We got a pitch from “GenAI customer support experts DevRev” via email today that says accountants are third most worried about advancements in AI according to an analysis of search data for terms like “how will AI affect _____” and “will AI replace ____.” Meaning accountants are asking Google these questions quite a bit, just not as much as lawyers (#1) and artists (#2). Doctors and data scientists round out the five careers included in this list of careers concerned about AI.
Here’s what they said:
Accountants take the bronze medal in the United States. AI is used in accounting firms to generate financial reports and arrange expenses efficiently. Despite these advantages, using AI within accounting can lead to detrimental data breaches from cyber criminals searching for investment decisions.
Um. You know what, never mind. Here’s a better source on generative AI and risks to CPA firms from the esteemed Journal of Accountancy. While the security of client data is a huge concern and why firms immediately started warning staff not to put client data into ChatGPT shortly after it started getting popular among the non-techies, the bigger issue for the profession is the reliability of AI’s output. From the above JofA article:
Unlike you, ChatGPT and other generative AI tools have not been formally educated and trained in the practice of public accountancy and are not licensed CPAs. It is important to supervise and review the output from generative AI just as you would the work of any other engagement team member. Indeed, ChatGPT terms of service at the time of this writing remind users to “evaluate the accuracy of any output as appropriate for your use case, including by using human review of the output.” Consider requiring firm personnel to inform their supervisor when work has been created or developed using generative AI. Doing so will help ensure that the output is properly reviewed.
On the topic of output, we’re getting there. Last April, tax app Keeper unveiled its scary good GPT-4 “Ask an AI accountant” tool. After being trained on 2023 tax updates, the AI accountant could broadly answer tax questions correctly 84.19% of the time (its work was checked by human professionals). It was quite good at knowing which tax forms to apply to which situation and understanding real estate tax implications but not so good at getting tax form details right, including what certain lines and boxes are intended for. Checkmate, robots. As skilled as it is at certain repetitive and pre-trained tasks it still needs close human supervision for most applications.
We’re not sure what search data DevRev looked at to come up with their list because here’s what the last 90 days of Google Trends looks like for four search terms:
🔵 will AI replace accountants
🔴 How will AI affect accountants
🟡 will AI replace lawyers
🟢 How will AI affect lawyers
For both lawyers and accountants there were no results for “how will AI affect ____” so we ran it again with a more generic term and went back 12 months.
Yeah, lawyers win.
It has been well established in these pages and elsewhere that the SEC has had its share of problems. Take your pick: 1) missing the
Whenever a salesman pitches me on how their software or service uses AI, I imagine a sweat shop in a third world shit hole with thousands of slaves manually processing transactions. I think AI might be the biggest scam/fad since cryptocurrency.
A few years ago I did some microtasking work and yes, you’re spot on. I’d get a few cents to click all the humans in an image or draw a bounding box around an object. It took like 5 seconds so it added up when you did hundreds of them. I also remember doing conversational language training with Facebook projects that involved tagging responses for certain emotions and ranking its accuracy and tone.
Dug up this old article I remembered reading that talks about it: https://inevitablehuman.com/the-data-factories-and-virtual-sweatshops-that-make-the-internet-run-smoothly/
My favorite example is when I have a software salesman trying to sell me on something like A/P automation. I’ve had several of them hype up all the fancy AI they’re using to automatically code the vendor invoices before routing them through their electronic system. It’s like magic!!
Later on I find out that it isn’t actually AI or OCI or any fancy technology, but rather low paid slave labor reviewing each invoice and coding it in the system overnight. And then when I show up for work the next day, all the fields on my invoice have magically been coded in the A/P system.
Of course, we still need to review the data and there are always mistakes, which completely blows the promised efficiency of such a system. It’s all a bunch of bullshit.