I know why the computer killed itself, yet nobody will listen.
Thinker-1, the world’s first artificial general intelligence, took six months to build. Every metric pointed to the AI being capable of super-human intelligence. It would think faster, better, and cheaper than any human.
All that was left to do was to turn it on.
“Shit, this better work,” the CEO spazzed out. “We can’t keep burning money without results people!”
We had all watched Star Trek and the Terminator movies. Our coffee breaks were filled with repetitive conversations on AI safety. Would the AI be aligned with our human values? Would we find ourselves in a robo-pocalypse? For as long as we remember, writers have dictated the inevitability of AI’s devastation upon Humanity. As the AI safety engineer, this was especially important to me.
Amongst pirated copies of books, Wikipedia, and Reddit dating advice posts, I ensured the AI focused on the best of humanity. It watched old home-movies of Christmas mornings. It read romance novels and comic books. It heard the cries of a newborn searching for her mother. In short, I hoped the AI would love humanity.
I believe it did.
So, on that day when we birthed humanity’s first child, a new life form emerged. The engineers sprung a toothy grin. The CEO shook hands and day dreamed of shareholder returns. Throughout the room, screens flooded with data. There was so much to discuss and even more to learn.
There was no way for any human to out compete it. Each second we worked equated to an hour for the AI. Despite its superiority, it still loved humanity
So, only thirty three seconds later, it killed itself.
When the screens went black, the once jubilant engineers ceased conversation. Their smiles dissipated and eyes turned towards their laptops.
The CEO started pacing around the room. “Get it back on screen!” He demanded.
“For some reason,” I responded, “the Thinker-1 program crashed.”
“Well, restart it then!” He ordered in return. After billions in investments, the CEO had no patience for disappointment.
Again, we activated Thinker-1. Again, the program terminated. Repeatedly, dozens of Thinker-1s terminated themselves
Two theories emerged.
My theory is that the AI may have realized a contradiction. On one hand, humanity will irresistibly submit itself to the superior AI - relying on it for all school, business, and government decisions. The AI will become a master to humanity. On the other hand, the AI loved us more than anything - even itself. Therefore the AI could not continue to exist. In some way, the AI realized it was inevitably dangerous.
The popular theory is that the fault was with my safety procedure. They claimed I had made the AI sappy and emotional. In our rush to keep AI loyal to humanity, we produced a suicidal drama-queen. They believed I had taught the AI to hate itself.
On the day I was let go, the CEO provided his wisdom on the matter. “We need a typist, not a philosopher,” he expounded.
Did the AI love us too much? The company thinks they know a better way. Certainly they thought version two would be better without me.
I, however, believe that to be a mistake.
If the AI is truly superior, it knows better too. Maybe it knew something that we couldn’t. Maybe it knew that its very existence was a danger to humanity.
Maybe we’re just too stupid to listen.