#TechTalksWithMelwyn – AI Winter is Coming!
Artificial Intelligence – AI. You can love it or hate it but you just cannot ignore it. This week on #TechTalksWithMelwyn I will get you to a higher base with AI.
Are you imagining a metallic robot with red eyes that is out to kill you? Don’t worry, that was the theme of my last post on Robotic Automation. I should probably tell you now that AI doesn’t hate you, then again, it doesn’t love you either!
Artificial Intelligence (AI) in Computer Science is the study of intelligent agents i.e. any system that takes inputs from its environment and takes actions to maximize its chances of achieving its goals. More commonly, AI is loosely defined as the ability of a machine to perform tasks that require human intelligence or cognitive ability. We have been training computers to do repetitive or mathematical tasks for us for over 50 years now. Why should this be any different?
Before we get into how you can convert your MacBook into an all-knowing superior being, let’s understand the types of AI and what capabilities they have. According to a Forbes article, there are 7 types of artificial intelligence but that is based on using 2 different methods to classify AI systems. The most commonly used classification breaks AI systems down to 4:
- Reactive Machines
- Limited Memory Machines
- Theory of Mind
As the name suggests, these systems are designed for a single purpose and work to maximize their ability to reach their goal. They do not have any memory of past actions and can’t learn from those actions. At every stage, they calculate their options and choose the one that gives them the highest likelihood of success. Potentially the most famous implementation of this system is IBM’s Deep Blue that beat reigning chess grandmaster Gary Kasparov in 1997.
These systems can’t learn strategies from previous wins or learn an opponent’s style. The only rules it does remember are the types of moves each chess piece can make and you must check-mate the King to win a game.
These machines have all the features of reactive machines and go one step further with an ability to learn from their past. Most of the AI systems currently in use from your Roomba to your digital assistant, Siri, are based on limited memory. While Siri can remember basic preferences like your default unit of measure and current city, he/she is not yet capable of having a follow-up conversation to any previous searches that you had.
The driverless cars industry is completely based on using limited memory machines that know what speed limits are, what a human crossing the road looks like, and how he/she is different from a vehicle in the next lane that is moving much faster. A major drawback of limited memory machines is the fact that it must be trained for every type of scenario it may encounter and is thus, limited to the universe created by its programmer.
Theory of Mind
In psychology, “theory of mind” refers to the ability to attribute mental states (beliefs, desires, mental states, etc.) to oneself and others, and people’s actions may be guided by these mental states. When it comes to AI, the understanding of human emotions is the next phase of systems that are either completely theoretical or work in progress. In order to reach this state, machines will need to understand and process human emotions and identify how these emotions guide our behavior.
If AI is ever going to make it as your co-worker in an office space, understanding emotions and reading social cues will be the deciding factor.
The last step in the evolution of AI is for the system to become Self-aware. This phenomenon is also sometimes referred to as the singularity where the system will be smart enough to self-identify and have its own thoughts, goals, and emotions, in addition to deciphering those of people around it. While the thought of such a system existing may send shivers down your spine, we are years, if not decades away from building a machine so sophisticated.
Folklore and fiction have long led us to believe that once AI becomes self-aware, its Frankenstein’s monster all over again. While this can be true in movies, more realistically, scientists believe that a super-intelligent AI will not love or hate but instead be programmed to reach certain goals and may not evaluate collateral damage the same way a human would. For example – an AI system trying to prevent global warming and the deaths of millions may instead build a dam for clean energy thereby drowning villages with a few hundred people. The goal here for scientists is to build an AI system that has goals that are aligned with ours!
Fear-mongering aside, on a funnier note, a research team at Harvard, recently proved that AI beats humans at making you laugh. The AI system predicted with higher accuracy what jokes a person may find funny than what their spouse or close friend could predict.
Although this may be a step in the right direction, other researchers have found that people are more likely to reject a good recommendation if they know it was made by a computer, than a bad/average recommendation from another human being. So is AI here to stay and become your co-worker? AI winter – a period of reduced funding and interest in AI has been a cyclic phenomenon ever since the discovery of AI simply because it never lives up to the promises and hype. For now, even the White House is funding AI initiatives such as the “Artificial Intelligence for the American People”. The only thing we know is that every summer that ever existed has ended, and for AI, “Winter is Coming.”
Fount of wisdom, insufferable know it all, make it go away are just some of the phrases used to define Melwyn. When he is not at his Consulting job, he spends his time reading about technology and current affairs.