In today’s era, if you haven’t heard about AI or Machine Learning, then dude/dudette, what the heck have you been doing? Having sex under a mountain cave with a charming beast that resembles a banana? Let’s dig a little deep on this topic and try to understand what this new thing is that people think is going to takeover humanity – for real!
So, with the introduction out of the way, let me clear myself. I have been researching on AI for past 36 hours, have written 30 papers on AI, all blank. Have a lot of knowledge about.. well, nothing and yet, I’m gonna explain AI to you! Well, maybe I can. Maybe, you don’t know anything about AI at all, and my slightest knowledge can help you understand AI better.
Inception of AI – Lazy Scientists wanted to be lazy and Productive
Well Scientists are lazy, and computer scientists are the benchmark of that. Otherwise, why wouldn’t someone play football? Or become a pop star? What motivates them to choose a career that has literally zero satisfaction and nobody gives a shit about. Seriously, when was the last time a girl wet their pants for scientists or engineers?
Oops! except him! Except Elon. Please. I am sorry.
Back in the days people had dreams, optimism and in 1965 one of them called H. A Simon said “machines will be capable, within twenty years, of doing any work a man can do“. He had to raise funds I know. But this is what computers in 1985 looked like.
I won’t even call it “ambitious”, though humans did land on the moon by 1969, yet didn’t make any progress in AI, not until 1997 when IBM Deep Blue defeated the grandmaster in chess! in the second attempt!
Today, a smartphone looks like this
Just to prove a point, this smartphone is 48 times more powerful than the IBM Deep Blue, just the GPU, Adreno 540 is 567GFLOPS while Deep Blue had 11.38GFLOPS and it has two AI assistants, Bixby and Google Assistant. At the same time IBM Deep Blue costed $100M in R&D, and Galaxy Note 8 probably didn’t cost as much as it is a consumer product and the cost have been recovered.
So what is an Artificial Intelligence (AI) anyway?
Glad that you asked. or I made you ask, to be honest. Artificial Intelligence is a concept where man-made (or machine made) machines are able to demonstrate learning, self awareness, knowledge, imagination, emotions, reactions, mistakes etc traits which we commonly associate with humans. These are part of Intelligence (please let me know if I missed something). And if we can make machines do it? We create Artificial Intelligence.
But today, AI usually is a program that doesn’t have to be coded specifically. We have a code that learns from examples. If we want to ask it to recognize a banana, we show it a thousand photos of bananas and it learns how to recognize bananas. If we want the same program to recognize mosquitoes, we show it a thousand pics of mosquitoes.
But true Intelligence will mean they have a lot of the characteristics listed above and will be able to demonstrate them.
Have we made it?
Are we close?
Does Kim Jung-un has a hydrogen bomb?
Yep!!! And I am shit scared of that!
Why haven’t we made one true AI yet?
Because it’s complicated. Making one is undeniably hard, but even if we make it, the short term benefits might outweigh the long term implications! There are some things to consider.
- Some people think AI may grow strong enough to kill everyone. On the other hand, some belief that it’s “just a software” and if we can code some hardcoded shit, it won’t go berserk. I believe that we can’t do that. There are far too many edge cases to code properly. We can’t even code a Code compiler without errors, what makes you think we can hardcode all the rules properly?
- Also, intelligent beings are independent. For example, twins who have been raised by same parents, went to the same school, came from the same set of chromosomes. Let alone personalities, they don’t have 100% identical genes!
- Outside stimuli are random which we can’t predict. We can’t predict what an AI is going to encounter. And we can’t predict what kind of reaction it might have. We can’t know it’s priorities.
Why are we pursuing AI at all then?
Again, It’s complicated. We think Artificial Intelligence is the future, and even more powerful than Hydrogen Bomb, Nuclear Fusion and everything else. Cause, AI can potentially solve those problems for us (and create a few maybe). But it’s mostly due to three reasons
- It doesn’t have to be so dystopian.
In all of Science, every technology that supposed to mean a utopia (good thing), was also used as a dystopia (weapon). But maybe AI can change that. Maybe AI, being intelligent can stop everyone from killing each other. When we read history from a thousand years, we say, so many wars fought for so petty reasons? A thousand years from now, (if people exist), they’ll look at us, and say, they killed so many lives for some petty reasons like religions, and beliefs and greed for power and shit like that. Maybe an AI can stop being the puppet of its owners and can take a stand for a greater good!
- All AI’s can’t even kill.
So, AI are of three types.
1. ANI (Artificial Narrow Intelligence): This kind of AI helps us solve a specific type of problem and doesn’t understand meaning. They are the working class, don’t argue, only obey.
2. AGI (Artifial General Intelligence): They are kinda rebels, they ask questions and have self awareness, can understand context and this is where the problem starts.
3. ASI (Artificial Superintelligence): Once an AGI forms, it will constantly improve itself. And become so intelligent that it will know virtually everything. And it is what we call an ASI. A being which is vastly superior to us and we have reasons to believe that it will kill us.
- We can’t go back now. Its too late.
With rising participation from teams such as OpenAI, Google, Facebook, Microsoft and Amazon, it’s impossible for anyone to stop it. The only thing that can help is making more and diversified AI Systems that are incapable of having General Intelligence. As long as we stay more intelligent. We reign supreme and AI won’t kill us.
How are AI designed??
Not easy. I tell ya. But if you really want to know, just read the free book called “Neural networks and Deep Learning” by Michael Nielsen. It’s informative, descriptive and if you can push through, you’ll learn it. But before, you need to learn about computers and coding (which I want to prepare you through my series).
Simply put, AI are not very complex things by themselves (yet). Because most of them depend on Statistical Regression, Interpolation and Extrapolation, Backpropagation and uses simple techniques such as Sigmoids etc.
Here, the term simple is relative. If you have no idea how coding and algorithms work, don’t know what is regression, interpolation, extrapolation etc, then it’ll be modestly complicated. But the term simple here reflects that, compared to other complex programs such as Computer Vision, NLP, CAS etc usually employs far more difficult and complex methodologies.
If you’re have no life, is a nerd, go for AI, it’s possible. If you have a life. Go get a girlfriend instead. If you can manage both, teach me how to do it!
Machine Learning (ML) vs Artificial Intelligence (AI)
Machine Learning is the ability for machines to learn from examples. The way we learned to connect seperate ideas into one abstract idea by observation, this is the process of learning. This enabled us to learn more. And in machine learning, we ask the machines to look at sample data, then ask it to find the correlations and learn to do things. This has enabled us to do things such as recognize images, play games etc.
Artificial Intelligence is the ability to transfer the learning into knowledge. We can see that if we drop something, it falls, but what did we conclude? The earth pulls things. This complex reasoning of the human mind gives rise to intelligence. If earth attracts, then the mars also attracts, so does the sun. In that case, they also attract one another, which means, due to the Sun’s attraction planets revolve. There goes a rudimentary planetary theory.
We can transform observations into ideas, generalize and build upon existing knowledge, this kind of abstract relationships form our intelligence, and incorporating them into a machine would imply an Artificial Intelligence.
How far have we gone with AI
Well, we have gone pretty far with ML but our understanding of AI is still rudimentary. Our achievement in AI includes
- IBM DeepBlue Beating Chess champion in chess.
- Google alphaGo beating Go champion in Go.
- OpenAI beating world champion in DOTA2.
- Google Translator understanding context.
- Facebook Image Recognition understands the picture.
- Google understands conversational context (Google Assistant)
- Google AutoML is able to judge seperate Neural Networks, and improve itself at the same time. An improving AI is judging other improving AI.
AI can be scary. Any new technology at this moment can be scary, but our fears should not hold us back, it should give us a cautious view into the world, fear is a good thing. We can stay alert. Things can go wrong, but let’s just thing the endless possibilities AI would bring us. Like and share this post with your friends if you liked it. Thank you.