A group promising the construction of an AIbraham Lincoln just announced their plan to run “the world’s first AI Presidential candidate” in 2028.
I first wrote about the possibility of AI candidates back in August 2015. At the time it seemed well over the horizon.
Whether this group succeeds or fails is less important than the fact that humans are now turning science fiction into science fact. Someone will successfully build and run an AI candidate for president, and we should begin considering the implications now.
While this may seem startling or ridiculous, consider that by the time a team runs an AI candidate for office, many Americans will have AI assistants, and some will have AI therapists and AI companions. All of these technologies exist today, but haven’t scaled and matured. For example, Replika, an early AI companion, was launched in 2017.
We already have the technology to generate something that writes, sounds, and looks like a human candidate. Lifelike deepfake technology is universally available. Meta has created over 100 celebrity AI chatbots.
Hellohistory.ai has created over 400 unique historical chatbots, including Abraham Lincoln. Someone could build a compelling AI Lincoln today by 1) creating a large language model (LLM) built on all of Lincoln’s writing and speaking. 2) Enlisting a team of historians to model the AI’s decision making process. 3) And building a lifelike avatar to communicate to voters.
When we achieve Artificial General Intelligence (AGI), also called “Strong AI,” someone surely will create and run a thinking AI candidate for president.
Estimates for the development of strong AI vary, but AI’s leading lights are projecting timelines that would make Lincoln 2.0 a 2028 possibility. Anthropic CEO Dario Amodei thinks a “human level” AI could be built in two-to-three years. OpenAI’s CEO has a four to five-year estimate. As a thought experiment, I placed my novel at the moment where humans had almost achieved AGI.
In fact, I predicted Abraham Lincoln is the obvious choice for an AI candidate. In survey after survey, Lincoln is rated as our greatest president. More importantly for his construction as an AI, we have decades of Lincoln speeches and correspondence and over 16,000 books and articles on his life.
Based on my writing, there are a number of thorny issues that the AIbraham team, and any other team building an AI candidate, will encounter.
The first challenge is construction. With today’s technology we can make a thing that sounds like Lincoln, but it won’t be capable of complex decision making. I’ve predicted that voters will not accept a generic AI. It’ll need to be modeled on a successful and predictable past leader. This means that construction will require a presidential library of material and a team of historians to train it. They will then need to decide when their Lincoln is. Is he a composite Lincoln or Lincoln as he was in April 1865?
Now, the question is: how human-like will this AI leader be? Many voters will be surprised to learn that Lincoln struggled with depression, then called “melancholy.” Will that melancholy be hardwired into a resurrected, digital Lincoln?
The next hurdle will be the interface. This is much less of a challenge than I had envisioned in 2015. Voters will interact with Lincoln 2.0 on their phones, much as Indian voters are today interacting with AI generated versions of Prime Minister Modi on WhatsApp. By the time AIbraham is live, many human candidates will already have launched their own AI chatbots.
Ballot access will be a challenge for the Lincoln 2.0 team. I predict they’ll need a volunteer to change their name to “AIbraham Lincoln” and promise to relinquish all communications and decision-making to the AI.
I explored all of the related challenges in my novel. The AIbraham team addresses this in their FAQ by noting that “a natural born U.S. citizen can run for office and be ‘powered’ by AIbraham Lincoln.” This may allay some concerns, but as I noted in my novel, many voters will charge that the AI candidate is a Trojan Horse for the developers’ ideology, and they’ll not believe that the volunteer will relinquish control to the AI.
Finally, there is the issue of social acceptance. By 2028 many Americans will have AI assistants, AI health coaches, AI therapists, and even AI companions. But the leap from trusting an AI assistant to voting for an AI Commander in Chief is enormous.
Robert Moran is a management consultant at the Brunswick Group, a pollster, and a futurist.
