Can Democracy Survive Artificial Intelligence? As the 2024 elections approach, we fear we will see an explosion of false and manipulative content generated by AI, especially by hostile foreign governments. concerns are growing.
The New York Times warned of deepfakes that could “shake elections.” The Biden administration has already warned against the use of AI, and the potential for election manipulation will put even more pressure on the White House and Congress to slow or halt its development.
But the real solution is to accelerate the deployment of AI. The strongest defense against artificial intelligence weaponized through elections is to apply it defensively.
Any discussion of AI and elections must begin with an uncomfortable fact. That means there is no way to stop AI from creating misinformation and disinformation. Powerful AI tools already exist, and as long as the United States has external enemies, it will try to use AI to influence elections.
The genie cannot be put back in the bottle, but more powerful techniques can negate its power.
The closest analogy is missile technology. The enemy builds and uses missiles, and we can intercept them with our own missiles. The most obvious application of this approach is Israel’s Iron Dome, which uses interceptor missiles to destroy incoming rockets.
That’s what America needs: an information steel dome powered by AI. Such systems leverage AI’s unprecedented pattern recognition capabilities to identify coordinated attacks using fake AI-generated content, such as articles, images, and videos.
seeing is believing:From Gaza to US politics, deepfake videos are spreading fake news
AI is more than powerful enough to play this role. Consider efforts over the past few years to train AI models to detect lip-sync deepfakes.
Humans have a hard time saying words that start with M, B, or P without closing their mouths (try saying “peanut butter crumbly”). However, deepfakes often ignore this reality. Although a casual observer may not notice it, researchers have found that the trained model can identify these instances with surprising accuracy. Their model, trained on videos of former President Barack Obama, detected deepfakes over 90% of the time.
AI is better at spotting deepfakes than humans
Given the potential volume of such content, identifying AI-driven attempts to influence elections would require a large human workforce, suffer from human error, and be costly and costly. It will take time.
In contrast, AI tools can scan the internet much more quickly, efficiently, and effectively and label deepfakes as soon as they appear in public view.
To use a rocket analogy, it’s the difference between a high-tech integrated missile defense system and a swarm of soldiers aiming shoulder-mounted rockets into the sky. There is a reason why Israel chose Iron Dome.
Private companies need to drive the development of this cutting-edge technology. Putting election-specific AI in the hands of politicized institutions will invite abuse and undermine the very democracy we are trying to protect.
We can save democracy.Will the 2024 election tear America apart? Here’s how to keep it that way.
It would also be unwise to give government agencies vast regulatory powers over AI and elections. The resulting burden is likely to stifle innovation and leave technology development in the hands of a few large companies.
It would also undermine Americans’ faith in democracy, given widespread concerns about bias within certain models.
AI development should be encouraged across the broadest and most diverse range of companies possible. Fierce competition will encourage companies to avoid ideology and offer the most effective products.
Tech companies can develop the tools needed to protect elections.
American companies are the world leaders in AI, so they are well suited to develop this technology. Startup DeepMedia is already helping the Pentagon detect deepfakes that threaten national security, including one that told Ukrainian President Volodymyr Zelensky to surrender his troops.
Further development of this same technology could protect elections by helping voters spot deepfakes when they encounter them.
Fast forward to Biden vs. Trump.Does anyone else not feel like the 2024 election has begun? Can we just keep going?
Similar technology discovered that a famous work of art by a previously unknown artist was actually a masterpiece by Raphael, while another AI program discovered subtle clues throughout the artist’s known paintings. Based on this, we determined that some of the works by another Raphael were likely painted by someone else.
If AI can decipher the secrets of the Renaissance masters, it might be able to spot deepfakes from Russia.
Even rudimentary AI programs need help: think of the troll farms in North Macedonia and the Philippines. You will quickly identify that relatively low-tech election interference efforts remain in widespread use.
As my colleague Neil Chilson, a former chief engineer at the Federal Trade Commission, explained in recent testimony before the Senate: Cheap fakes, carefully selected edits, foreign content farms, and plain old Photoshop are cheap and effective enough. ”
Just like the real Iron Dome protects you from artillery shells and rockets, the AI Iron Dome can help you identify these threats even faster.
Uncertainty caused by regulatory threats and continued calls for a “pause” on AI development have hindered maximum innovation in the area of election protection. If policymakers make clear that companies can act within current law, the benefits are almost certain to be felt before the November election.
Foreign countries such as Russia, China, and Iran are acutely aware of the power of AI, and are guaranteed to utilize it in their election manipulation efforts, even if they have not yet deployed it.
If America did anything other than fully unleash AI to protect democracy, it would be a particularly foolish unilateral disarmament.
Christopher Koopman is executive director of the Center for Growth and Opportunity at Utah State University.
