The Advancement of AI - Are We Doomed?
Before we get into the meat and potatoes of AI, it is important to realise that the advancement of AI is inevitable. I have had met a lot of people who doubt the prowess AI may possess in the near future, they believe that the claims of AI sound far-fetched. Others who believe that AI can supersede the abilities of the human brain in the future have their doubts over the time frame by which AI will be capable enough to emulate human functions. The time estimates range from fifty to seventy years. However it is important here to realise that the time frame is not of much concern and importance as the acceptance of the fact that AI is only going to grow from here on and there will be a time in the near future when AI will be intelligent enough to carry out tasks which we currently believe can be only carried out by 'logical' human beings. People believe that machines don't have any thinking abilities, so how can they make progress. Machines however have had made phenomenal progress over the span of last few decades. We will address this issue later on though.
Curiosity is what has had driven mankind since it's inception. From space explorations to solving the mysteries of functioning of the human brain to finding the fundamental particles that make up matter, it has been our desire to know more and improve ourselves that has motivated us to carry out all these endeavours. And till the time that desire is there, there is absolutely no stopping. AI can not only help us in carrying out our daily jobs more efficiently, but it can also help us in making breakthroughs in sciences, including in finding cures to Cancer and Alzheimer's. To press on my claims further, we can refer to the laws given by Moore and natural selection. Natural selection, as many of you would know, is what has determined evolution to it's very core. Over time, biologists and psychologists have realised that natural selection has somehow influenced our thoughts too and because our thoughts govern our actions, we can certainly award the reason of our progress till date to natural selection. And because of the fact that we have had made progress up to now, despite many hindrances, we have no reason not to believe that we will only make progress in the future. To quote from Wikipedia, Moore's law is as follows:
Moore's law (/mɔərz.ˈlɔː/) is the observation that the number of transistors in a dense integrated circuit doubles approximately every two years.
It's been more than sixty years since Moore's Law was brought into existence in 1965, about a decade after the concept of Artificial Intelligence was originally conceived at the Dartmouth Conference in 1955, and he is yet to be proven wrong. Thus, unless the Doomsday is around the corner somewhere, the progress of AI is not going to be halted anytime soon. In fact, the rate of progress will only grow from here on.
Why do we need AI?
We have been stormed by an unbounded flow of information lately. Various estimates suggest that the amount of information we have been producing has been doubling every year. The graph below represents how the number of hours of video uploaded every minute on YouTube have increased over the last couple of years.
|Image credits: Tubefilter|
By the end of 2014, about 300 hours of videos were being uploaded every minute, which amounts to a shocking 432,000 minutes of videos uploaded everyday. That two-year old number would have only increased by now. Just take a moment and absorb this. Thus, just to watch videos uploaded in the last minute to YouTube, one need to spend 12.5 days. Obviously, no matter how much we divide this workload of processing this outrageous inflow of information, we cannot catch up with this kind of stellar pace of information inflow. Forget YouTube, as of October 2014, there were more than one billion websites on the internet. Apart from that, we click more than hundreds of thousands of photos everyday, some of which make their way over to social networks or cloud backup platforms such as Google Photos and Dropbox.
There is thus a tremendous amount of information that needs to be sorted out. One might ask why we need to sort all this information after all. Well, broadly, there are two reasons for this. First up, sorted information will make it a lot easier for us to find what we are searching for. This is very much like finding a book in a well stacked and ordered library. Books are sorted into various headers and then they are arranged as per their name or the author's name. Talking about books, a dictionary is the best example of sorted systems. Imagine looking for the meaning of a word in a randomly ordered dictionary. You will have to spend a lot of time just to pull out the meaning of one word, on the other hand, a correctly sorted dictionary will make this less tedious.
Secondly, a well sorted information makes it easier for us to process it and draw conclusions from it. Have you ever wondered why video suggestions on YouTube or content suggestions on Amazon more often than not end up matching your interests. Moreover, the ads that you see over here on AndroGuider or the related posts suggestions at the end of this article, these are all examples of machine learning. Machine learning is a type of artificial intelligence which allows machines to learn from the data that they are given as an input. Machines then process this data and draw patterns and conclusions from it. Google Photos had recently introduced a search feature. Using this feature, you could have searched for the images related to any keyword you entered, say 'hugs'. And if you would have hit the search button with the keyword 'hugs', Google would have correctly pulled out all the images showing people hugging each other. Do make a note here that it was a machine, not a human, which had automatically figured out that the people were hugging in the photo. You now might be curious to know as to how a machine learnt to identify two people hugging each other. This is where machine learning comes into the play. Quoting Margaret Rose's definition of machine learning:
Machine learning is a type of artificial intelligence (AI) that provides computers with the ability to learn without being explicitly programmed. Machine learning focuses on the development of computer programs that can change when exposed to new data.
Google's systems were given an input of thousands of images of people hugging each other. Through the means of machine learning, Google's systems were able to find out and learn about the various patterns which all those 'hugging' images had in common. From there on, whenever Google's systems encountered an image which had similar patterns as to the 'hugging' images, Google would automatically conclude that the image being dealt with would have two people hugging each other. This is how machine learning works, machines are inputted with a sorted data in the first place and then after processing the data machines are able to learn more artificially by figuring out the various patterns.
The applications of machine learning and artificial intelligence also extend into different domains such as cyber security. With the vastly growing number of Internet users, cyber security has emerged as another important domain. In fact, Google rolls out monthly security patches to the various Nexus, Pixel and Android One owners to safeguard them from multiple cyber security threats. To combat hacking attempts, organisations all over the world make use of various security systems too. With the help of various tools empowered by machine learning, modern security systems have given rise to Security Information & Event Management (SIEM) software products, using which organisations can foil hacking attempts. SIEMs scrounge through a vast data log and alert a security analyst in case they come across the familiar patterns of break-in attempts.
Artificial Intelligence is all about information processing. We have (and are on course of building) algorithms that interpret the abundant information we have in different ways. There is a reason why people prefer Google over other search engines like Bing and Yahoo. It is because Google has better search algorithms than the two. Over the next few years, we are only going to better our abilities of interpreting this information. As you read this, many scientists across the globe are working hard on various projects which intend to use artificial intelligence to make scientific breakthroughs. In fact, it might be AI that may help us invent cures to diseases such as cancer and Alzheimer's. Not only that, physicists believe that AI can help them in learning more about quantum gravity.
AI and machine learning will also help us in making our daily lives much more efficient. Virtual assistants like Google Assistant, Siri, Cortana, Hound etc have already learnt a great deal about natural language processing even in various voice accents. Various mechanisms of machine leaning will only empower these assistants further to assist us.
What does the future hold?
Artificial Intelligence is like a double edged sword. Before I get to my opinions on how AI may pose a threat to the human race, it is only fair enough to listen to the science experts, before nerds like me come forward with our opinions. Here is what the great physicist Stephen Hawking had to say about AI:
The last invention we will ever make and the last challenge we will ever face.
As described earlier too, the developments in AI are going to only progress from here on. If we are on the same page regarding AI's inevitable progress in the future, then it is a likely possibility that we will create some form of super-intelligence one day. This form of super-intelligence will have the capacity to improve itself too. However it is important to know here that the very reason behind AI's remarkable progress will be the unimaginable speeds at which it can process the information it is supplied with. And then by the virtue of machine learning, AI can further expand it's knowledge across various domains.
IBM Watson's victory in general knowledge game Jeopardy is an apt example of the pace at which future computers can work. Watson can process more than two hundred billions worth of information in a matter of few seconds. Competing with such rates of powerful processing is just beyond the scope of humans. Moreover, AlphaGo's victory over Go champion Lee Sedol showcased that machine learning is headed in the right direction. Go is a an ancient Chinese board game that is far more complex than chess and the number of probable moves in Go is more than the total number of atoms in the visible universe. Since Go is a game based more on judgement than feasible predicted moves, programming a computer to play Go is very tough and even if one is able to program a computer to play Go, the chances are very likely that the computer will still lose.
AI's super-intelligence can thus even help it achieve human-level general intelligence. When you come to think of this, it looks as if we are in the process of creating someone like God. And just imagine that if there is a conflict of interest between ours' and the future robot's, the future robots might think of wiping us off the planet. This is happening for the first time that we ourselves might actually be planning our own destruction, or better say extinction.
I share Sam Harris' opinion on this. We know that AI just might destroy us all and still we haven't done anything yet to guide the development of AI in the right direction, ensuring safety of humanity at the same time. The reason why we haven't been able to do anything about this till now is because we haven't been able to evoke the apt emotional response that is needed to counter this. In fact, death through the hands of man-made robots actually sounds fascinating to many people. I'm unable to marshal anything on the current atmosphere of fascination and excitement that many people have seemingly expressed. All I can say here is that this is the most weird corollary and analogy I have ever seen.
No matter how weird the climate of thought might be, it is still however extremely important for us to realise that AI is still not at that level where it poses a threat to us, we are the ones who will build AI. The very fact that WE would be the creators of AI is of great essence here. If we develop AI in the right spirit and bound it with restrictions that safeguard humans, we are definitely headed for a great future. We need to keep in mind why we are developing AI. We are building AI for our own benefit, not harm.
We thus need some kind of organisation that looks after all these affairs and governs the actions that are being taken up in bringing up AI. Hoping that we keep the principle reasons why we are building AI in our mind, I wish all of us a great future, if it exists.
The Advancement of AI - Are We Doomed? Reviewed by Krittin Kalra on 2/23/2017 12:29:00 am
Subscribe To Us
Get All The Latest Updates Delivered Straight To Your Inbox For Free!