Artificially Intelligent LLC completes research on limitations of current AI systems and creates first AI with potential for super artificial intelligence.

THORNTON, CO, USA, May 6, 2023/EINPresswire.com/ -- Very few people understand what is possible today in artificial intelligence. Even the world's experts can only guess when a super AI will be created or if it is even possible. Artificially Intelligent LLC has conducted extensive research to determine if super AI is possible, and have discovered that it is and when it became possible. Using the research they have created a new AI capable of emerging into a super AI. The company's research starts at research conducted 64 years ago, at Cornell, where Frank Rosenblatt created the first component of AI, a working neural network called Perceptron. This was the first time a computer made a prediction, an educated guess based on its understanding of data. This changed everything, but it would be many years before our computer technology was advanced enough to utilize this turning point in our technological history.

Decades went by in repeated cycles of trying to do more with this technology and failing and trying again. But eventually, the hardware became available to create reliable and useful capabilities with AI, and the exponentially exploding race of artificially intelligent capabilities had arrived.

The typical approach to technology in general is like building a never-ending staircase. Each new achievement is adding another step to the staircase. Each engineer building the staircase adds their contribution on the concepts and achievements of previous steps. Even after 64 years, no one had created a new staircase and we were barely starting to make the idea behind the Perceptron into useful features. Will humanity eventually achieve super AI through this method, probably, but it will take far too much time. Almost everyone had failed to see the truth of what was learned back in 1958, and it seemed that no one really knew what to do with that knowledge. What was the truth then and now? The truth is this...

- The Perceptron unlocked the missing step to achieve super AI, the ability to learn and predict.
- AI is currently limited by its training data or actions list.
- AI is narrow, only good at one kind of task.

These are the problems that needed to be solved to achieve a system capable of advancing to a super AI. Making small advances until we eventually invent the final feature that pushes our AI across the finish line will take a long time.

Artificially Intelligent's approach is to unburden the AI with human bias, allow it to advance outside of its data and actions, safely, and document its advancements while it achieves super AI.

Artificially intelligent has solved the limitations of current AI and created an AI that can be guided to relearn our technological capabilities until it reaches the point that guidance is no longer needed. This process is like training a dog tricks. Trevor Chandler, CSO says, *Imagine we teach a dog how to sit, and then how to lay down, then we teach it how to roll over on its back. Next, we give it an objective of performing a barrel roll. The barrel roll is an emergent capability achievable through a combination of actions it already knows, laying down and rolling on its back. Our job to obtain a super AI is just to guide an AI across a set of objectives that are of known value to humanity, and to continue this until our needs are eventually met without guiding the AI to learn new specific actions. As this AI is learning, everything it does that is considered of value is persisted into a general memory that all subsequent systems can access and evaluate for their current goals. Between a code generating reinforcement learning system mapped to the full grammar of the python programming language and this ability to utilize all previously useful actions, we have a system that is not narrow.".

Trevor Chandler, Chief Scientist states, "These advances resulted in the first of its kind, a new AI, one not travelling the road of human limitation that drives on the single threaded road of the 400-year-old scientific method, but a new vehicle on the new superhighway to super AI."

A video of this new work has been included with this press release. In the video, the mazes being solved are insignificant, it is how the system does this that is incredible. The video depicts an AI that starts without any actions and a single environment and objective, nothing more. The AI expands its objective into multiple, similar objectives, and generates and evaluates its own actions in each environment until its original and all the expanded objectives are achieved. Every useful action it makes is persisted for subsequent clones to use for their current objective. Each new clone can choose the most effective action, either their own code-generated action, or an action persisted into the overall system by previous clones attempting to solve other objectives. Their CTO says, "If this sounds much more to you like how we learn to do things as human beings and how we use our previous experiences to solve new challenges, then you're catching on."

Trevor Chandler states, "As the world sees the mistakes that ChatGPT has made, it's critical that we think about safety when using AI, we shouldn't be reckless or move in haste. With my systems, I develop in air-gapped environments with a custom language created solely to define boundaries the AI cannot advance outside of. If you connect a water hose to a gun trigger and turn the water on and get shot, you don't write a letter to try to outlaw water hoses. This is the same with AI, we must control its access."

Trevor Chandler
Artificially Intelligent LLC
Trevor.chandler@researchgroup.ai
Visit us on social media:
LinkedIn

Emergent Self-Creating AI

Legal Disclaimer:

EIN Presswire provides this news content "as is" without warranty of any kind. We do not accept any responsibility or liability for the accuracy, content, images, videos, licenses, completeness, legality, or reliability of the information contained in this article. If you have any complaints or copyright issues related to this article, kindly contact the author above.