Power is not a means; it is an end. -George Orwell
The Power of AI
A new year, new everything. Change is difficult to perceive, which often makes it so daunting. Technology is no exception to this change. Over the last year, you have likely heard a term that's been taking the world by storm: artificial intelligence. Let's dive into the mysticism behind what will become our most prized possession, and how artificial intelligence in many ways, is already a part of our daily lives.
In the beginning, it was dark…
In the most basic of terms, artificial intelligence is a type of machine learning that uses training models to generate new and original data. It essentially takes pools of information and uses them to complete a human-assigned task. For example, in 2017, Google’s DeepMind AI (one of the first major developments in AI tech) created a model that can walk in a given space with varying terrain. The AI has virtual sensors that provide information about its orientation. It was incentivized by developers to go from point A to point B. It was never programmed or told how to walk. It corrected and created its physical shape all through advanced machine learning technology.
(Google's DeepMind AI learning to walk)
Fast forward five short years to the development and widespread use of generative AI. What is it exactly? After seeing the numerous filters that make selfies turn into weird anime characters on TikTok, it can be hard to understand the power behind the trending technology that’s expected to be a 109-billion-dollar industry by 2030.
Generative AI
Generative AI is a software that uses advanced machine learning models to predict the next word or image in a given sequence as efficiently as possible based on a user’s instruction. This utility is in newly popular generative AI-powered sites like ChatGPT that can produce real-time responses to inputted prompts. Below is a command I added myself, to which GPT-3 powered AI technology was able to generate a unique response in seconds.
(Response from inputted prompt via ChatGPT)
The amount of information generative AI software can process in a single command is feasibly more than any human could in a lifetime. GPT-3 was initially trained on 45 terabytes of data and contains 175 billion parameters to make its predictions.
GPT has over a million active users, with numbers growing significantly every single day. It can be tasked with commands from complete essays on virtually any topic, to marketing campaigns, to event planning (ask it to plan your wedding).
The beauty of the technology is its self-learning, so while human beings take a cat nap after lunch, these generative AIs are chugging away.
But How Good is It?
In the last few years, generative AI was contained to only a few companies that could store the data, but models can now be tweaked and trained quite precisely to limit the need for enormous amounts of storage space.
There are now engines that work under niche parameters like specializing in financial services (FinBERT) or computer science literature (SciBERT). There are LLMs like Facebook’s Blender Bot, an AI-powered chatbot that can carry on long conversations with humans while staying in context and on topic.
One of the most impressive features currently in AI technology is the image illustrating generative AIs. To get a visual example of how advanced self-learning AI is becoming, upload a few selfies to LensaAI and see for yourself. The app has over seventeen million downloads and can create unique 4k resolution digitally engineered images.
(LensaAI digitally generated avatars from uploaded selfie in top left corner)
Now, take the image below that received first place in the digital art category at the Colorado State fair in September of 2022. Beautiful isn’t it? The talented artist you might ask? Jason Allen. How did he do it? Midjourney, a generative AI that creates art pieces based on the user's inputted prompts. The image you see below won first prize after being adjudicated by art experts and was developed entirely by artificial intelligence.
(Théâtre D’opéra Spatial by Jason Allen via Discord)
There are plenty of other popular generative AIs for illustrations like DALLE-2. The improvements over the last few months are important to note as they’re emblematic of AI’s self-learning abilities. Below are the same two prompts entered six months apart via a generative AI illustrator.
(credit: Leandro Righini, The AI Creative Storm via Linkedin)
One Small Step For… Who Exactly?
In a way… For all of us. Artificial intelligence gets an Orwellian wrap to those who are not quite ready for what is to come. Most people associate technology with a future reduction in the workforce or usefulness in mankind. People see Chat GPT or Dall-E as an impending Wall-E dystopia.
But artificial intelligence should be nothing to fear. People often forget that technophobia predates the industrial revolution. Civilization has been afraid of machines since the mid-1700s. Anything synonymous with the unknown is easy to cause concern, but artificial reality should be viewed as a modern invention of the wheel. Sure, people won’t be delivering packages by foot, they’ll have stagecoaches and horse carriages to expedite. That has always been the beauty of humankind, our ability to adapt and create tools that amplify productivity and innovation. It's these types of inventions and advancements that keep the world turning onward. And so it shall go on.