2,082 karma · joined May 14, 2019
Prime example of trade-off between cleanliness of implementation and backwards compatibility. As much as I like a pristine, pure spec, the market winners are clear. It pays to keep product debt around.
* Say yes first. Acknowledge everything and say what you agree with.
* Use the phrase "tell me more." Use this to get context in a nonconfrontational way (as opposed to "why" which can get people in defensive mode).
* Avoid using "but". People don't like to hear it and it makes them forget the past before it. Everytime you want to use but, see if you can use "and" or "at the same time", or nothing, and almost always the result is better.
Yes there is, per the trade regulations you're entitled to a refund
.
Wait
I'm only partly sarcastic. If you think you're entitled to anything provided as a free service then the problem starts there. Do I miss the days when ISPs provided email as part of your subscription? Then made you pay for anything over 100 Mb? Then held your email address hostage if you wanted to switch providers? Hell no. But I also don't think government should force anyone to provide services to me for free.
Edit: downvotes without replies do not help me understand where this feeling of entitlement to a free service comes from; insight would be appreciated.
(But otherwise +1 to schema awareness during authoring.)
Design goal question - is the goal to have a language that blindly compiles to SQL, or will compilation require data schema knowledge?
Suggestion: where possible stick to the well established SQL keywords (prefer "group by" to "aggregate by")
It's the same reason why lottery revenue or (in California) bond approval propositions are typically earmarked for "noble causes" (schools, firefighters, parks...). It's a cheap promise to make, and the share of general operating revenue allocated to these departments can quietly decrease as lottery or bond financing replaces it, completely working around the spirit of the earmarking.
It sounds trivial to say but control of budgeting decisions is obtained through control of budgeting decisions; earmarking constraints are just accounting games to make spending more palatable to taxpayers.
How do truck shippers protect their wares? 1 driver per truck with eyes on the load all the way through. If they have to take a nap, it's in the truck...
How do valuables travel? In armored cars with armed personnel.
The train company isn't entitled to its business model of very long trains with loose security.
Or, partner with Mark Rober to create a giant train-car-sized glitter bomb that will douse thieves in glitter and fart spray when entered.
It is much more important for a company to filter out likely low performers than it is to avoid filtering out likely high performers, because the costs of hiring an incompetent person are enormous. Not only do they not produce as much, they eat up others' bandwidth by asking too many questions, implementing super buggy code, or not communicating effectively. They make other people on the team feel like they're not pulling their weight. They're a pain to fire depending on company culture and laws.
This process is designed to overfilter and maintain an A-player roster as much as possible, vs putting butts in seats. It's hard to tell whether the pool is exhausted here - by some indications, there's a lot of attrition but also a lot of hiring...
A couple questions:
1) I didn't really understand how you went about identifying what vectors of the latent space stand for various things, like pose or color. Did you train one of the AIs to that effect, or did you manually inspect a bunch of vectors, twiddling through them one by one, did to the outcome?
2) If one were to train an AI to the same level using commodity cloud services, what's the order of magnitude cost that you would pay for the training? More like $100, $1,000, $10,000 or $100,000?