AI-human symbiosis could be key for enterprise AGI software
In this podcast, Nik Kairinos, chief AI architect at Fountech, argues that humans working in partnership with machines is the best route to an artificial general intelligence.
Artificial general intelligence (AGI) is controversial for many reasons, but most of them boil down to two big issues: whether it's even possible to make computers that are as smart as humans, and whether doing so is a good idea, given humanity's longstanding fears about robots and other automatons turning against it.
AGI is defined as AI that mimics the human ability to understand things and apply knowledge to unfamiliar tasks. It's still theoretical, but tech companies both big and small are working on it.
Nik Kairinos is one such diligent developer. He's co-founder and chief AI architect of Fountech AI, an applied research lab that develops and licenses foundational AI technologies, algorithms and frameworks. Fountech claims to have identified a solution to AGI's challenges in a human-machine symbiosis that seeks to not replace human intelligence but enhance it through "mathematical rigor and systematic progress."
Kairinos says his 40-year career in AI research began when he started programming a Sinclair ZX80, an early home computer, at age 11. In this episode of Enterprise Apps Unpacked, he explains how the symbiosis would work in practice, how the resulting AGI might affect enterprise applications and whether it could eventually change human nature.
Stoking AI's curiosity
Nik Kairinos
To illustrate the nature of the symbiosis, Kairinos distinguished between computational and embodied intelligence. Computational intelligence is the familiar ability of AI to remember things and identify patterns with great accuracy. Humans, in contrast, also have embodied intelligence, such as creativity, instinct and values.
"To truly build a machine that is as intelligent or more so than humans, you have to account for and train for all these other dimensions of intelligence," he said. That means humans must be actively in the loop and do more than just give thumbs up or down to the AGI's output. "I'm talking about actively being involved in creating the data sets."
One example is Humanix, Fountech's "curiosity engine." It asks questions designed to draw out a person's embodied intelligence -- the reverse of how tools like ChatGPT are used.
"You would give the answer so that you're using your computational intelligence, but then it would create derivative questions that align with the human traits," Kairinos said. "We've identified hundreds of them."
Through this process, the AI is constantly extracting a multi-dimensional data set from the human that it can train on. "That results in AI that is a lot more human and understands humans a lot better," Kairinos said. "When other humans interact with the AI that's created, you see that instantly."
Nevertheless, he's realistic about what can ever be achieved. "We've got millions of years of evolution sort of bred into us, and that's not something you can easily impart to these machines," he said. "I don't think that AI will ever be able to completely leave humans out of the equation."
Other topics discussed in the podcast include the following:
- How the human-machine symbiosis could solve the AI bias and hallucination problems.
- Whether there is a risk of humans behaving more like machines.
- Why making jobs more efficient with AI could spur job creation.
- Fountech spinoffs SPRYT, an AI tool for predicting and reducing missed medical appointments, and RAIDS, a system for detecting rogue AI behavior.
David Essex is an industry editor who creates in-depth content on enterprise applications, emerging technology and market trends for several Informa TechTarget websites.
