Artificial Intelligence has become ubiquitous in the technological sector: we find it in factories, in call centers, in advertising, everywhere. But what is it exactly and what are its technologies?
What is Artificial Intelligence?
Artificial intelligence is a part of computer science that allows you to simulate human abilities through algorithms, thus creating machines that can act and think like humans by performing complex tasks that would otherwise require human intelligence. AI uses deep learning, machine learning and programming to process a huge amount of data that is used to recognize patterns and perform actions, mimicking the decision-making and problem-solving abilities of the human mind.
How many forms of AI are there?
Based on their skill level, we can identify three main types of Artificial Intelligence.
Weak Artificial Intelligence
Weak or Limited Artificial Intelligence, also known as Artificial Narrow Intelligence (ANI), or simply Narrow AI or Weak AI, focuses on a narrow scope and has a limited range of abilities. For this reason, this type of AI, which is currently the only one to have been realized, allows computers to overcome humans only in defined tasks, unlike AGI, which could have different implications. ANI is therefore the highest level of Artificial Intelligence we have, however there is still plenty of room for its improvement.
Narrow AI includes all software that uses machine learning or data mining to make decisions.
Examples of ANI are:
- IBM’s Watson supercomputer: an expert machine that uses AI to mimic the cognitive abilities of humans and answer questions to users
- Alphago, the computer program of deepmind Technologies that, in 2015, beat the best world player of GO, a board game of Chinese origin, in 4 games out of 5
- Siri, Google Assistant, and Cortana: they were trained in speech recognition to interpret human speech and provide a consistent response.
Social media such as Facebook and Twitter also use ANI to present to users the topics that most match their interests, so that they see posts that they will click on more likely.
Strong Artificial Intelligence
Another AI technology is Strong Artificial Intelligence, also known as General Artificial Intelligence (AGI). This technology would be at the level of a human mind and could therefore think like a human being. It will take some time before we fully understand and discover AGI as we have yet to discover many aspects of the functioning of the human brain.
This line of AI is also known as StrongAI or Human Level AI, precisely because it will be able to perform the same intellectual tasks as a human being. In fact, unlike ANI, AGI will succeed in imitating the general abilities of human beings, not limiting itself to specific restricted tasks.
ANI may be better at its specific task than AGI, but General AI can do more than one thing. The fundamental difference between the two in fact lies in the field, which as far as ANI is concerned is only one.
General Artificial Intelligence will be achieved when AI reaches the human level in the following aspects:
- Ability to reason, solve problems, use strategies and make decisions in uncertain conditions
- Represent knowledge
- Ability to plan
- Learning ability
- Ability to communicate using natural language
- Integrate all previous points towards a common goal
Tests confirming AGI at the human level
Different tests have been created to confirm the AGI at the human level, the most famous are:
- The Turing test: is a method used to assess whether a machine is able to exhibit an intelligent attitude. The test consists in human machine interaction: if the chatbot can simulate intelligent human responses thus deceiving the human examiner about its nature, the test can be considered passed.
- The Wozniak test: also known as Coffee Test, in this test the AGI must be able to enter a house and, without any help, understand how to make a coffee (find the machine and coffee, add water, find a cup and prepare the coffee by pressing the appropriate buttons).
- The robot university student test: devised by computer theorist Ben Goertzel, this test claims that an Artificial Intelligence that is able to obtain a degree in the same way a human being should be considered conscious.
Super Artificial Intelligence
The last aspect of AI is Super Artificial Intelligence (ASI) which refers to technology that will surpass the human mind, even the most gifted and brilliant minds. This type of AI could not only perform tasks and actions, but also be able to have emotions and relationships.
Among the researchers there are doubts about how to achieve this type of AI but we often talk about recursive self-improvement: the ASI, with the intelligence and creativity of a human being but with a much better processing speed, would be able to improve itself very quickly and without human intervention.
The exponential growth that would occur with the transition to the ASI is called “intelligence explosion“, in fact the ASI could surpass the human brain in all areas.
The evolution from AGI to ASI could be much faster than the current transition from ANI and AGI, since AGI would allow computers to think and improve themselves once learned from experience, trial and error.
What awaits us in the future?
To conclude, we can say that we are still far from AGI and ASI technologies. However, it is a scenario that could come true one day and, for that, we should consider all the implications, positive and negative, which would involve to ensure that these technologies are safe and beneficial for us and ensure a fair collaboration between humans and machines.
As for the ethical and moral part of Artificial Intelligence, we must not wait for the development of the ASI, in fact it is already an important issue. Man must ensure the ethical and safe functioning of AI in all areas and make this a priority for future developments. The speech on Responsible AI, however, still remains poorly defined: at the moment there are few companies (e.g. Microsoft) that try to create guidelines in this perspective.
Another aspect to consider and to work on is that of the Explainable AI that works on a practical level to create a way to explain the answers provided by AI so that they can be easily understood by humans.
The development of the AGI is a source of wide debate, both in terms of feasibility and the risks that it would bring with it. Some researchers and experts believe that within years or decades the Strong AI will be developed, others say that it could take centuries, others still think that it will never come to its development.
At the time of ChatGPT, some even wonder whether modern deep learning systems such as GPT-4 are a not yet complete form of AGI. In fact, as we have all seen in recent months, GPT-4 is not limited to conducting a specific task but is able to manage problems of all kinds (fundamental ability of General Artificial Intelligence) and to reason, plan and learn from experience.
Blue BI has gone further in this field and has decided to integrate ChatGPT to its BBI x Chatbot solution . If you want to know more, read our article about it.