'We should be scared': Experts warn of advanced AI impact on US elections
Experts have warned about the potential impact of advanced AI systems that could generate realistic fake news at unprecedented levels as 2024 the U.S. presidential election approaches.
Software for voice cloning from ElevenLabs and photorealistic photography from OpenAI's DALL-E 2, Stable Diffusion, and Midjourney could potentially be used for misinformation, according to a report by Fortune Magazine on Saturday.
"We should be scared shitless" about the threats that the huge language models that underpin the technology offer to democracy said Gary Marcus, emeritus professor of cognitive science at New York University and an expert in artificial intelligence.
"It is hard to see how A.I.-generated misinformation will not become a major force in the next election."
Although the technology for making films from text prompts is still in its infancy, deepfake experts can create convincing fake videos.
Chris Meserole, a fellow at the Brookings Institution, cautions that voice clones could play a significant role in the 2024 elections, allowing audio clips to surface at crucial moments in the election, claiming to be a recording of a candidate saying something scandalous.
While not everyone is convinced that the situation is as serious as Marcus suggests, he does, per the Fortune report.
Fact-checking AI
Large language models akin to those that power ChatGPT are being used in Spain by Newtral, a business that specializes in fact-checking assertions made by politicians, to identify and disprove lies.
Ruben Miguez Perez, chief technology officer at Newtral, thinks that these technologies can identify when a piece of material is making a factual assertion that needs to be verified and evaluate the chance that it is disinformation based on the sentiments it conveys.
Although Vallas, a candidate for mayor of Chicago, was unaffected by a phony audio in which he appeared to minimize police shootings, experts caution that AI-generated false material will play a significant role in future elections.
These technologies have the potential to undermine trust and confuse people by making them doubt the accuracy of information provided by experts and social connections.
The scenario will need further consideration and research to combat the possibility of widespread, machine-generated misinformation, even if AI technology may potentially be the best protection against these instruments.
For more, you can read the full report here.