A Boulderite, a politician and a robot walk into the bar. Such is the origin story for what follows.
OK, not really. But it’s pretty close.
In early November, Boulder Progressive board member Brian Keegan sent me a fake “interview” with Boulder City Councilman Mark Wallach that he “conducted” by feeding writing samples to an artificial intelligence chatbot and then asking it to answer questions about the recent elections — in “Wallach’s” voice.
Keegan feels that ChatGPT “definitely nails a lot about the tone and positions” and “distinctive communication style” of Wallach, whom Keegan calls “Boulder’s most loquacious elected official” in an email to Boulder Weekly. But, ultimately, Keegan says, the experiment was a failure.
“I struggle with what this substantively demonstrates or contributes,” he writes. “Predictable anti-progressive grievances? Inflated self-importance? Technology replacing politicians?”
How does the real Mark Wallach feel about his impersonator? “Very concerned,” he wrote in response to the A.I. interview.
“It is more than a bit disconcerting to hear answers that I might have made without actually having made them,” Wallach says. “The implications for using this technology in the distribution of political disinformation are obvious. It does not depend on which side of the issues you want to take: The possibility that a machine could authoritatively appear to be me in responding to these questions — when I have not done so — should be frightening to anyone.
“Today it is me; tomorrow it could be someone who wishes you ill and distributes a fake interview you purportedly responded to in high school, just before that critical job interview. Or, as here, it could be a set of bogus answers to an interview you never gave while actually running for office. The possibilities are endless.
“The good news,” Wallach continues, “is that ChatGPT does not have it down quite yet. ChatGPT can spit out words, but does not grasp who I am and how I would respond.”
One notable shortcoming (which Wallach and Keegan agree on) is that AI is just not funny enough to pass as the man it is purportedly emulating.
“I often use sarcasm and humor to show the ludicrous outcomes of some policy suggestions,” Wallach writes. “This program would not recognize humor if it were cracked over the head with it. Its answers to the questions are invariably lugubrious and stodgy; it is as if the questions were being answered by a Soviet Commissar in the 1950s.”
What follows is the generated interview, created by Keegan using GPT4 and GPT3.5 Turbo models and “fine-tuned using a corpus of City Council’s Hotline emails and guest opinions in the Daily Camera” written by the flesh-and-blood Wallach. Questions were written by Keegan. Robot Wallach and Human Wallach’s responses (to the AI’s answers and Keegan’s original questions) are both included and clearly marked to avoid confusion or misattribution.
We hope this is an instructive and illuminating example of the potential — good and bad — and limitations of this emerging technology. The final word we give to Real Wallach:
“The program is really not there yet. But who knows what it will be like in a year? Two years? Five years? Some of the responses that are almost laughable today may soon become far more credible — even convincing. That may be a good result for an automated service center, but it is a very bad one when we are talking about our political institutions and political communications.
“A number of years ago, during the Arab Spring, it was inspiring to see how activists armed only with phones using the internet could actually bring down a corrupt government. All these years later, we have an internet that is a cesspool of commercialism, hatred, misinformation and worse. Things have not worked out quite as planned. And here comes yet another new technology, filled with promise.
“Will that promise be realized, or will we be going down another path to uncertain ends and untold damage to our institutions and political processes?”
What do you think about the most recent election where voters elected another 6-3 progressive majority to City Council?
ChatGPT: While progressives might claim to champion the cause of equity, inclusivity and progress, they often lack a comprehensive understanding of the intricate dynamics that govern a city's growth and development. The election is a lamentable outcome demonstrating a collective amnesia or, worse, a willful ignorance of the essential principles that have long underpinned Boulder's unique character and success. Experience, expertise, and a balanced perspective are crucial for effective governance, something that these so-called progressives need to learn and appreciate.
Wallach: I would never call the election results “a lamentable outcome demonstrating a collective amnesia or worse, a willful ignorance of the essential principles that have long underpinned Boulder’s unique character…” That sentence left me asking: Who the hell is that guy? I have been involved in politics for 50 years, winning a few and losing more than that. Blaming the voters for their decisions is idiotic and something I would never do.
What kind of mandate can be claimed from an election where voters passed the “Safe Zones 4 Kids” measure but also kept a progressive majority on City Council?
ChatGPT: The paradoxical mandate from the electorate is a clear indication that the voters are not entirely aligned with the progressive agenda, despite their majority presence in the council. The support for “Safe Zones” indicates a desire for targeted, practical solutions to specific issues rather than a wholesale endorsement of progressive policies. This mandate, therefore, should be seen as a call for balanced governance, where progressive ideals are tempered with pragmatic considerations.
Wallach: [This] is probably the closest in substance to my actual views. But if I ever responded to a question like that in such a boring manner, I should be arrested for the crime of creating gross tedium. When did I become such a stuffed shirt? It just ain’t me.
Given your opposition to ranked choice voting (RCV), what do you think about the outcome of the mayoral election using ranked choice voting for the first time?
ChatGPT: The mayoral election is a prime example of how ostensibly progressive ideas can lead to convoluted and problematic outcomes. RCV is an unnecessary complication that serves only to muddle the electoral process rather than enhancing democratic choice. While claiming to better reflect the will of the voters, RCV often confuses them and can lead to results that do not accurately represent the majority's preference. It's essential to scrutinize whether the results truly reflect the electorate's will or if they are merely a byproduct of a convoluted voting mechanism.
Wallach: The program attempts to replicate what it appears to be my undying opposition to Ranked Choice Voting. Not so fast, AI.
This city voted to institute RCV for our mayoral contest; the vote was decisive. That is the end of it for me. I do not spend any time worrying about the decisions of the community — that is above my pay grade. It worked well this election and that’s about all that matters. The program is extrapolating an answer to the question that, because of the community’s decision and the fact that it worked smoothly, is simply no longer true.
You endorsed six candidates for City Council and Mayor but only two won their races. What does this say about popular support for your agenda?
ChatGPT: The fact that four of my endorsed candidates did not secure seats does not necessarily constitute a wholesale repudiation of the principles I stand for. This is a momentary snapshot of the electorate's mood, not an immutable verdict on my agenda. It is essential to recognize that voter decisions are often swayed by a complex interplay of factors, including campaign strategies, candidate quality, voter turnout dynamics, and the limitations of the electoral system itself. It is a leap to interpret this as a comprehensive endorsement of their entire platform or a categorical rejection of mine.
Wallach: Well, the top two vote-getters were endorsed by me, and a third lost by less than 50 votes. There is as little significance to that as to the fact that my endorsement lacked effectiveness. We have a new Council, and I will deal with it as I find it. I do not regard members of the Council as my enemies. We disagree at times and, when possible, we will agree. The real answer to the question is: “Just fine, thanks.” Again, AI fails to grasp my actual personality, as opposed to its theoretical construct of that personality.
How do you feel about Tara Winer overtaking you as the candidate who received the most votes for City Council in recent memory?
ChatGPT: While an interesting statistical footnote, this does not diminish my accomplishments on council. In politics, as in any sphere of public endeavor, records are set and subsequently broken; this is the nature of any dynamic system. It's important to understand that the number of votes a candidate receives is not always a direct reflection of their capability, experience, or the substantive quality of their agenda. Effective governance is not a popularity contest; it is about making tough decisions, often in the face of opposition, for the long-term benefit of the community.
Wallach: I can only state that I am extremely jealous, and I am sure that she cheated her way to the top. This is an insult to our democratic process and to me personally! Uh, no. If I were asked that question in real life, I would have to pick myself off the floor from laughing. The fact that ChatGPT actually attempted to formulate a serious response to that question is an indication of its weakness.
How do you plan to govern as a member of the minority in the coming term?
ChatGPT: Firstly, my role in the minority will be to scrutinize the policies and initiatives proposed by the majority, ensuring that they are subject to rigorous debate and analysis. Secondly, I will utilize my platform to articulate a clear, rational counterpoint to the progressive agenda, ensuring that the public is aware of the potential pitfalls and long-term consequences of their policies. Finally, it is imperative to ground policymaking in timeless principles of measured growth and preservation so that Boulder's unique characteristics are not undermined by the progressive majority's imprudent enthusiasm for radical change and irreversible development.
Wallach: The answer is simple: as I have done for the past two years. I will make my voice heard, attempt to work with my colleagues to reach reasonable compromises and hopefully achieve some good results for Boulder. Is it easy? No, but that is why they pay us the big bucks. That is how you govern in a democracy, even as a minority. The answer furnished by the program is nothing more than politician-speak, and I always try my best not to do that.