In yesterday’s publish, I calculated the profitability of public software program corporations. To calculate these figures, I constructed a bit of Rube Goldberg machine.
I didn’t obtain the information into Excel. As an alternative, I complexified issues by sending the evaluation to 4 AIs to see if they might agree.
The inspiration : many corporations have used Amazon’s Mechanical Turk to crowdsource duties, & choose a consensus reply throughout three employees to enhance accuracy.
Why not do this throughout 4 AI employees as an alternative?
Immediate : “calculate the common internet revenue margin and money circulation from ops margin from this information set” plus the information set. Notice that CFOM isn’t a easy common however requires dividing money circulation from ops by income beforehand.
| Mannequin | NIM, % | CFOM, % |
|---|---|---|
| Claude | 4.99 | 27.31 |
| Gemini | -9.29 | 16.2 |
| Perplexity | -8.67 | 14.4 |
| ChatGPT | – 9.29 | 1,433.01. / 14.9% |
| My Evaluation | -9.29 | 16.2 |
Gemini scored prime marks for tabulating accurately on each columns. ChatGPT did properly with NIM however “forgot” to finish the extra division step, which I corrected with a observe up, however nonetheless not the suitable determine. The opposite programs missed the mark altogether.
It could be a mistake to attract any broad conclusions from my little experiment.
However on this case, consensus doesn’t but work as a method which implies I nonetheless have to double verify calculations myself.
Sooner or later, AI will mechanize the illusory Mechanical Turk & I’ll restart my Rube Goldberg math machine with confidence.
