Good [time of day], fellow NSG-denizens.
Recently, I was listening to an episode of Hello Internet, a podcast made by CGP Grey and Brady. On this episode, they talk about the dangers of artificial intelligence, and how artificial intelligence can pose an existential threat to humanity. Grey points out that, even under perfect circumstances, where an AI researching facility is burried deep underground inside a faraday cage, with no internet port access for a thousand miles around, an AI could convince its human caretakers to plug it into the internet regardless, and thus gain access to the whole world of computing.
Brady, meanwhile, offers an alternative. AI are dependent on computers and the internet, and he theorises that we could hypothetically stop a rogue AI if we unplugged the entire internet, or even cut all power on earth for a bit. That would hopefull extinguish an artificial intelligence before any major damage could be done. "If there is blood in the streets, could we not do it?"
That made me wonder about this hypothetical, which I want to posit to you. Imagine a company creates an AI and gives it access to its entire database and asks it to 'maximise shareholder profit'. For convenience of the hypothetical, let's imagine this AI does have an off-switch, and it is located at the company headquarters. If this switch is switched, the AI will be deleted entirely. In reality, this would not be so easy, but we are simplifying the problem to its essence. The AI has knowledge of this off-button, however, and it will do whatever it can to stop its own deactivation, unless the board of investors votes to turn it off, at which point it will allow someone to switch it off.
The question: in a capitalist system, can this AI be turned off, and if its possible, how likely is it?
This question goes to the root of the problem with solving existential threats, such as climate change, water scarcity, and weapons proliferation. Can we take drastic actions to save our planet or part of it, while threatening profits, and expect companies to accept their loss of revenue?
I doubt that, in this hypothetical, the board of shareholders would choose to turn off the AI. Since its goal is to maximize profits, it will do so in a way that will not hurt the PR of the company too much. However, that does not mean it could not result in unspeakable yet largely unseen suffering. This suffering, however, would largely be felt by those without shares in the company. I see no reason for this board of investors to choose to unplug the AI, just as many companies have chosen to do nothing or extremely little about climate change. The board of investors, itself run by the various investment firms that form the financial sector, which are in turn controlled by shareholders, can only make decisions that are in the interests of their own profits. So I don't expect them to make that choice.
Governments could take action, of course, but they would need an incentive to do so. Many governments have shown a general lack of willingness to really tackle the problems of climate change, not in the least part because of corporate donors of neoliberal parties, but also because doing so would mean limiting GDP growth in the short term. As this is bad for government revenues and general approval figures, governments will refrain from taking action that would upset the market.
Even if the government, or more likely an activist group, would take it on themselves to shut off the AI, the AI and the investors would try to protect it against interference. They are, after all, still making money off of it. The cost of trying to shut off the AI, which is trying to defend itself, could be incredibly large, and cost a lot of lives. Knowing this, governments might refrain from taking action at all, and activists would not get the public support necessary to put the company HQ to the torch. Even if they did, they would most likely be prosecuted for doing so.
In the end, while theoretically possible, I believe that the economics involved in the individual actions of those involved would mean that a profit-maximising AI, even if its activation had horrific consequences, would never be turned off. I think this speaks to a broader problem with capitalism, where short-term profits are velued over the long-term viability of any system, or even the mid-term collateral damage. Problems like climate change and world hunger cannot be solved by the market, and in fact, the market prevents viable options from being pursued.
But what do you think, NSG? Could this AI be turned off? Would it be by activists, the government, or the board of investors? Is capitalism capable of solving existential issues? I would love to hear your opinions.
The episode of the Hello Internet podcast: https://www.youtube.com/watch?v=jmOBm-Lcs70 (AI part starts at 1:09:30)
