The folks who think that artificial intelligence, which is nothing more than complex software, can replace tax return preparers face a stark reality. For some taxes, surely artificial intelligence has advantages. But for any tax preparation that requires judgment, wisdom, experience, and intuition, artificial intelligence fails. Perhaps decades from now, when neuroscientists have figured out how judgment, wisdom, experience, and intuition are reflected in the biochemical and electromagnetic functions of the human brain work, and software engineers have figured out how to translate those functions into computer code, the idea of robots doing federal income tax returns might come to a worthwhile fruition. Until then, the likelihood of crashes that weren’t supposed to happen and time wasted at badly programmed traffic signals will make the robot tax return preparer a fine wine that no one should drink before its time.Nothing during the past 15 months has changed my mind.
Recently, I heard a story about artificial intelligence being used to do tax planning. At a meeting of tax professionals, the presenter shared the advice generated by an artificial intelligence system designed to do tax return preparation and tax planning. The system generated some very bizarre ideas. For example, it suggested that an elderly couple whose only income was social security benefits should start their retirement planning by setting up an IRA and contributing to it. It also suggested that a middle-aged couple who had paid off the mortgage on their residence should borrow money to purchase another home in order to create mortgage interest deductions and additional real estate tax deductions. Those sorts of responses on a tax exam would earn an inescapable F grade.
As I suggested in Robots Doing Tax Returns?, “technology needs to generate results that have at least the quality they would have if an expert did the work.” I understand technology. I understand how it can fail. I understand that failure can be at least as bad, if not worse, than the outcome when an expert fails. For that reason, I cautioned, “For me, until a technological ‘advance’ is ready for prime time, it needs to remain in the world of testing and experimentation. It ought not become mandatory or widespread until it proves its superiority.”
The rush to let artificial intelligence and robots “do the work,” a temptation perhaps fueled by expectations of leisurely lives, poses grave risks. Quoting again from Robots Doing Tax Returns?, “Technology is no better than the programmers who design the hardware and software. Sometimes I wonder if the advantages of multiple sets of eyes and brains reviewing the product are being lost on account of cost-cutting goals that misperceive the difference between long-term and short-term success.”
At the moment, very few people are likely to communicate directly with an artificial intelligence system to get advice. But most people are communicating with experts, advisors, and guides to get help. It makes sense to ask questions to get assurances that the assistance isn’t simply being cranked out of an artificial intelligence system, bereft of judgment, wisdom, experience, and intuition.