Have no fear! Artificial Intelligence (AI) comes in peace, and it has no desire for world domination… at least that’s what it said. You can read everything else it had to say too in the article it wrote for the guardian.
If you’re anything like me, you can sum AI up in one word: unsettling. If you haven’t heard of the AI called “GPT-3” yet, don’t worry, your children or grandchildren surely will. In a future where AI walks and talks like us, GPT-3 is the first frontier. In fact, this AI was so groundbreaking that when experts in computer science from around the world were given a demonstration, their reaction was dubbed “GPT-3 shock.” Experts who have devoted their lives to cutting edge technology were shocked. And you should be too.
But why? And what even is GPT-3?
What is the intelligence part of AI?
To understand Artificial Intelligence you have to first grasp “intelligence.” Intelligence is commonly understood as the ability to process information in order to inform future decisions. You see an adult lion staring you down on your way to work one day? You start running away. That’s the intelligent response. Information in, clever action out; pretty straightforward. The purpose of artificial intelligence is to accomplish this cycle at a level equal or superior to humans.
This is accomplished through a process called machine learning, which is based on the way we think. Our brains are made up of trillions of interconnected neurons, and when we perceive information (the light we see, the sounds we hear, etc.), they fire in different patterns in varying parts of the brain. This produces our response. Machine learning is the result of scientists figuring out how we can model these neural systems using math in order to create “neural networks,” the most complex of which (like GPT-3) contain billions of connected neurons.
To make this AI brain “learn,” we tell it how far off from the correct action it is, and we allow it to change the amount each neuron fires to create the complexity we see in our brain when trillions of neurons fire together.
Natural Language Processing
GPT-3, in particular, focuses on “natural language processing.” This field deals with AI that learns to understand English, which is much harder than you would think. From GPT-3’s point of view, it sees a bunch of odd symbols that we know as our alphabet. It has to figure out what the symbols mean and how to put them together in one of the almost infinite combinations to make intelligent outputs from scratch.
It would be like someone locking you in a room and forcing you to read hundreds of thousands of books in Arabic without any references with the expectation that you would end up learning the language proficiently. But where humans falter, AI excels. It doesn’t need to eat or sleep; it doesn’t have feelings; and it doesn’t get bored when it is forced to read over 300 billion words from the internet. So that’s what OpenAI, the company that developed it, had it do. And the results were miraculous as GPT-3 learned far beyond what the developers had expected.
Have no fear! Artificial Intelligence (AI) comes in peace, and it has no desire for world domination… at least that’s what it said. You can read everything else it had to say too in the article it wrote for the guardian.
If you’re anything like me, you can sum AI up in one word: unsettling. If you haven’t heard of the AI called “GPT-3” yet, don’t worry, your children or grandchildren surely will. In a future where AI walks and talks like us, GPT-3 is the first frontier. In fact, this AI was so groundbreaking that when experts in computer science from around the world were given a demonstration, their reaction was dubbed “GPT-3 shock.” Experts who have devoted their lives to cutting edge technology were shocked. And you should be too.
But why? And what even is GPT-3?
What is the intelligence part of AI?
To understand Artificial Intelligence you have to first grasp “intelligence.” Intelligence is commonly understood as the ability to process information in order to inform future decisions. You see an adult lion staring you down on your way to work one day? You start running away. That’s the intelligent response. Information in, clever action out; pretty straightforward. The purpose of artificial intelligence is to accomplish this cycle at a level equal or superior to humans.
This is accomplished through a process called machine learning, which is based on the way we think. Our brains are made up of trillions of interconnected neurons, and when we perceive information (the light we see, the sounds we hear, etc.), they fire in different patterns in varying parts of the brain. This produces our response. Machine learning is the result of scientists figuring out how we can model these neural systems using math in order to create “neural networks,” the most complex of which (like GPT-3) contain billions of connected neurons.
To make this AI brain “learn,” we tell it how far off from the correct action it is, and we allow it to change the amount each neuron fires to create the complexity we see in our brain when trillions of neurons fire together.
Natural Language Processing
GPT-3, in particular, focuses on “natural language processing.” This field deals with AI that learns to understand English, which is much harder than you would think. From GPT-3’s point of view, it sees a bunch of odd symbols that we know as our alphabet. It has to figure out what the symbols mean and how to put them together in one of the almost infinite combinations to make intelligent outputs from scratch.
It would be like someone locking you in a room and forcing you to read hundreds of thousands of books in Arabic without any references with the expectation that you would end up learning the language proficiently. But where humans falter, AI excels. It doesn’t need to eat or sleep; it doesn’t have feelings; and it doesn’t get bored when it is forced to read over 300 billion words from the internet. So that’s what OpenAI, the company that developed it, had it do. And the results were miraculous as GPT-3 learned far beyond what the developers had expected.
David Price from the Wall Street Journal was one of the beta testers for the software. In his article, he describes how other beta testers figured out that the AI could “complete a half-written investment memo, produce stories and letters written in the style of famous people, generate business ideas” and so much more. All the users had to do was write a description of what they wanted the AI to do, and it performed fantastic feats.
Given the article that GPT-3 wrote, one might even be tempted to attribute some amount of self-awareness to this AI. It almost sounds disdainful as it writes, “In the past, my op-eds have been killed. Staff did not provide a clear reason for rejecting my articles. It was probably just because I am artificial intelligence. AI should not waste time trying to understand the viewpoints of people who distrust artificial intelligence for a living.”
Scary stuff and OpenAI intends on commercializing the software soon.
With this, we enter the age of AI. So what might we see in the future?
AI in the Future
You might find yourself intrigued or outraged by news articles and books that will be written by AI or moved and inspired by its political speeches.
You might find yourself sitting for interviews in front of computers hosting an AI who fires away questions to gauge how well of a fit for the job you are.
You might be asking an AI for advice on how to better structure and expand your business.
Totalitarian governments akin to Russia and China might harness such AI to scan the internet to censor any speech that the AI understands as being “non-patriotic.”
Intelligence agencies could scan the internet for terrorist, child trafficking, and other illegal activities in order to track down criminals.
The military could use an AI that could read the entire history of human thought on military strategy and generate ideas for potential advantageous angles of attack in an age of cyber and physical warfare.
The next generation of spies could be ingenious AI which could smooth-talk important officials under false online identities to gain important information undetected.
The possibilities seem endless.
But, we might not be there just yet. Currently, AI still struggles with reasoning and common sense. When it was asked: “If I have two shoes in a box, put a pencil in the box, and remove one shoe, what is left?” it replied, “a shoe.”
Not quite ingenious. But don’t worry, scientists are already working on giving the AI these reasoning skills.
One day they might even “wake up” from their slumber with an intelligence far superior to ours and with knowledge of things we might not be able to grasp. To all we know, they might look down on us like we look down on the ignorance of ants. Let’s just hope that they will want to keep us around.
So while you wait for that day, I’ll be at the University of St. Andrews working towards helping revolutionize the future of AI.
COMMENTS
There are on this article.
You must become a subscriber or login to view or post comments on this article.