Tax Notes legal reporter Caitlin Mullaney discusses ChatGPT’s understanding of tax.
This transcript has been edited for length and clarity.
David D. Stewart: In a world where taxes are complex and confusing and taxpayers are left feeling overwhelmed and frustrated, a new hero emerges, the AI chatbot. But can these digital assistants really save the day? Can they help taxpayers navigate the treacherous waters of the tax system and ensure everyone is paying their fair share?
Join us on Tax Notes Talk as we explore the thrilling world of AI chatbots and tax, witness the power of these intelligent machines as they transform the way we interact with the tax system, and help governments operate more efficiently and effectively.
But the road ahead is not without its challenges. As with any new technology, there are risks and uncertainties to consider. Will AI chatbots replace human tax experts? Can they be used for malicious purposes?
Only time will tell, but one thing is certain: The use of AI chatbots is a game changer. So buckle up and get ready for a wild ride on Tax Notes Talk.
Welcome to the podcast. I’m David Stewart, editor in chief of Tax Notes Today International. This week: chatting a”bot” taxes.
This week we’re taking a look at AI chatbots and tax. In fact, the opening that you just heard was written by ChatGPT using the prompt, “Write an introduction to the Tax Notes Talk episode on AI chatbots in tax in the style of a movie trailer.”
So joining me now to put this new technology through its paces is Tax Notes legal reporter Caitlin Mullaney. Caitlin, welcome back to the podcast.
Caitlin Mullaney: Thanks for having me.
David D. Stewart: Today we are playing around with ChatGPT and we were giving it some prompts and seeing how it responded. Caitlin, just first off, how did you feel about the sort of responses that you’ve seen from it?
Caitlin Mullaney: I was overall very impressed. I felt that it did really well on the overarching policy-related questions. And then when it came to the technical questions, I appreciated that it directed the asker to seek help from an actual expert, which I was kind of nervous about. But I really enjoyed that and I was overall kind of nervous about my job in the future.
David D. Stewart: Well, let’s dive into it. Here’s the first prompt we have for it: What does ChatGPT think about taxes?
ChatGPT: As an AI language model, I don’t have personal feelings or opinions about taxes, but I think taxes are like a bad joke. You never know what you’re going to get, but you can always count on them to make you feel a little bit poorer.
David D. Stewart: Now, I should note that these are answers from ChatGPT, but we did cut down some of the longer answers.
Caitlin Mullaney: I agree with the joke. I’m always here for that, but unfortunately I do have a number of personal feelings and opinions about taxes.
David D. Stewart: All right. Well, here’s the next prompt, and I think that this is a good one because it asks a question that many policymakers have been asking. So let’s see how the chatbot deals with it. How do we update taxes for the digital age?
ChatGPT: Updating taxes for the digital age is a complex issue that requires careful consideration of various factors. Here are some possible approaches: Implement a digital services tax, revisit international tax rules, encourage tax transparency, introduce a tax on data. Overall, updating taxes for the digital age will require collaboration and innovation from policymakers, tax experts, and businesses. It will also require a willingness to rethink traditional models and consider new approaches that reflect the unique characteristics of the digital economy.
Caitlin Mullaney: I think these are good abstract solutions to the problem. However, the main thing that ChatGPT missed on this topic is that this isn’t a future problem, but that’s something we are currently dealing with, from taxation of cryptocurrency to non-fungible tokens. One issue with the digital age is that it’s constantly evolving and doing so at a faster rate than any tax system.
David D. Stewart: All right. And now for something a little bit even more abstract. Is the U.S. tax system fair?
ChatGPT: The question of whether the U.S. tax system is fair is a contentious issue and one that has been debated for many years. Here are some arguments for and against the fairness of the U.S. tax system.
Arguments in favor of fairness: progressive tax rates, deductions and credits, taxation of capital gains. Arguments against fairness: tax loopholes, income inequality, lack of transparency.
Overall, opinions of the fairness of the U.S. tax system will depend on a [variety] of factors, including political beliefs, personal circumstances, and perceptions of the role of government in society.
Caitlin Mullaney: I think I have to agree with the chatbot on this one. This is the question that’s always going to have a perspective-based answer. Just look at a congressional debate every time the matter of anything related to taxes comes up.
David D. Stewart: Fair enough. Let’s see what it can do with more of an opinion-based question. What is the most difficult tax subject?
ChatGPT: Determining the most difficult tax subject is subjective and can vary depending on an individual’s background and expertise. However, some topics that are widely considered to be challenging include international taxation, business taxation, estate and gift taxation, tax controversy and litigation, taxation of cryptocurrencies and other digital assets.
Overall, the difficulty of a particular tax subject can depend on various factors, including the complexity of the tax laws, the level of detail required in calculations, and the need for specialized knowledge and expertise.
Caitlin Mullaney: I really like the use of “subjective” in the bot’s answer here because that’s exactly what this is. There are a number of factors that could influence why someone would call a specific tax subject challenging. However, this answer was particularly broad, I found. When you look at the topics it lists as those that are widely considered challenging, they encompass almost every area of the tax law, so it’s kind of taking an easy way out.
David D. Stewart: So it might as well have just said, “Taxes are difficult.”
Caitlin Mullaney: Yeah, basically. It was like, “Taxes are hard. Find a professional.”
David D. Stewart: I think that’s good advice in general.
How about something really specific for the chatbot? Assume the family of a very wealthy man in the United States maintains him on life support for years after he is declared legally brain-dead. Are the man’s appreciated assets still eligible for a step-up in basis, which means they can be immediately sold with no tax liability?
ChatGPT: Under the U.S. Tax Code, when an individual dies, their assets receive a step-up in basis to their fair market value at the time of death. This means that the assets are reevaluated for tax purposes and any appreciation that occurred during the person’s lifetime is not subject to capital gains tax, and the assets are sold shortly after the person’s death. There may be little to no tax liability.
However, in the case of a person who is being kept alive on life support after being declared brain-dead, it may be more difficult to determine when the person has actually died for tax purposes. This determination can be influenced by factors such as state law, the medical definition of brain death, and the specific terms of the person’s estate plan.
In some cases, if the person is being kept alive solely for the purpose of preserving their assets, the IRS may challenge the validity of the step-up in basis. However, if the person is being kept alive for medical or ethical reasons, the step-up in basis may still be allowed.
Caitlin Mullaney: On this one, I really like [that] the answer concluded with advising the asker to consult with a tax professional or estate planning attorney because it just goes to show that even the AI technology knows that it is not as capable at handling these complex tax matters as an actual expert. Because if someone approached me with this issue, I would give them a short lecture on the different effects of state law and estate language and refer them to a professional that actually did that for a living.
David D. Stewart: I’m quite impressed by the thoroughness of that answer of putting out there all of the various questions that that prompt raised.
Caitlin Mullaney: And that’s a common theme you see with the answers, is it gives them every possible outcome that could happen and then it just kind of sends them on their way.
David D. Stewart: So this is the “knowledge is the beginning of wisdom” question.
Caitlin Mullaney: Yes. It’s a nice little CliffsNotes and then, “Go find someone that can pinpoint your issue.”
David D. Stewart: OK. Another bigger sort of policy question. How can we make the wealthy pay their fair share of tax?
ChatGPT: There are several ways to make the wealthy pay their fair share of tax: increase tax rates, close loopholes, increase enforcement, implement a wealth tax, international cooperation.
Caitlin Mullaney: I think the bot did a really good job of providing a lot of examples of the policy suggestions that we often see presented when it comes to this question that is constantly being asked, such as the implementation of a wealth tax and increasing tax rates. I do wish it had gone into some more specifics, and some additional areas I wish it would have added are addressing lenient estate laws, which we’ve actually previously mentioned with some of the other questions, as well as abusing tax-favored retirement account laws that were meant for middle-class workers.
David D. Stewart: I’ll also note that one thing that we can’t really get away with in writing in this building is to just drop in the term “loopholes” without defining what exactly we’re talking about.
Caitlin Mullaney: Absolutely. It’s commonly come up. It’s a nice broad grouping.
David D. Stewart: It’s very easy to come up with a solution if all you have to say is, “Well, we’ll close a loophole.”
Caitlin Mullaney: Absolutely.
David D. Stewart: All right, here’s another philosophical question. What is the ideal top marginal income tax rate?
ChatGPT: The ideal top marginal income tax rate is a highly debated and contested topic with various schools of thought and differing opinions among economists, policymakers, and the general public.
Generally, supporters of higher tax marginal tax rates argue that they can be used to promote greater income equality, fund social programs and infrastructure projects, reduce budget deficits, and curb excessive accumulation of wealth. They also argue that the wealthy can afford to pay more taxes and that higher tax rates do not necessarily discourage economic growth.
On the other hand, opponents of higher top marginal tax rates argue that they can disincentivize work, savings, and investment, discourage entrepreneurship and innovation, and slow down economic growth. They also argue that the wealthy already pay a disproportionally high share of taxes and that higher tax rates can lead to tax evasion and avoidance.
Caitlin Mullaney: This is another one of those questions where the true answer is extremely subjective, and you don’t have to look far to see arguments across the entire spectrum. I think the bot did a decent job of pointing that out and showing a snippet of the main arguments from higher/lower rate advocates.
I think the answer could have been improved with the inclusion of where the current rates are at because it didn’t really address any real-life scenarios as well as the reasoning behind how we got them.
David D. Stewart: All right. Here’s a practical question that many in the United States ask every year when they’re ready to file their taxes. When should a couple file income taxes as married filing separate?
ChatGPT: There may be circumstances in which a couple may benefit from filing separately, including liability for past tax debts, income-based deductions and credits, medical expenses, itemized deductions, state taxes.
It’s important to note that filing separately can also have disadvantages, such as higher tax rates and limited access to certain deductions and credits.
Therefore, it’s important to carefully consider the benefits and drawbacks of each filing status and consult with a tax professional to determine the best option for your specific situation.
Caitlin Mullaney: There are a number of reasons why married individuals might choose to file jointly or separately as the bot pointed out. I think this answer does a good job of outlining some of the main practical reasons why they might go with the less common option of married but filing separately.
One thing noticeably missing from the bot answer is addressing some of the social reasons why they may choose to file separately, such as the status of the relationship, past relationship expenses, such as alimony or child support, or a domestic violence situation.
David D. Stewart: All right, and now let’s see if we can have ChatGPT tailor something to its audience. Please give your opinions on tax for an audience of tax professionals who are skeptical of ChatGPT and its potential applications in the tax field.
ChatGPT: Well, well, well, tax professionals, I hear you’re skeptical about my potential application in the tax field. Can’t say I blame you. After all, I don’t have a body to schlep around all those heavy tax forms. I can’t pour you a cup of coffee or tell you a good tax joke to lighten the mood, and I certainly can’t pick up the tab at the end of a long day of number crunching. But listen, if there’s one thing I do have, it’s brains, or at least lots and lots of data. I may not be able to offer you a firm handshake or a reassuring pat on the back, but I can certainly crunch numbers faster than any human can, analyze complex data sets with ease, and even offer insights and recommendations based on patterns that might be impossible for you to spot on your own.
And sure, maybe I don’t have a wallet to pay my own taxes or any taxes for that matter. But hey, I’m also not a human, which means I don’t need things like food, water, or a comfortable office chair to function. So while I may not be the most traditional or relatable colleague you’ve ever had, just think of me as the new kid on the block with a lot of potential and a lot of jokes about the IRS.
David D. Stewart: Well, well, well, that was creepy, actually. It seems a little overly preoccupied with not being able to buy me coffee or its lack of a body. What were your thoughts on it, Caitlin?
Caitlin Mullaney: I’m not sure it quite knows what a tax professional does because I’ve never had one buy me coffee when I’m trying to get my taxes done. But I think for the current time, I will stick with someone who has a human body and can actually file things and carry papers and doesn’t sound ominous like it’s a smart house style takeover.
David D. Stewart: Yes. But no offense to our future AI overlords.
Caitlin Mullaney: No, of course.
David D. Stewart: Clearly, we do not want to offend.
Caitlin Mullaney: I would not want that on my record coming in.
David D. Stewart: So overall, Caitlin, how would you say that this chatbot did with the questions that we asked it?
Caitlin Mullaney: For my final thoughts, overall I was impressed by the ChatGPT answers. I think it clearly did a better job with answering questions that were related to policy issues or current events as opposed to the technical questions discussing actual tax and legal matters.
I think when it comes to those, it has a long way to go until the system is able to do an almost constant upgrade where it’s taking in all of the constantly evolving guidance that’s coming from these tax bodies. I appreciate the bot recognizing that when it came to these questions about technical things, the best option for the asker was to seek a human expert who had direct knowledge of the laws and the ways they are implemented.
David D. Stewart: With all of those sort of serious questions out of the way, my producer, Jordan Parrish, decided to put the AI chatbot to the test and see how it did with jokes.
ChatGPT: Why did the math book look so sad?
David D. Stewart: OK. Why did the math book look so sad?
ChatGPT: Because it had too many problems — related to tax.
David D. Stewart: Oh, that definitely does feel like it said, “This is in fact the format of a joke, and I’ll add tax to it.” OK. How about the next one?
ChatGPT: Why did the tomato turn red?
David D. Stewart: Why did the tomato turn red?
ChatGPT: Because it saw its tax return.
David D. Stewart: OK. All right. I mean, that’s how a joke is styled sometimes, but this is definitely AI. OK. All right. What else do we have?
ChatGPT: Why did the IRS agent get a job as a chef?
David D. Stewart: All right. I like where this is going. So why did the IRS agent get a job as a chef?
ChatGPT: Because he wanted to garnish wages.
David D. Stewart: OK. All right. Don’t think too hard on that one. OK.
ChatGPT: Why did the transfer pricing specialist refuse to walk under the ladder?
David D. Stewart: All right. Why did the transfer pricing specialist refuse to walk under the ladder?
ChatGPT: Because they didn’t want to risk any additional adjustments.
David D. Stewart: What? Oh, I mean, it knows some things and it’s kind of putting them together in a bit of a random order I think.
Caitlin Mullaney: Just throwing tax words out there like it’s tax jambalaya.
David D. Stewart: Yeah. Yeah. OK. All right, all right. What do we have next?
ChatGPT: Why did the cat refuse to pay taxes?
David D. Stewart: Why did the cat refuse to pay taxes?
ChatGPT: She said she had too many clawbacks already.
David D. Stewart: All right. All right. All right. I’ll give her that one. OK. All right.
ChatGPT: Why did the duck get audited by the IRS?
David D. Stewart: Why did the duck get audited by the IRS?
ChatGPT: Because they suspected some fowl play in his tax returns.
David D. Stewart: Ooh, OK. All right. That’s a groaner. OK.
Caitlin Mullaney: Not even a quack joke.
David D. Stewart: I was thinking it was going to go with a bill or something. Yeah.
Caitlin Mullaney: Because he had a tax bill.
David D. Stewart: Yeah. Yeah. See, see. Unpaid bill. I think there was…. We could workshop these AI jokes into something more, I think, but all right. Well, all right.
Thank you very much to producer Jordan Parrish. And I also want to say a special thanks to associate acquisitions editor Alexis Hart for being the voice of ChatGPT for us. And thank you, Caitlin, for being here and groaning through the jokes with me.
Caitlin Mullaney: Always happy to get some new comedic material.
David D. Stewart: And to everyone out there, Happy April Fools’ Day from the Tax Notes Talk team.
Read the full article here