Rep. Yvette Clarke (D-NY) is one of the handful of Democrats who has been trying to get ahead of the possible threats — some that may seriously disrupt the country’s elections and threaten democracy — posed by ever-more-rapidly evolving AI technology.
We are going to have the same problem with AI that we have with all forms of data and information, conservatives will use AI to spread lies and propaganda, just like they do now with all other forms of media they control. Conservatives are not honest participants in democracy because they are ideologically and demographically outnumbered, yet they refuse to temper their radical agenda to gain a legitimate majority. Their answer to minority status is to cheat and lie and manipulate.
It won’t be just conservatives abusing AI, not by a long shot. In this case specifically, I’m 100% certain both sides will do it. The text2img interface is just ridiculously accessible for everyone, and the temptation is too rich. Generating inflammatory images and videos is an obvious risk, but I’m more concerned about the deployment of finely tuned chatbots. These tools will be extremely diligent and persuasive with minimal investment, and average people are, to say the least, vulnerable.
In the world of The Golden Compass, everybody has a daemon, an accompanying spirit. Similarly, we’ll all soon be born with our own AI that we build out during our life. Your personal AI can be tasked with tussling with the disinfo AIs to reclarify political focus. The problem is low-information voters, not voters with too much information, even wrong information. Modern ballots in some jurisdictions can be multiple pages and a profoundly challenging reading test, especially if you’re not fluent in English or Spanish. An AI could be really useful in breaking such stuff down, and if nothing else, could go ask the League of Women Voters AI if it gets stuck. George Santos, for example, got elected in a low-information environment, with those choosing him party-cueing to the “R” by his name. A personal AI could have been asked to screen the local press about an unknown candidate.
You’re assuming that personal AIs will be under the control of the person. I’ve got this really nice offer of a free AI for you that will negotiate extra good deals and explain things just the way you want them explained.
The pool-pah is upon us, and it’s our duffle if we expect some granfalloon to save us.
Both sides do not do it equally, conservatives use free speech to lie and propagandize while liberals use free speech to tell the truth, AI will be no different. It always boils down to the character and honesty of the folks pushing the narrative.
I know the market for ideas and otherwise has to be excited about the new thing (and there better be one or one be made), but:
Intelligence is self directed.
AI is exactly what we don’t need in a world increasingly full of voters educated in schools that are run by people devoted to book banning and ideology. If the impact of AI can be solved, and I am not yet sure it can, it is going to be solved by an increased emphasis on teaching people how to evaluate the world around them.
Vonnegut pointed out that a “granfalloon” was also “any nation, anytime, anywhere”.
Clearly, you would need some sort of agency to intermediate between the AI builder and the AI owner/user. It would be a bleak future indeed if the AI builders owned your AI. The AI could take all your information and leave you unable to buy it back or even get access. Such things have to be regulated. Money is regulated through the banks as intermediaries. Energy is regulated through utility intermediaries. We haven’t yet reached a point where you can be thrown out of the airlock for not paying your Netflix bill, or TPM subscription for that matter. Moreover, by forcing the intermediary to carry risk as banks do, you at least fend off the ugliest predations under such a system.
There is a fix for that: 100% vote-by-mail like we have here in Washington state. You get a ballot with an info pamphlet two weeks ahead of the election, with time to study the issues before voting, instead of having to make a snap decision in a voting booth. You can verify online that your ballot was received and counted.
Not everyone will take advantage of studying the accompanying info pamphlet, but it solves a host of other problems like eliminating voter intimidation at the polls, and expanding the franchise for those with difficult work schedules or illness that makes in-person voting hard.
As a completely non-electronic system (until the ballots are received and counted) it isn’t vulnerable to AI interference. Until we get artificial general intelligence robots running around I guess, but it will all be over for humans by that point.
AI, meanwhile, is advancing at a ferocious speed, and experts warn that lawmakers are not treating this issue with the seriousness they should given the role the unprecedented technology could play as soon as the 2024 election.
Then we’re doomed. Because if it’s one thing humans have in common is not treating what should be non-partisan issues like, say, global warming ,with the seriousness they should be given.
who’s to say it will tell you the truth?
Good point. Personal buddy AIs are already here and will become more popular. They will act to reinforce whatever bias you already have, not correct your biases. Similar to the way the Facebook algorithm is designed to increase engagement by feeding you what it thinks you want to see. Your personal buddy AI will be like this on steroids.
There is a way to control this with legislation protecting personal data privacy online, at least for the large AI systems that operate remotely. The algorithms/AI can’t reinforce bias without knowing a shitload of information about your tastes, your history, your personal life, and who your friends and family are.
Of course this would kill the personalized advertising model that FB and other social media is built on, so those companies would lobby against it. But personal data privacy is a popular issue. I think consumers would support this, if Congress would get off its ass and do something about it.
One of my friends was really surprised to find out that ChatGPT lied to him! Truly we are babes in the woods.
98% of children admit to lying to their parents.
100% of parents admit to lying to their children.
Lying is intrinsic to the human condition.
If AI intends to pass a Turing Test it needs to lie.
