Boris Kashirin — LessWrong

LESSWRONG
LW

Replying toThe Weirdness of Dating/Mating: Deep Nonconsent Preference

The Weirdness of Dating/Mating: Deep Nonconsent Preference

Just one data point: to me your post feel shallow in a good way. I can say: "I can't help but note how all this annoying behaviour suddenly become endearing to me when I am in love." but it feels like I am deepening your post, not contradicting it. Am I gesturing at non-terminal property you mentioned?
P.S. just noticed I am using "shallow" when there is literally "deep" in headline. Yet this is my impression.

-2

Boris Kashirin's Shortform

Boris Kashirin

7mo

This is a special post for quick takes (aka "shortform"). Only the owner can create top-level comments.

Boris Kashirin7moQuick Take

I see some discussion about worth of bee vs human, and I feel there is big piece missing there. Why do we think that we are the only consciousness inhabiting our body? If bees are expected to be conscious why not subsystems of human brain? Possible a lot of them?

Replying toTormenting Gemini 2.5 with the [[[]]][][[]] Puzzle

Boris Kashirin11mo

Tormenting Gemini 2.5 with the [[[]]][][[]] Puzzle

When I see the question, I know I am on LW. It allows me to deduce that "arcane runes" part is not important, but LLM don't have this context. Maybe it sounds like crackpot/astrology question to it?

-2

Replying to"It's a 10% chance which I did 10 times, so it should be 100%"

Boris Kashirin1y

"It's a 10% chance which I did 10 times, so it should be 100%"

e^3 is ~20, so for large n you get 95% of success by doing 3n attempts.

Replying toThe Online Sports Gambling Experiment Has Failed

Boris Kashirin1y

The Online Sports Gambling Experiment Has Failed

Thinking about responsible gambling, something like up-front long-term commitment should solve a lot of problems? You have to decide right away and lock up money you going to spend this month and that will separate decision from impulse to spend.

Replying toMomentum of Light in Glass

Boris Kashirin1y

Momentum of Light in Glass

I tried to derive it, turned out to be easy: BC is wheel pair, CD is surface, slow medium above. AC/Vfast=AB/Vslow and for critical angle D touches small circle (inner wheel is on the verge of getting out of medium) so ACD is right triangle, so AC*sin (ACD)= AD (and AD same as AB) so sin(ACD) = AB/AC= Vslow/Vfast. Checking wiki it is the same angle (BC here is wavefront so velocity vector is normal to it). Honestly I am a bit surprised this analogy works so well.

Replying toMomentum of Light in Glass

Boris Kashirin1y

Momentum of Light in Glass

I read about better analogy long time ago: use two wheels on an axle instead of single ball, then refraction come out naturally. Also I think instead of difference in friction it is better to use difference in elevation, so things slow down when they go to an area of higher elevation and speed back up going down.

Replying toThe Sun is big, but superintelligences will not spare Earth a little sunlight

Boris Kashirin1y

The Sun is big, but superintelligences will not spare Earth a little sunlight

It is defecting against cooperate-bot.

Replying toThe Sun is big, but superintelligences will not spare Earth a little sunlight

Boris Kashirin1y

The Sun is big, but superintelligences will not spare Earth a little sunlight

From ASI standpoint humans are type of rocks. Not capable of negotiating.

Replying to"Which Future Mind is Me?" Is a Question of Values

Boris Kashirin2y

"Which Future Mind is Me?" Is a Question of Values

This experience-based primitivity also means inter-temporal self identification only goes one way. Since there is no access to subjective experience from the future, I cannot directly identify which/who would be my future self. I can only say which person is me in the past, as I have the memory of experiencing from its perspective.

While there is large difference in practice between recalling past event and anticipating future event, on conceptual level there is no meaningful difference. You don't have direct access to past events, memory is just an especially simple and reliable case of inference.

How tokenization influences prompting?

Boris Kashirin

I was thinking about how prompt differs from training data in terms of tokenization. If i am to prompt with "solution:" as opposed to "solution: " it seems like it can influence the result, as in training data last token contain some information about next token. If there is token ": T" but my prompt ended in ": " it can be inferred than next token can't be "T[something]".

Is this real effect, or I just misunderstand how tokenization works?

GTP4 capable of limited recursive improving?

Boris Kashirin

Apparently with reflection technique (answer-critique-improve) GTP4 capable of giving much better answers. But that implies it should be capable of doing essentially Alpha Go Zero type of learning! It can't do complete self play from zero as there is no ground truth for it to learn from, but that basically burns all hopes of having further progress bottlenecked by data. Also, while still severely lacking, it constitutes limited self-improvement capability. Not only we have unclear distinction of GTP4 being an AGI or not, but it also have some slight self improvement capabilities! We really are boiling frog, aren't we?