Welcome to Neural’s beginner’s guide to AI. This multi-part feature should provide you with a very basic understanding of what AI is, what it can do, and how it works. The guide contains articles on (in order published) neural networks, computer vision, natural language processing, algorithms, artificial general intelligence, and the difference between video game AI and real AI.
As legend has it, a reporter once asked Mahatma Ghandi what he thought of Western Civilization. His response was “I think it would be a good idea.”
The same sentiment could be applied to artificial intelligence if you compare it directly to human intelligence. That is, the most advanced AI systems in the world (DeepMind’s, GPT-3, etc.) pale in comparison to a human infant’s intellect: artificial intelligence would be a good idea.
Thankfully for everyone in the industry, the rubrics we use to measure machine intelligence are entirely different than the ones we apply to ourselves. It can be difficult to suss out what “AI” or even just “intelligence” means from one source of information to the next.
But the reality isn’t all that complex. Humans experience reality through a theater of the mind. We inherently define our existence by the time, place, and sensations we observe. That’s a fancy way of saying we have imaginations.
We experience time as a UI for memory, place is defined by where we think we are in relation to things outside our observation (that we’re assuming still exist), and sensation is just one of the many languages our brains speak. Our experience of reality, our basis for intelligence, is like a movie that lasts as long as we live.
Computers experience intelligence as an exponentially unfolding series of ones and zeros. We can reverse engineer any current AI system (because we’re the original engineers of all AI systems) and we ultimately drill down to ones and zeros. (Quantum algorithms not withstanding).
And, though we still haven’t sorted out all of the human brain’s mysteries, it’s safe to say we’re not binary thinkers.
That’s the simple explanation. But it doesn’t clear up much when it comes to what AI can and can’t do. Because, binary or not, it doesn’t seem too far-fetched to imagine humankind could be one or two eurekas away from inventing a sentient machine that is capable of imagining things and forming its own theater of the mind.
Yet, to be clear: no current AI system we’re aware of has the ability to think or imagine. This theoretical idea for an artificial stream of consciousness is the closest thing we can find.
AI can’t do much right now. But what it can do, it does extremely well. Deep learning – computer vision, natural language processing, and similar disciplines – excels at mundane tasks that would take humans too long to do.
There’s no way you or I could search through 75 million images to figure out which ones looked like cats. Despite the fact we’d perform the task with far more accuracy than any algorithm, we wouldn’t live long enough to finish the job. An AI might do it in seconds.
So when you hear something like “AI can diagnose cancer with 97% accuracy,” the reality is this: they taught an AI to look at the pixels in a photo and tell us where Waldo is. And “Waldo” is just whatever it is that oncologists look for in images when they’re searching for signs of cancer.
But deep learning isn’t the only form of AI there is. Thousands of researchers are developing new classes of algorithms, advanced neural networks, and hybrid learning technologies designed to better imitate the human brain.
In the meantime human intelligence and machine intelligence simply aren’t comparable. However, in the future, technologies such as quantum AI, hybrid approaches involving symbolic AI, or new class calculus could go a long way towards spanning that gap.
Published December 11, 2020 — 20:00 UTC