What is AI
Welcome again! In this blog post we are going to ask the first and most important question there is when discovering Artificial Intelligence, we will dive into what it represents, how it is created, and the general public perception of this amazing new technology.
A computer program is said to learn from experience E with respect to some class of tasks T and performance measure P, if its performance at tasks in T, as measured by P, improves with experience E.
This is the textbook definition of Machine Learning AI, the most popular amongst programmers, engineers, and people working in the field of data science. It is what we do every day at Cortex, to be more specific it’s deep learning, but we won’t get into that right now. Machine learning is the activity in which a statistical model “learns” useful patterns in data. It is quite similar to a toddler who absorbs all the information surrounding him like a sponge.
This is why ml is usually compared to a human who learns, in reality, machines are far from the human gift of learning, it being merely a simulation of the process. Even so, we are getting there, baby steps as they say. Famous AI researcher Ben Goertzel predicts that AI singularity (the point of no return for Artificial Intelligence when it will surpass human intelligence) will happen as soon as 2045, which is very close if you think about it. We will just have to wait and see for ourselves until then it is our duty as responsible engineers to ensure its safety and moral standards.
Besides being a useful tool and exciting technology AI presents a part of our human fantasy. Coming from sci-fi books as those of Isaac Asimov’s and Arthur C Clarke, the concept of a thinking machine has perpetuated imaginations for decades. It is an interesting prospect, to create and program a digital being that thinks and acts like a human, maybe even feels emotions, and why wouldn’t it, as far as we know the activity inside our brains is just the firing of neurons. Science has long known that you can map brain activity to mathematical operations, and computers are great at dealing with huge amounts of data and performing complex arithmetic on it.
Thousands of experiments and countless hours of research after the publishing of these books and no definitive result. We haven’t gotten closer to cracking the consciousness questions and the challenge of building AGI (artificial general intelligence) stands before us as prominent as ever before. This dip in the public interest has lead to what is now known as AI winter
The winter as they call it is a period in time when public interest has faded and no there was little to no funding in this area. Fortunately, in 2006 the University of Toronto develops Deep Learning, this is what many call the turning point which sparked interest again. This effort has lead to numerous new inventions and findings. We at Cortex truly believe that DL can revolutionize industries across the world and solve countless problems in human society and multiple other major international problems. Well if DL is so effective, then why haven’t we seen a conscious machine yet? Well, you see things are not so simple as they seem, DL is good at finding patterns and all the research before now targeted specific tasks — CNN’s for computer vision and RNN’s for text analysis are an example. Few organizations devote their time and effort towards building the omniscient and omnipotent machine that people envision when thinking of AI. Most organizations devote time and funding to make algorithms perform better on large amounts of data, optimize them, and develop tools to manage models. MLOps is an area of research that has stolen the spotlight of the ml world lately, following the path of traditional computer software.
So what is exactly AI, in the view of most of the population its an evil robot whose goal is to destroy the world or save it depending on who you ask. Computer scientists will tell you that it is a system that can automate human tasks and statistical models that predict some useful values, and everything in between. To be sure we’re on the same track here’s a diagram that shows the relation of all the aforementioned technologies, hope it offers a clear view of the subject.
As you can see AI is a general term or an umbrella term covering lots of technologies and methods, it is more of an idea and can be used interchangeably to mean many different things. Machine learning is a subset of artificial intelligence and covers the art of using algorithms on data, it is associated with data science in some sense, but the differences are not as clear, many use ml and data science mutually since they cover many similar themes. And then we have deep learning, a machine learning technique where layers of “neurons” are stacked to form artificial neural networks.
Deep learning in itself is a method of forming complex structures that perform many calculations, they are great at detecting special characteristics in data and especially good at finding patterns. Take for instance Convolutional Neural Networks, they are great at detecting little “deviations” and memorizing them. Essentially they execute multiple computations on data extracting only important values and feeding them further to the next layers. The larger the neural network the more operations it makes, increasing its complexity and hence its possibilities. However, don’t be fooled by this idea, if there is little data the neural network will perform poorly even with complex architectures, essentially starving it of possible outputs.
So where do we draw the line between sci-fi AI and real-world sentient digital agents? It seems as these concepts have merged over the years and modern pop science hasn’t helped in this regard. We need to go back in time to understand where the similarities begin and end between these two notions. In 1965 mathematician I. J. Good coined the term “Intelligence Explosion”, it refers to an Artificial Intelligence system that exceeds any known limits of knowledge, entering a cycle of an unstoppable learning cycle and reaching unprecedented levels of never before seen performance.
It is expected that such a system will accelerate and evolve exponentially approaching to a godlike figure, omniscient in some sense, therefore surpassing all the collective capacity of human intelligence. Of course, such a virtual being is just being hypothesized, there isn’t a single company capable of maintaining a structure for such a system, nor is there any progress of getting closer to understanding the necessary recipe for consciousness.
The current AI system can be described as really powerful statistical models that analyze vast amounts of data to find useful patterns in them, and the deep learning models are used for more precise applications as computer vision. You can imagine the advancement of classical AI that exists nowadays to meet the awaited super-intelligent one in the near future, somewhere around 2035 to be more specific.