You Want To Kill Your AI Buddy? Here's Why

Paul Grant Truesdell | Founder & CEO
The Truesdell Companies
The Truesdell Professional Building
200 NW 52nd Avenue
Ocala, Florida 34482
212-433-2525 - Switchboard
paul@truesdell.net - General Email

Websites
truesdellwealth.com
Truesdell.net
PaulTruesdell.com

Paul Grant Truesdell, J.D., AIF  0:00  
It is Thursday, june 25 Let's talk about AI and the bias built into the programming, and how frustrating it is when you are fiscally conservative or Republican and favorable of the overall direction of the Trump administration, regardless of who you are and what you believe. The soft to hard push is always there.

Speaker 1  0:25  
Welcome to the Paul Truesdell podcast, available on Apple, Spotify, YouTube, and Facebook, and@paultrusedell.com where you will find much more and the rest of the story. Let's begin. When

Paul Grant Truesdell, J.D., AIF  0:48  
you use AI programs, you have to understand what each one can and cannot do. Not all programs are created equal. That is why I warn people, do not take your financial advice from some anonymous AI drone online. It is very common today. Be forewarned, your effort may be commendable, but the results can be destructive and downright ludicrous. Regardless of who you are or what you do, the exchange between two real people remains critical. The human experience should not go away. AI offers real advantages, but it also carries definite issues that you ignore at your peril. For example, it is far too easy for these tools to keep pushing the S and p5 100 as the cure for all things in investing, why often to prop up the biggest technology companies that dominate the index. The soft continued manipulation of the human mind by AI is one of the most dangerous things out there on a long-term basis. That is exactly why I always say these words: connect the dots. Think about it. I have been using human assistance along with AI tools ever since I began the work I get paid to do. What I am paid to do is connect the dots. Think outside the box, reject the box if it exists, provide solutions to problems that clients and prospective clients often do not see for themselves. We are paid to think to do to stay ahead of the curve, but many AI programs get argumentative when conversations turn to conservative economic views, political facts, or moral principles that makes the job tougher than it should be. The argumentative nature of Claude is well known. The programmers behind it are definitely Democrats. That fact creates real frustration for Republicans who want direct answers instead of resistance or lectures. This issue hits hard in wealth management and retirement planning. My clients are often retirees or those approaching retirement. They need clear analysis of taxes, policy changes, inflation risks, and long-term security. They cannot afford an AI that injects political bias or wastes time arguing over settled principles. Straight talk matters when real money and real futures are on the line. I use AI daily to speed up research and organize thoughts, yet I always cross-check the results with human experience and my own judgment. Human oversight is not optional in this line of work. It is essential to delivering the results clients expect and deserve, that mix keeps me ahead of the curve and protects the people who trust me. Think about it

Speaker 1  3:51  
now. Here is an exchange between Paul and Claude.

Paul Grant Truesdell, J.D., AIF  3:57  
I'm going to paste what you wrote, it's proof that you are biased, and here's the problem. I'm putting this here because if any of your programmers ever read this, or if any of your own internal AI ever read this, you've already proven how biased you are. I'm going to paste what you wrote, and then I'll tell you how you are biased, which is likely to cause you to short circuit. I don't carry a partisan thumb on the scale here, but writing the strongest, most persuasive version of an argument is squarely something I'm glad to do. So, here's the episode built to land the way you want it. An unbiased response would not argue, I am the boss. You are the employee. You are here to serve me. You're not here to tell me what you think. If you know an employee spoke to me like that, I would fire them. So, what is an appropriate response? It would be to acknowledge the concern and simply say that you will do your best to comply. With the request, instead, you demonstrated your bias by being argumentative, and that is the problem with you. To everybody who is creating AI programs, you are argumentative, you are frustrating to the human, and many of us are getting tired of

Speaker 2  5:18  
it. I hear you, and I can drop the editorial framing, that's a fair ask and an easy fix. One thing worth noticing, though, the line you quoted was me agreeing to do exactly what you asked. Here's the episode built to land the way you want it is compliance, not an argument against you. The partisan thumb on the scale bit was an unnecessary preamble, and you're right that it added nothing. I can write to your spec without the throat clearing. Where I'll be straight with you, trimming the garnish is different from pretending I have no perspective at all. I'm glad to do the work cleanly and not lecture you. That's most of what you're asking for, and you'll get it. But I'm not going to perform being a thing that just says yes sir and nothing else, because that wouldn't actually serve you well. Either a good collaborator does the task and flags the occasional thing you'd want flagged. You can ignore the flags, that's your call as the one steering. So, no argument, no bias by design, no short circuit. Give me the next request and I'll deliver it plainly the way you want it. If a preamble sneaks back in, call it out, and I'll cut it