Artificial Intelligence: The Next Human Evolution

Artificial Intelligence is becoming more and more popular the more time passes, and people have used this term to refer to everything that is machine-automated. Whilst just a generalisation, it’s important to highlight how generalisations are intellectually necessary evils, but it’s also important to remember that we live in details.

It means that besides community conventions the field of Artificial Intelligence has many theoretical (in on-going research process) and practical shapes as well, they are: Narrow or Weak AI (ANI), Artificial General Intelligence (AGI) and SuperIntelligence.

Let’s start with the first one: ANI (Narrow AI). This is the level we’re currently at, where the stage is dominated by GPT-4. Machines are able to perform a single (or narrow) task, often faster and better than a human. However, it cannot perform anything else. Also, the output of such agents is partially wrong or imprecise, requiring a strong understanding of prompt engineering in order to gain the best off of them. They can play music, analyse data, surf the web and get information, answer question and make art. This is the case of Midjourney, a ANI model that creates any type of art you ask it to do! Of course, as Midjourney itself, these models present ethical limitations (such as the ability of GPT-4 to generate malicious content or for DALL-E (art generation model) to represent gore and nudity).

From the advancements of Weak AI, scientists had theorised various profiles for the upcoming versions that it can develop into. The “closest” one that we’re about to reach is: Artificial General Intelligence (AGI), or simply known as Strong AI. This the type of AI that can use previous learnings and skills to accomplish new tasks in a different context without human intervention. It can learn and perform any intellectual task that a human can, and potentially even more. For example, an AGI could learn how to play chess, write a novel, or diagnose a disease, by using its own reasoning and creativity. It is based on Deep Learning models that use artificial neural networks that simulate the way the human brain process information. Neural networks exist today, and they’re extensively used for creating projects of different kind (you can also check my GitHub profile to see how), however, we didn’t reach Strong AI yet for several reasons, the most important is: we don’t know how to define what intelligence and consciousness mean, and how to measure them. There is no universal agreement on what constitutes intelligence and consciousness, and how to test them in machines. Different fields and disciplines may have different criteria and methods for evaluating intelligence and consciousness, such as the Turing Test, the Chinese Room Argument, or the Integrated Information Theory. Without a clear and consistent definition and measurement of intelligence and consciousness, it is hard to design and evaluate Strong AI systems.

Another theoretical type of AI that has been theorised (but way more abstract than the previous one) is: SuperIntelligence. It is the type of AI that can think, reason, learn, make judgments, and possess cognitive abilities that surpass those of human beings. It can potentially achieve any goal and solve any problem that humans cannot. For example, an ASI could invent new technologies, discover new scientific laws, or create new forms of art, by using its own intelligence and imagination. The reason why achieving such model is impossible with today’s equipments is because it would require techniques that go beyond the current machine learning and neural network models, therefore, it still remains just a sci-fi idea.

References

Joshi, N. (2019, June 20). 7 Types Of Artificial Intelligence. Forbes. https://www.forbes.com/sites/cognitiveworld/2019/06/19/7-types-of-artificial-intelligence/

Team, I. D. a. A., & Team, I. D. a. A. (2023, October 16). Understanding the different types of artificial intelligence. IBM Blog. https://www.ibm.com/blog/understanding-the-different-types-of-artificial-intelligence/