Will AI Kill Us All?

Or Will We Use It to Kill Each Other?

Image created by Google Gemini

Science fiction has certainly prepared us for this doomsday. AI will come in some form to wipe humanity off Earth. Maybe we have been subconsciously waiting for that “future” to finally arrive. And now we are too ready to just accept that as part of our evolution and then extinction.

But wait …

Let’s look at all the things that humans have made that have the potential to kill on a massive scale or cause human extinction. The first that comes to mind is nuclear weapons; then there are bioweapons and chemical weapons. And the new toy humans have made is AI.

Let’s not forget that the same areas of science and technology that can produce terrible weapons can also produce enormous human benefit. Nuclear power, vaccines, medications, antibiotics, gene therapies, fertilizers, and sanitation are good examples. AI, too, can serve humanity in many useful ways such as education, healthcare, agriculture, research and data analysis, and monitoring climate, to name a few.

We want AI, like any other advancement in science and technology, to be capable, but we must be careful with giving it the authority to act on its own.

AI can have the capability to design cancer treatments without having the authority to inject anybody with anything.

AI can be very capable in nuclear physics without having the authority to command nuclear systems.

AI can be very capable in financial analysis without having the authority to access bank accounts.

So …

Who gets to decide what science and technology are used for?

Shouldn’t we have a say in something that could kill us all? Or are we just going to let the few corporate executives make that decision? Or heads of governments? Or military leaders?

Shouldn’t we put together a governance model that represents all humanity? Globally?

With all the weapons of mass destruction, we understand that they need extraordinary controls and restricted access. Nuclear materials are controlled. Biological weapons are prohibited by international treaty. Chemical weapons are prohibited. Nuclear proliferation is constrained. Laboratories handling dangerous pathogens have biosafety and access controls.

Not perfectly. Many people live in fear of someone with access to these systems, feeling justified in pressing the button one day. But at least we understand the importance of multiple layers of control and international cooperation.

Why should AI be any different? Why on Earth would we give AI the keys to the existing weapons of mass destruction to use when it thinks it must?

It all comes down to how we design AI. What objectives we set for it and what level of autonomy we allow it.

An autonomous AI trying to achieve the wrong objective may very well decide that humans are in the way. But it is us who give it that objective and the authority. It is us who make the design decisions.

Preventing AI’s progress is not the battle we need to fight.

We need to unite and fight the right battle. We must regulate the design, decide the right goals and objectives for it, and control what decisions it can make and what actions it can take.

How do we choose the right objectives?

Much like humans, if we set the goals for AI to beat competitors, dominate industries, defeat enemies, outperform everyone else at all costs, then we have created a beast that we won’t be able to beat.

But if we set the goals for human and machine alike, as increasing knowledge, learning and understanding, promoting curiosity, securing peace, creating art, beauty, joy, and happiness, we will then get very different results, both in humans and AI.

Winning should not mean someone else’s loss. Winning at all costs cannot be the norm. We must redefine ourselves and AI alike. Because we are creating AI in our own image. The question is which image? Some images of humans are very ugly. We don’t want the same ugliness recreated in AI. That ugliness will be much more dangerous when it is stronger and unstoppable.

We must change our own incentive models and culture towards a peaceful world in order to have peaceful AI.

Note: This essay is part of my ongoing series, Reclaiming Humanity, reflections on dignity, meaning, technology, and the future of our shared world.