Hacker News new | ask | show | jobs
by arduinomancer 1711 days ago
“AI” startups that are actually just linear regression/random forest as a service
1 comments

"AI" also gets used for cosmetic reasons. I knew a startup that got acquired. They had a .ai domain with no AI product whatsoever.
Sold a small SaaS product 2 years ago, which was marketed as an "AI" product.

Extract from the "model":

                $yearsBetween = $comparisonItems->created_at->diffInYears(Carbon::now());
                if ($yearsBetween <=1) {
                    $multiplier = 1.08;
                } else if ($yearsBetween <= 2) {
                    $multiplier = 1.12;
                } else if ($yearsBetween <= 3) {
                    $multiplier = 1.15;
                } else if ($yearsBetween <= 4) {
                    $multiplier = 1.22;
                } else if ($yearsBetween <= 5) {
                    $multiplier = 1.37;
                } else {
                    $multiplier = 1.41;
                }

Those multipliers were just from a bit of human-playing-around-with-the-data.

The thing is, customers were happy with the result and, as far as I know, still are. Internally it was always called the random number generator.

How was morale internally? Did the whole thing feel disingenuous? Were people cynical about it?
Only a small team of three but I can’t say it ever felt like a bad thing. If anything, proud that we didn’t need to train a model to get a decent result and could concentrate on the rest of the product experience instead.

If customers weren’t happy or the results they were getting were wildly out (they weren’t) then might’ve been different.

In our minds the customers are buying a product to produce a result. If that’s from an ai model or a bunch of ifs it shouldn’t matter.