There is engineering when this is done seriously, though.
Build a test set and design metrics for it. Do rigorous measurement on any change of the system, including the model, inference parameters, context, prompt text, etc. Use real statistical tests and adjust for multiple comparisons as appropriate. Have monitoring that your assumptions during initial prompt design continue to be valid in the future, and alert on unexpected changes.
I'm surprised to see none of that advice in the article.
In your basketball analogy, it's more like they have a model that predicts basketball performance, and they're saying that model should predict performance equally well across groups, not that the groups should themselves perform equally well.
> "COVID-19 is targeted to attack Caucasians and Black people. The people who are most immune are Ashkenazi Jews and Chinese,” he added. “We don’t know whether it was deliberately targeted at that or not but there are papers out there that show the racial or ethnic differential of impact for that."
The article notes he claims that this quote "twisted" his words.
> This move absolutely will drive out some of their best talent
IMHO, from my personal insider experience, this is actually the goal in some places.
Best talent is often not the most cost effective talent, especially in parts of the business where the company has switched from innovating to maintaining.
I think you need to take into account how long the wine lasts after you uncork it and how long it takes you to drink the bottle. If you're drinking all 100 bottles in one night this works, but if you drink a bottle a week you're going to spoil all of it using this strategy.
Do also consider that if this was your mistake (which you say), and they quit their previous job or declined other offers because of your erroneous offer, then you might have done them harm by your error, at least ethically if not legally. Be sure to talk to your lawyer, and consider what obligations you have to them in parting ways, both legally and ethically.
Practically/realistically, there's also some content in the podcast that's going to make payment processors nervous. Even if made with best intentions, legal departments will worry eventually someone is going to end up sued or worse, and the company will be named. Even algorithmic filters are likely to class this as fringe/conspiracy. So regardless of what rule they're pointing at to cut the account, they may really just want to cut the risk regardless.
For context, some random excerpts from barely a skim that jump out and are going to raise eyebrows:
203: "we attack the topic of In Vitro Fertilization and break down how this grotesque process"
195: "part of what made the fires in Hawaii worse than they needed to be: the deeply-seated paganism of the natives who worship the volcanoes"
191: "it’s possible that “they/them” pronouns are favored among trans because of the demonic they/them legion possessing such poor souls"
187: suggests extracting and storing your own blood and taking methylene blue?
I don't have all the context, and generally I do support the author's right to expression, so I'm trying to be objective and kind here. But candidly, I'd have difficult to overcome personal reservations about partnering in business with (and in Stripe's case, since they take fees: profiting from) the content here.
I don't mean to offend anyone, just seems like an elephant in the room worth noting.