attempting to not get into the habit of sharing bad tweets on here, but this one is emblematic of a common, bizarre, and dangerous misapprehension of what "artificial intelligence" is and what it's good for, and I want to write a bit about it:
Though I can't speak much to this post technologically, as I am not a person anywhere near involved with developing machine learning or AI, I do want to offer my perspective as a used car salesperson who uses this technology every day.
Though I sell cars, my main job is to be an underwriter. I am the last bastion between the unwitting public slavering to get into a Nissan Altima and the bank I work for betting on whether the loan they serve will function or flop. My job is to sit at a computer, try to get customers into the store, and collect their information for underwriting. Drivers licenses, paystubs, utility bills, SSA award letters, bank statements, insurance cards, social security cards...I have to collect it all, scan it, and funnel it into the bank's system.
As of a few years ago, well before I started with the company, they developed a system that uses OCR, algorithmic calculations, and machine learning to interpret the thousands of documents a day they receive from us lowly underwriters...and in an industry where speed is key, gave us another option aside from emailing it to the bank and waiting for a call back. My goal is to get a customer their terms, sign a contract, and get them into a car. Every minute counts, and having an "AI" that can look at a customer's paystub and calculate their verifiable income in less than a minute does, indeed, make my job easier.
Every minute of collecting paperwork, contacting the bank, and waiting for the financial processes behind the scenes to crank it all through the gears to give me a number I can use is a minute that I could better be using to engage my customer, showcase cars, and even get them on a test drive. Scanning a paystub and having it verified within minutes is a godsend, taking what would have taken me a good hour of emailing, waiting, underwriting, and waiting some more into productive time.
As to the original tweet: yes, automating things such as data collection and mathematical verification saves me time that I could use to productively engage a customer and roll a car. When that system fails, we have a backup: email the funding desk, send along the scanned information, open a ticket, make your case, and hope for the best. But for the standard cases, every minute counts, and getting lightning-quick results saves me time and helps me sell more cars.