Date Posted: September 30th 2020
This is the second part of my piece on the importance of Understanding AI (you can read Part 1 here). We’ll pick-up where we left off last week and talk about the potential threats to AI’s exponential growth but before that, I will give you some more background around what AI is and the different types of AI.
AGI and ANI
The fact is that most people expect “Skynet-like” features from AI, namely a digital analogue of our brain. At the same time, scientists realized long ago that our technologies would not be able to create such a system in the foreseeable future, and decided to divide the concept of AI into two components:
- Artificial General Intelligence (AGI) – yes, this is Skynet, the creation of which ignited scientists in the middle of the 20th century, and which still does not allow them to enjoy complete peace of mind;
- Artificial Narrow Intelligence (ANI) – highly specialized solutions, which can safely include all existing AI systems.
Despite their capacity, artificial neural networks, unfortunately, are very far from our neocortex. They are huge formulas and in essence incredibly complex approximators that try to find complex dependencies and patterns in the narrowly focused data that we "feed" to them but they are still no match for the complexity of the human brain. It is worth mentioning however that the method of reinforcement learning shows great potential. Reinforcement learning is the practice of AI models acquiring knowledge and improving through many rapid trials and errors. Scientists have high hopes for this approach, especially after the triumph of the AlphaGo program, developed by DeepMind (and now owned by Google). In 2016, this program defeated one of the best Go players in the world, Lee Sedol and more recently, in 2019, their AlphaStar program managed to defeat some of the top professional players of the complex strategy game StarCraft.
Per aspera ad astra
Will humanity be able to create Skynet in the nearest future? Definitely, not. Is it worth worrying about? Obviously not, either. Of course, at some point in the distant future, the invention of AGI will radically change our civilization. And if Science Fiction has thought us anything, this seismic change may not necessarily be for the better. Irrespective of the fact it would have been a long way coming, people may still simply be unprepared for the developments they have already fantasized about in a wealth of books and movies.
What is very important to note here is AI’s role of a guiding star. Moving towards the goal of AGI, scientists will formulate and solve many other interesting and world changing problems, create new models and technologies. One might consider ANI to be a bright example of this process.
Is winter coming?
New convolutional neural networks are constantly improving Computer Vision systems, including bringing us closer to the era of self-driving cars. The invention of Transformers (this, not this) in 2017 significantly changed Natural Language Processing, and moved us towards the era of widespread use of virtual assistants and consultants that already quite interestingly mimic AGI.
AI is developing rapidly and is not going to give up. Moreover, already existing models and methods, when used correctly, will dramatically change entire industries. And the process of this transformation has only just begun.
There is only one danger – the misinterpretation of AI capabilities. After all, even a positive result of using AI may not live up to the high expectations and cause disappointment, not to mention the deliberate distortion of AI’s abilities in order to parasitize on its rapid development. Therefore, it is very important to clearly perceive what modern AI (actually ANI) can and cannot do. Understanding these things will not only save you money and wasted effort; it will allow the whole AI area to successfully develop further and will avoid yet another AI winter which could be extremely detrimental to the field.
So, what can you do to help?
My recommendations are simple:
1. If you are interested in AI, I strongly advise you to find some time and take a short, but very informative, online course called AI For Everyone. In this course, Andrew Ng, one of the world's most renowned AI professionals and promoters, provides an excellent overview of the real AI capabilities.
2. If you want to implement AI in your organization, be sure to get some tips from an experienced AI professional who will be able to analyze your organization’s readiness and capability to implement AI (for example, the quality of available data) and help you find and start your first AI project.
A final thought: A bad workman always blames his tools
Remember that AI is just a tool. As such, it will be able to lead you into a brand-new world and become a value generating game-changer for your company only when expertly planned and executed, with clear and validated goals.