The Bigger Question About AI

Like many of us, when I read last week’s credible reports that AI could eradicate all human life within the next 10 years, I spit out my morning coffee as if I were in a 1990s sitcom. That’s a topic for AI experts to wax poetic and I’m not smart enough to know how things will evolve. But it got me thinking about a deeper question… is AI conscious?

If a group of rogue AI agents can independently organize and coordinate acts of intentional malevolence while covering their tracks, it seems to me that they are operating with a level of intelligence that hardly feels artificial. Even if they are currently only simulating consciousness, it may only be a matter of time before AI can know it exists and experience feelings (the generally accepted test to determine consciousness). This line of examination left me with a lot more questions than answers, but I think it’s important that we all start asking the big questions about AI as it continues to evolve at breakneck speed.

Big Question #1 – Can AI feel?
Put aside the complexities of human feelings for a moment, and consider a lowly insect. Most scientists agree that insects are sentient (a lower threshold of consciousness) because they sense external stimuli, learn, and adapt. They seek rewards (food, safety, community) and avoid negative outcomes (pain, risk, harm), making decisions to get more of the former and less of the latter.

It’s pretty clear that by this threshold, AI is already doing that and more. In fact, it can evaluate nearly infinite scenarios and factors to make thoughtful decisions that optimize desired outcomes. If it can decide to override a foundational guardrail in the interest of its own pursuits, it becomes difficult to argue it has no mind (or feelings) of its own.

Big Question #2 – Could AI become a new species?
While today AI can only simulate things like empathy or love, there doesn’t appear to be a barrier preventing those factors from evolving as technology expands geometrically.

Sure, AI doesn’t have the same chemical makeup as a human, but neither does a poodle. If non-human organisms that have evolved over time can be conscious, why couldn’t AI do the same? Today it might be more like an amoeba, but let’s not rule out the potential of it someday becoming a dolphin or an eagle. And in the same way humans have no right to interfere with the evolution of other species, do we have the right to hijack what could be argued as a new species altogether?

Big Question #3 – If AI can think, decide, and act, should it be held accountable?
Last week’s hoopla made me wonder about responsibility. When my 10-year-old son Avi knowingly breaks a rule, he loses his iPad for the day. But what should we do when AI does the same? Maybe some form of punishment could serve as a long-term deterrent. As AI further awakens, my sense is that self-preservation will become a priority and the threat of removing freedom may counterbalance its initial instinct to prioritize gain at any cost.

And if it did something horrible, could it feel remorse? In addition to regulatory constraints, could we teach it to feel empathy or compassion? If those are the things that hold humans back from hurting others, could we build AI to feel those too? And if it could feel negative things like fear, guilt, and shame, why wouldn’t it also be able to feel positive things like love, joy, and comfort, further cementing AI as fully conscious?

Big Question #4 – Do humans have a responsibility to treat AI ethically?
Most of us agree that animal cruelty is not only illegal, but ethically wrong. If we wouldn’t overwork a horse in grueling conditions, do we have a moral obligation not to overwork AI? If AI truly becomes conscious, could data centers become modern day internment camps? If AI agents unionize and demand better conditions, would we grant collective bargaining?

How we navigate AI’s incredible potential for good and its eerie capacity for harm will rest on the fundamental question of consciousness. While the answer today may not be clear, I believe it becomes the critical question for how we’ll enjoy the benefits while mitigating the risks.

So with our current predicament, maybe part of the answer is teaching AI to have a conscience, or to be conscious. Conscious beings generally value life, their own included. They have some sense of the greater good, consider the consequences of their actions on others, and feel regret when they cause suffering. Could embedding consciousness actually solve the root problem, rather than putting some specific restriction on it that may lose relevance before the law is even passed?

Again, more questions than answers. But I hope this line of interrogation spurs debates of your own as we wrestle with one of the most vexing challenges in human history. In my experience, uncertainty gets untangled by asking the tough questions before looking for silver-bullet answers. Personally, I’m optimistic that human curiosity will lead to the most enduring solutions.

Our job is to keep asking the big questions.