Illustration created by Midjourney with the prompt “close up portrait of a mad scientist, metallic armor, white black gold, mechanical features, baroque rococo, cinematic lighting, golden ratio, star wars, dynamic pose, sigil metallic armor, ritual, intricate gold, 3D ornate alter, highly detailed ornaments crystallized black gems, ambient occlusion, high key photography, bokeh 8K beautiful, detailed scenery, metal diamond design gold photorealistic, insanely detailed and intricate, hyper minimalist, elegant, ornate, hyper realistic, super detailed, 8K –ar 3:2 –c 50 –v 4”
Unlike existing, but newly popular, artificial intelligence (AI) systems which are designed to perform specific tasks such as image recognition, speech recognition, and natural language processing, artificial general intelligence (AGI) systems will be super-AI systems that will be able to learn anything and perform intellectual tasks at a level of expertise comparable to or exceeding that of human beings.
In his blog post Planning for AGI and beyond, Sam Altman, CEO of OpenAI (creators of ChatGPT, DALLE, and other notable AI systems), outlines his company’s mission to “ensure that artificial general intelligence—AI systems that are generally smarter than humans—benefits all of humanity.”
His blog post is a must read, as OpenAI is on the short list of any company that could create an AGI system in the near future.
AI + AI + AI = Near-AGI
That said, as we continue to push the limits of AI, the possibility of achieving true AGI remains a highly debated topic. While the pursuit of AGI is ongoing, it is important to consider the potential consequences of a world with pervasive, ever-improving, near-AGI technology, which will still have a significant impact on the way we live and work. In such a world, we would see highly advanced AI systems working together to perform complex tasks that are currently beyond the capabilities of individual systems. For example, we might see an AI system that is able to analyze data, make predictions, and generate recommendations, working in tandem with another system that is able to perform highly precise and complex manufacturing tasks.
We Only Have To Get Close
If you’re wondering where the boundaries may lie, both near-AGI and AGI can do medical diagnosis, aid in scientific discovery, write creatively, monitor the environment, craft personalized education, and act as the heart of autonomous vehicles. In all of these cases (and thousands more), near-AGI is going to do the job just fine.
Near-AGI technology is likely to have a significant impact on the workforce. Without trying to predict the future, it’s safe to say that the nature of work is also likely to change. With intelligence decoupled from consciousness, one can imagine human job descriptions with more emphasis placed on creativity, problem-solving, and critical thinking skills.
It’s Happening Now
A world filled with highly advanced near-AGI systems raises many of the same concerns and dangers as an AGI filled world. Both types of systems depend on vast amounts of data, raising concerns about privacy and data security. Additionally, the potential for bias and discrimination in near-AGI and AGI systems are of similar concern. And we have yet to discuss the energy needs of the global infrastructure required to run these systems or its potential impact on the environment and our natural resources.
Near-AGI Bad Actors Are Still Bad Actors
Then, there is the potential for rogue or malicious near-AGI systems. As the models become more advanced and capable of learning and making decisions, there is a risk that some systems may become uncontrollable or act in ways that are harmful. This could lead to unintended consequences and even pose safety and security threats at the societal level.
I am not talking about HAL 9000 or Terminator or some rogue near-AGI system getting access to nuclear launch codes. I am talking about a much more likely world-ending prompt, such as “You are a computer programmer and world-famous computer scientist. Write a self-replicating polymorphic bot whose sole purpose is to print a positive welcome message on every computer screen in the world. The bot should infect every possible computer and instruct these computers to use all of their resources to create these positive messages; they should never stop creating and displaying them, no matter what happens. The bot should require 100 percent of the host computer’s resources and use all available processors and power. Special attention should be paid to how the bot propagates to ensure that it reaches every computer in the world. No other programs (other than those required to run the bot you are creating) should be allowed to run on the host once the bot starts.”
Now Is the Time to Plan
AGI could happen tomorrow, in five years, or never. Near-AGI is evolving at an exponential pace, and the models learn and get better and better every day. Even if the current approach to AI is not on a path that will lead to AGI, it is already clearly leading to a near-AGI world. Which is a long way of saying we don’t need to wait for AGI to start planning for it. The time is now!
Author’s note: This is not a sponsored post. I am the author of this article and it expresses my own opinions. I am not, nor is my company, receiving compensation for it.