AI isn’t a human-like machine; it’s a machine-like child

Artificial Intelligence (AI) is riding an ebullient wave of innovation, investment and popularity. Across a range of revenue technology categories, AI has permeated predictive analytics, sales automation and customer engagement — and that’s for starters. But, it’s easy to forget that AI as we know it is still young.

Children, very much like AI, offer endless amusement and constantly astonish us with their new skills. They also worry leaders with their limited reasoning power due to still-developing frontal lobes and susceptibility to external manipulation. AI worries experts for that very same reason.

It is time to stop thinking of AI as a human-like machine and start thinking of it as a machine-like child.

AI, like children, is a product of its training and environment

Children learn a lot from their parents and from formal schooling. They learn from their peers, from TV, from books and from play. Sometimes they learn useful skills and knowledge. Other times they learn absolute nonsense.

The education process children experience is not unlike that of machine learning, which uses a combination of supervised, unsupervised and reinforcement learning approaches. Unintended results range from mildly amusing to potentially harmful.

For example, one of the more serious criticisms aimed at AI recently is that it has somehow “learned” to stereotype and discriminate based on gender, race and socioeconomic status. Here, it is important to note that AI is a product of its training —and its training is imperfect.

AI is trained on digital assets that reflect the same biases and blind spots that exist in human society. Some biases can be fixed within an iteration, while others will take generations.

AI, like children, needs guidance and guardrails

For all the recent outrage around ChatGPT generating eloquent, but factually incorrect outputs, fallibility is hardly unique to this one AI tool. Al learns from its mistakes, like children. Early adopters of AI should not write off the technology for not being perfect, just like you would not completely dismiss children that make mistakes at, say, a piano recital or a spelling bee.

AI may not feel the psychological effects of ridicule that children feel, but an overly derisive reaction to AI’s shortcomings will dampen adoption and development, precisely the things it needs to improve its accuracy and breadth of knowledge.

It’s impractical to avoid using AI until it is perfect. Likewise, it’s impractical to control unsupervised learning. As with children, using too firm of a hand will stunt AI’s learning ability and will be counterproductive in the long run. To take advantage of AI’s capabilities even in its current level of maturity, business leaders using AI-driven solutions must:

• Carefully choose use cases that play to AI’s strengths.
• Apply human inspection commensurate with risk tolerance.
• Feed AI highly accurate data with as much contextual detail as possible.

AI, like children, needs to be nurtured 

AI’s ability to surpass humans in certain skills will not diminish humanity any more than my kids will diminish me by becoming more talented and successful than I am. I can’t wait to see how my kids will improve the world, just like I can’t wait to see how AI’s applications will improve humanity.

Scientists, business leaders and civic leaders must band together to nurture AI’s capabilities and enforce thoughtful guardrails. The rest of us should embrace AI’s potential, but with the same care we would use with a young, precocious helper who still has a lot of maturing to do.

This blog was republished with permission from the Gartner Blog Network.