the Turing Test, though this has not yet been independently verified. 

Because the notion of ‘intelligence’ itself is difficult to define with reference to AI, it can be broadly outlined in three ways: Artificial Narrow Intelligence, Artificial General Intelligence and Artificial Superintelligence.

Artificial Narrow Intelligence (ANI), sometimes called Weak AI, has limited application and succeeds at performing single tasks very well. Narrow AI operates within a narrow context and is an imitation of human intelligence applied to a specifically designed problem (e.g. changing voice to text). Narrow AI models include: Siri and Alexa; Grammarly; social media filters; self-driving cars (so far, only manufactured by Tesla and Mercedes Benz); web searches; conversational bots (customer service chats on websites); email filters; and streaming service recommendations, to name just a few. At present, ANI is the only type of AI in existence. 

Artificial General Intelligence (AGI), or Strong AI, is a machine that is considered to be ‘on par’ with human intelligence, simulating behaviours with the ability to learn and apply its intelligence to solve any problem. AGI would behave in a way that is nearly identical to that of a human in any given situation; AGI would even possess the ability to have and understand emotions. However, this type of AI remains theoretical and doesn’t yet exist – rest assured though, there are many companies accelerating efforts towards achieving such advanced levels of AI. 

Artificial Superintelligence (ASI) would far surpass its predecessors, with the ability to become cognisant and exceed the intelligence of humans. While such capabilities might seem like the plot to a futuristic, dystopian novel, ASI is presently hypothetical and not a level of advancement that we are likely to encounter in our lifetime.

How does AI work?

The way AI systems function is incredibly complex with many different facets. In essence, these systems function by combining large data sets with well-programmed processing algorithms. These systems then execute numerous tasks at high speed. Bottom line? AI systems are все powered by алгоритмы.  

Most current AI systems ‘learn’ through Machine Learning; a type of learning that requires an input of data, which the system will then process with intuitive algorithms (algorithms designed to mimic human thought processes). After the data has been entered and processed, the AI can then recognise behaviour patterns and errors, adjusting its functions and algorithms as needed. For example, streaming services (Netflix, Prime, Spotify, etc.) use machine learning to generate recommendations for their users. 

Other, more sophisticated, AI models use Deep Learning, which is a more complex form of Machine Learning that uses large neural networks (which mimic the neural pathways of the human brain) to learn multifaceted patterns and make predictions of outcomes without human participation. For example, ChatGPT uses a deep neural network with billions of criteria to analyze and generate text. These AI systems allow for a more personalized user experience.

The algorithms, used to program and help AI systems ‘learn’, are all written by highly-skilled computer programmers (coders).

What are the different forms of AI?

There are four forms of AI: Reactive Machine, Limited Memory, Theory of Mind and Self-Aware.

Reactive Machine

AI systems that have no memory, are unable to store data, and mimic the ability of the human mind to process unexpected stimuli without prior knowledge (i.e. not programmed into the algorithm) are called Reactive Machines. 

The most well-known example of a Reactive Machine is IBM’s Deep Blue. This machine is programmed to understand the rules of chess, including the recognition of all of the pieces on the chessboard and how each piece can move. Deep Blue calculates the most effective next move for itself, whilst predicting the opponent’s most likely course of action, through an analysis of the current situation on the board. However, since there is no memory, Deep Blue is unable to ‘learn’ from previous chess games. The decisions of this Reactive Machine are based only on the current situation and its available options. 

Streaming service recommendations are a more contemporary example of a Reactive Machine AI system. This type of AI processes huge amounts of customer data in order to recommend specific films and TV programs relevant to previous viewing history. 

Limited Memory

Limited Memory AI systems are the most commonly used type of AI in the world today. Unlike Reactive Machine AI systems, Limited Memory AI systems can refer to past experiences over time, much like the way a human brain’s neurons create pathways. These interpretations are then used to improve the algorithm by programming the AI system to allow its actions to be based on past and present data. While it might seem this is the same way a human might garner information from successes and failures (memories), the AI system only temporarily stores past experiences. The information is then exploited by the AI system to ‘train’ itself and improve its capabilities over time, with the more data that it interprets. 

One of the most impressive current examples of a Limited Memory AI system is self-driving cars. These cars operate based on data (from the algorithms they were programmed with), as well as interpreting information gathered by its sensors, adjusting to the environment as necessary. The information these cars interpret includes, but is not limited to, dynamic road conditions, other cars and static objects. 

Theory of Mind

Theory of Mind AI systems are the next frontier for computer programmers. When these AI systems are capable of decision-making equal to that of a human, we will have achieved Theory of Mind; this type of AI system is anticipated for the near future. At the heart of the theoretical Theory of Mind AI system lies the proposed capacity to recognize and remember emotions, which will then enable adjustment to behaviors based on this information – similar to how humans react in social situations. 

While Theory of Mind AI hasn’t yet been achieved, robots like Sophia (created by Hanson Robotics) have come quite close; Sophia is a humanoid robot able to “see” emotions and respond accordingly. 

Self-Aware

Going one step further than the Theory of Mind AI lies Self-Aware AI, which will be an AI system maintaining an awareness and understanding of its own existence. This type of AI system would possess human-level consciousness and intelligence, with the potential to exceed both.

Self-Aware AI is a long way off from becoming a reality, as research continues to understand how the human mind operates and how memory, learning and decision-making abilities function. Furthermore, the algorithms and hardware that currently exist would be unable to support such a sophisticated and complex AI system. 

What are the benefits and applications of AI?

Even though we are continually developing and adapting the AI systems that currently exist, the applications of AI are vast. From healthcare to agriculture and everything in between, AI systems have found utility in almost every sector. 

One of the most important ways that AI systems will benefit the average person is through their applications in the healthcare industry. From diagnosing to treatment to post-medical care, AI systems can streamline and make certain processes more efficient. For example, those with diabetes can wear automatic insulin pumps which are programmed with algorithms reacting to measurements of their glucose levels which automatically inject the correct dose of insulin as and when required. You can read more about the applications of AI systems for healthcare from Imperial College London здесь.  

About agriculture, AI systems are used by farmers to complete a number of tasks that would previously have been difficult and/or time-consuming. For example, AI systems can predict how long a certain crop will take to grow or detect pests. AI systems can even plant, spray and harvest crops through automated machinery, which will save farmers from employing farm laborers or doing the work themself. Read more from AgfunderNews about how the use of AI systems in agriculture is benefitting farmers around the globe. 

The above examples are just two of the many ways that AI systems are used in our everyday lives.

What are the challenges and limitations of AI?

While AI systems seem like they can do it all – in fact, some computers have now crossed the exascale threshold, meaning that they can perform as many calculations in a single second as an individual could in over 31 billion years – there are still many limitations and challenges faced by those creating them. As mentioned earlier, AI systems, no matter how advanced, still only ‘know’ what the programmer (through the algorithms) has allowed them to ‘know’. The primary limitation of AI systems are their lack of empathy and true emotions. However, as robots like Sophia demonstrate, the emotional intelligence of AI systems is progressing quickly. There are many other limitations of AI, such as: potentially inaccurate data analysis, potential bias, and cost. 

What are some of the ethical and safety concerns of AI?

Many people have a fear of AI systems becoming too intelligent, with some going as far to say that robots could one day “take over the world”. In response to this matter, many governments and industries are already taking steps to address the ethical and safety concerns of AI systems. Some of the safety and ethical concerns of AI are: lack of transparency; lack of neutrality; potential for surveillance; further division between the privileged and non-privileged; concerns of human rights; unemployment; and the replacement of independent thought.

Despite AI systems being relatively new, world-renowned science fiction author Isaac Asimov often wrote about robots, technology and AI systems. Interestingly, because of the inclusion of robots within some of his work (I, Robot 1950), he developed Three Laws for Robotics, which are still referenced today when discussing the ethics of creating and using AI systems.

Three Laws for Robotics:
1. A robot may not harm a human being or allow a human being to come to harm through its inaction.

2. A robot must obey the orders given to it by a human unless it is in direct opposition to the first law. 

3. A robot must protect its own existence unless that protection conflicts with the first two laws. 

_{*Asimov later added a fourth law: A robot may not harm humanity or allow humanity to come to harm through its inaction.}

This month, November 2023, safety concerns regarding the creation and use of AI systems were addressed at the global AI Safety Summit held outside of London, UK. In a symbolic gesture to its historic contributions to cryptography, the summit was held at Bletchley Park (famous for the development of many codes and ciphers that led to Allied victory in WW2) and was attended by delegates representing 27 countries, as well as many titans of industry, including Elon Musk. King Charles also recorded a message for those in attendance, stating that, the risks of AI need to be addressed with a “sense of urgency, unity, and collective strength”.

Those attending this significant summit did not attempt to come to an agreement for a shared set of rules for the enforcement of AI. However, AI companies did agree to provide early access to governments of AI systems in order to evaluate safety. The UK has also launched the AI Safety Institute.

What does the future hold for AI?

Get your sunglasses out – the future of AI is bright. With so much potential to improve existing systems and create new ones, the possibilities of AI are endless. Furthermore, with only having achieved a surface level of what could be possible, there will be (and already is) an enormous demand for highly skilled computer programmers to write algorithms for AI systems. These programmers will not just be sought after by robotics companies either, but virtually every industry will be looking to snap up these expert coders.

With such a promising outlook for careers in AI, children as young as four can (and should) be учиться программировать to take advantage of such rewarding and wide-ranging career opportunities in the future.           

As you can see, AI is a complex and multi-faceted technology that features greatly in our everyday lives. The AI systems in existence are already advanced, but the possibilities for future innovation of AI systems are seemingly endless. An AI system is only as ‘smart’ as the algorithm that it has been programmed with; highly-skilled programmers across many industries will be highly sought after, with lucrative salaries likely to be used to entice the best of the best. With ethical and safety issues of AI increasingly concerning many governments and leaders of industry, it is clear that AI is here to stay and is poised to make a difference to the world we live in – hopefully for the better. 

Do you want to deeper explore the world of AI? Check out our Artificial Intelligence course and learn crucial AI concepts and practical applications.