Artificial intelligence (AI) is increasingly permeating the world around us and rapidly changing our lives. It offers some very exciting opportunities, but in some cases, it can also be more than a little scary. And without a doubt, the big development in AI that’s making waves right now is generative AI.
As the name suggests, it is an AI that can create videos, music, computer applications, and even entire virtual worlds from words and images.
What makes generative AI different and special is that the power of machine intelligence is available to almost anyone.
We are accustomed to using AI-powered applications and tools in our daily lives. Google uses this to find the information you need. Amazon uses it to suggest things we should buy. Netflix uses this to recommend movies. Spotify uses it for music – all powered by AI.
But a new generation of generative AI tools goes even further, giving us the power to build and create in amazing ways. With a little practice, you can even use them to build your own AI-powered apps and tools. This is the long-awaited beginning of breaking down technological barriers. democratization A.I.
So in this article, we’ll give you a quick overview of why it’s so powerful and what you can do with it. We’ll also look at how it works from a technical perspective, but most importantly, why it will change the world and what everyone should do to prepare for it. I will explain.
What is generative AI?
As used today, the term AI refers to computer algorithms that can effectively simulate human cognitive processes, such as learning, decision-making, problem-solving, and even creativity.
It is in this last, and perhaps most human quality, that generative AI comes into play. Like all modern AI, generative AI models are trained on data. It then uses that data to create more data according to the rules and patterns it has learned.
For example, if you train it with a picture of a cat, it will learn that cats have four legs, two ears, and a tail. Then, when you tell it to generate your own cat image, it will create as many variations as you want, following some basic rules.
One distinction worth understanding is between generative AI and discriminative (or predictive) AI. Identification AI primarily focuses on classification, learning the differences between “things” (for example, cats and dogs). This is a recommendation engine, like the one used by Netflix and Amazon, that helps differentiate between what you might want to watch or buy and what you might not be interested in. Or used in a navigation app to distinguish the appropriate route from A to B. Something to probably avoid.
Instead, generative AI focuses on understanding patterns and structures in data and using them to create new data that resembles it.
So what can generative AI do?
The first use cases for generative AI typically involved creating text and images, but as the technology became more sophisticated, a world of possibilities opened up. Here are some of them:
Image: Many generative AI tools – etc. The middle of a journey or stable diffusion – You can take a natural language (i.e. human language) prompt and use it to generate an image. Tell them you want an image of a two-headed dog in an Elvis costume flying a spaceship into a black hole, and watch it (or something close to it) appear before your eyes.
Text: ChatGPT is probably responsible for starting the intense hype around generative AI at this point, but there are other generative text tools out there like Google. bard And Meta’s llama. You can use it to write everything from essays and articles to plays, poems, and novels.
Audio: Generative AI tools can be used to create human-like voices (text-to-speech), allowing computers to speak words never before spoken by humans, as well as music and sound effects.
Video: Although not as sophisticated as text or image generation, tools are starting to appear that allow you to create and edit videos simply by describing what you want to see.
Data Augmentation: Generative AI allows fully synthetic data sets to be used to train other AI models that follow real-world rules without imposing privacy and data security obligations on those who store and use the data. can be easily created.
Virtual environments: Think virtual reality (VR) environments or video game worlds that you can explore and interact with, or the much-hyped concept of the Metaverse. Designing these is a very complex task, but it can be greatly accelerated with the help of generative AI.
How does it work?
Generative AI, like all the AI we see today, comes from an area of AI research and practice called machine learning (ML).
While traditional computer algorithms are coded by humans to tell a machine exactly how to perform a particular job, ML algorithms perform better at that job the more data they are fed. .
These algorithms are put together to obtain a model so that new data can be generated based on what is learned. This is essentially an engine tuned to produce a specific type of data.
Examples of models used in generative AI applications include:
Large-scale language models (LLMs) – Learn semantic relationships between words by ingesting large amounts of text and use that data to generate more language. An example of an LLM is GPT-4, created by OpenAI, which powers the ChatGPT tool.
Generative Adversarial Networks (GAN) – These work by pitting two competing algorithms against each other. One task is to generate data similar to the training data, and the other task is to determine whether the output is real or generated. This type of generative model is typically used to create images, sounds, and even videos.
Variational autoencoder – This is a model that learns how data is constructed by encoding it in a simple way that captures its essential properties and understanding how to reconstruct it. It is a type. Often used to generate synthetic data.
Diffusion models – These work by adding random data (known as “noise”) to the data being trained on and figuring out how to remove it while preserving the original data. This way, you will learn what is important and what can be discarded. Diffusion models are most commonly used in image generation.
Transformer Model – This is something of an umbrella term for a group of models, including LLMs, but covers any model that works by learning the context and relationships between different elements in the training data .
Generative AI in action
There are already many amazing examples of generative AI being used to create amazing (and sometimes terrible) things.
Please pick up a Coca-Cola masterpiece For example, advertising, a collaboration between human artists and AI, will bring many of history’s greatest works of art to life on screen like never before.
It is also used to create . Beatles new song Lyrics partially recorded by John Lennon are reconstructed and combined with new songs by Paul McCartney.
Generative design is a term that describes the emerging field of using generative AI to create blueprints and production processes for new products. For example, General Motors uses a generation tool created by Autodesk to new seat belt bracket It is 40% lighter and 20% stronger than existing components.
It is also being used to speed up drug discovery, with one UK company recently announcing it had developed the world’s first AI-powered immunotherapy. cancer treatment.
Generative AI is also the technology behind the recent phenomenon of deepfakes, which blur the line between reality and fiction by making it appear as if a real person did or said something fake.
Deepfake Tom Cruise is one of the earliest and most famous examples. Even more insidious, potential candidates on both sides of his upcoming 2024 US presidential election are Starred With deepfakes aimed at discrediting them for political purposes.
Spreading propaganda is bad enough, but there are also outright criminal uses, such as staged extortion of money. kidnapping hoax The act of using a cloned voice to impersonate someone and defraud them of money. company CEO.
Ethical issues surrounding generative AI
Generative AI clearly enables amazing things, but it’s clear that its existence also poses some difficult questions and challenges for us.
Perhaps one of the biggest questions is when will there come a point when it will be impossible to distinguish between the real thing and what is generated by AI?
Given the incredibly fast pace of innovation in this field, it’s likely to happen sooner rather than later.
This leads to the question of what (if anything) to do about it.Countries including China have already passed bill Make it illegal to deepfake people without their consent – should the world follow suit?
And there is also the question of how this will affect human work. If the companies that employ creators could simply tell a computer to create as many images, sounds, and videos as they want, would their livelihoods be in jeopardy?
Another issue that needs to be addressed is copyright. If AI is used to create a work of art, who owns it? Who is the person who used AI to create the art? Author of AI itself? Or all of the (probably) thousands of artists whose work was used (actually, often without permission) to train AI?
All of these questions will need to be answered in the near future, given the accelerating pace of development of this technology. How we answer these questions could play a key role in determining the future of generative AI in society and our lives.