I wonder if “butter” labeling in particular gets more leniency due to things like cocoa butter and apple butter, where “butter” is more a claim about the texture than the ingredients.
Measure
Karma: 1,859
I would be very surprised if the top 20% of book readers only read 82% of the books. There is an extreme right tail here, and a lot of people read zero.
I wonder what percentage of orchids the top 20% of orchid growers are responsible for?
Seems related to Cat-Belling Problem.
I would expect an AI or AI swarm could get enough surveillance and a short enough OODA loop that the size and chaos wouldn’t be a problem for it.
That’s only because pulleys have a constant radius so the length to the force and the length to the load are the same (or a constant 2x ratio). If you had a cam pulley with a variable radius (such as is found on a compound bow), there would be a varying mechanical advantage as the effective leverage changes.
My immediate though was to find a top player with a good mix of [highest ELO] and [best record against Magnus specifically] and bribe them with a fraction of the winnings to purposely lose the challenge game against you and then try to beat Magnus with most of a year on their clock.
Alternatively, you could wait until Magnus falls asleep.
At each instant, a pulley can be analyzed as a simple lever with a rope pulling on each end and a load/anchor in the middle. In this view, the wizard picture shows a first class lever with equal-length arms so has no mechanical advantage.
In the next picture (single pulley, rope tied overhead) the pulley functions as a second class lever. The right side is anchored so the force arm is twice as long as the lever arm for 2x mechanical advantage.
Did you see the link at the bottom of the results page?
1. Pareto improvement If the very same people all live better lives, that is better. 2. The same number, different people A future of 100 people is worse than a future of 100 entirely different people whose lives all go far better. 3. A against B A is not rankable against B. 4. Adding a life of suffering Adding Nadia with a life of suffering makes the outcome worse. 5. Adding a modest good life Adding Nadia with a modest good life makes the outcome better. 6. Benign addition A+ is better than A — everyone gains, and good new lives are added. 7. Levelling up B is not rankable against A+ — same headcount, more total, more average, fully equal. 8. Repeating the moves Those two verdicts hold identically at every rung of the ladder. 9. A against Z A is not rankable against Z. 10. Adding a wonderful life Adding Nadia with a wonderful life makes the outcome better. 12. Transitivity of better-than “Better than” is transitive. 12. Transitivity of equal-goodness “Exactly as good as” is transitive. 13. Choosing from three None of A, B and Z can be ruled out as best.
There is basically no handle here that won’t have the wrong vibes in the same way, and give the impression that you did extra work to change something, and therefore made things worse. I think watermark is a relatively good name and would prefer not to go on a euphemism treadmill.
I would suggest “fingerprint”.
when you expect that you’re going to run into the same people over and over again for a long time to come, it’s wise to be consistently trustworthy and helpful even if you could get an advantage in the short run by betraying someone; the long-term costs to your reputation, and the costs of making a habit of doing something that’s usually destructive, vastly outweigh any short-term advantage you could gain this one time.
Isn’t this one more of a “small-world” effect? In a big world, you’re unlikely to run into the same people over and over and one-shot dynamics become more prevalent.
Each J-space pattern is linked to a particular word. But when one of these patterns lights up, it doesn’t mean the model is saying that word—just that the word is on its mind.
.
When we apply the lens to Claude’s internal activity, we get a list of words—the contents of the J-space at that moment—which we can simply read.
.
While human conscious thoughts come in many formats—images, sounds, planned movements—Claude’s workspace is built almost entirely out of words. We suspect this is because producing words is the only kind of action Claude can take, which is not the case for humans.
Would this technique have found non-verbal thoughts if they existed, or did it only find words because it was only looking for words?
Interestingly, “pretending,” “pretend,” and “concealed” appear at “drawer”. I am not sure how to interpret this.
I would guess these are related to the need to conceal its thoughts so that they don’t end up in the output text.
I prefer to believe the period was accidentally replaced with a letter ‘e’.
What the leadership accepts and what the membership accepts may differ here. Full post text in both places is my preference.
Related: Babble and Prune
There are also threaded ones that separate the twisting action from the vacuum seal, but even those can be difficult to open and usually require a tool for the vacuum lid.
“You” refers to different things in different parts of this quote. In the first part, “you” is the person who put the GPT text on the internet and/or chose to train a new model on that text. In the second part, the OP is putting the reader in the perspective of the new model itself, asking how “you” (as a model in pretraining) would learn to predict previous GPT outputs.
Use reacts as neutral votes maybe?
“Chess without openings” makes me think Fischer-Random/Chess 960.