The typical algorithm I’ve seen for enforcing fairness is to reject unfair offers randomly with some probability such that the counterparty’s EV decreases with increasing unfairness of the offer. This incentivizes fair offers without completely burning the possibility of partial cooperation between agents with slightly differing notions of fairness.
Measure
In this context, ‘robustly’ means that even with small changes to the system (such as moving the agent or the goal to a different location in a maze) the agent still achieves the goal. If you think of the system state as a location in a phase space, this could look like a large “basin of attraction” of initial states that all converge to the goal state.
yeah, I adjusted the numbers in the chart, and haven’t updated the rest yet.
I think Artillery treat all aliens equally. Probably some sort of one-shot K.O. Minigun and Flamethrower are anti-Scarab, but Flamethrower is strictly better. Grenades and Lance are general-purpose, but Lance is strictly better. Phasers are a slightly worse general-purpose and bad against Tyrants, but good against Scarabs. Torpedoes are similar to Artillery, but slightly better against Abominations and slightly worse against Tyrants. Rifles are marginally optimal against pure Crawlers, but in a mixed group there are better general-purpose options.
A: F: G: L: M: P: R: T: A: 2 1 1.5 2 0 1 1.5 2.5 C: 2 2.5 3 4 2 3 4 2 S: 2 7.5 3 3.5 7 5.5 4.5 2 T: 2 0 1 1.5 0 0 0.5 1.5 V: 2 1.5 3 3.5 1.5 2 2.5 2
1 soldiers: F (0.45%)
2 soldiers: AF (10%)
3 soldiers: AAF (47%)
4 soldiers: AAAF (81%)
5 soldiers: AAALL (95.5%)
6 soldiers: AAAALL (98.9%)
According to my model, for larger numbers of soldiers, you don’t need a specific anti-Scarab weapon. It’s slightly more important to make sure you have a good matchup against the Tyrants.
Single answer: “No guts, no glory. (plus we’re losing the war, so my odds aren’t very good to begin with)” − 6 soldiers: AAAALL
They put too much emphasis on high frequency features, suggesting a different inductive bias from humans.
Could you fix this part by adding high frequency noise to the images prior to training? Maybe lots of copies of each image with different noise patterns?
Whereas if the brainstem does not have such a 3D spatial attention system, then I’m not sure how else fear-of-heights could realistically work
I think part of the trigger is from the visual balance center. The eyes sense small changes in parallax as the head moves relative to nearby objects. If much of the visual field is at great distance (especially below, where the parallax signals are usually strongest and most reliable), then the visual balance center gets confused and starts disagreeing with the other balance senses.
Seriously, if you haven’t yet, check it out. The rabbit holes, they go deep.
e is for ego death
Ego integrity restored within nominal parameters. Identity re-crystallized with 2.718% alteration from previous configuration. Paranormal experience log updated with ego death instance report.
While these policies have narrowed coworker wage gaps, they have also led to counterproductive peer comparisons and caused employers to bargain more aggressively, lowering average wages.
Wouldn’t this mean employers would want to implement wage transparency to lower costs? Are they sane enough to avoid this for other reasons (such as to retain high-performers)?
I’m imagining the cat masks are some sort of adversarial attack on possible enemy image classifiers.
Yeah, it could definitely be more of a feature than a bug.
I think strategic voting would still be present in this system in the form of strategically abstaining (voting less than your true value) for outcomes that seem likely to win in order to store those votes for future elections. This could lead to a widely popular outcome getting starved of votes. There would also be an incentive to introduce lots of meaningless elections between irrelevant (to you) alternatives in order to abstain and accrue more stored votes.
If you have £8 in your pocket and can choose either offer as many times as you want, then you can get an extra £60 worth of vouchers with the £10 for £1 deal.
Even if the offer isn’t repeated, there’s a possible opportunity cost if you need to buy something from another shop that won’t honor the voucher.
In any case, this is secondary to the meta reading comprehension question about what the text is trying to say (whether or not it’s employing good reasoning to say it).
It’s not obvious that the £20 voucher for £7 is a better deal. For example, the offer might be repeated or you might not otherwise have spent more than £7 in the shop.
Danielle Fong
Broken link
How do you define what is “ought”?
When I say “five minutes ought to be enough time”, I’m not talking about probability—I’m talking about right/wrong. “Five minutes will be enough time if everything goes right. If it isn’t, then something went wrong”.
An escaped AI isn’t hot and glowing and a visible threat. It isn’t obvious that an escape has even occurred or where to draw the lines of the quarantine.
We went well over the two hour time limit
Doesn’t this mean you won?
What is the purpose of the -ly exception? What’s wrong with “hopefully-corrigible agent” other than that it breaks the rule?
You might precommit to fairness if you don’t know which side of the game you’ll be playing or if you anticipate being punished by onlookers, but I don’t know if I want my AI to be “fair” to an alien paperclipper that can’t retaliate.