Community Archive

🧵 View Thread

🧵 Thread (11 tweets)

Placeholder
UltimApe@ultimape• about 4 years ago

Makes me wonder if anyone has used emulators + phase space simulation to generate TASs of physics games like Kirby's Dream Course.https://t.co/bPkJF1FEPS https://t.co/c8pWn7YGaC

0 0
7/10/2021
Placeholder
UltimApe@ultimape• about 4 years ago
Replying to @ultimape

I once got holes in ones in all the levels.Theory crafting about probabilities and costs across different domains like that is why I think so much about Bayesian optimization problems. https://t.co/yahy9UiBrr

Placeholder
UltimApe@ultimape• about 8 years ago

Also a good description of bayesian optimization method: https://t.co/T1C5lgo9mQ

2 1
0 1
7/10/2021
Placeholder
UltimApe@ultimape• about 4 years ago
Replying to @ultimape

I go really meta though. I think about the phase space of people exploring phases spaces under constraints of speed/accuracy and imperfect knowledge games.https://t.co/0UTlR6lPcU

Placeholder
UltimApe@ultimape• over 8 years ago

Bayesian collectives"studies of spatial behavior have usually been carried out on individuals…"https://t.co/Ai5XiCobJI ht @GoatsThatStare

4 1
0 0
7/10/2021
Placeholder
UltimApe@ultimape• about 4 years ago
Replying to @ultimape

Effective exploring of a phase space is basically what slimemold do.https://t.co/Gg9fQjdMv4

Placeholder
UltimApe@ultimape• over 6 years ago

"Decision-making without a brain: how an amoeboid organism solves the two-armed bandit"https://t.co/GV2EXbNwEa"the learner's aim is to collect enough information so that it can predict the next best arm to play"https://t.co/mptdAAbhkihttps://t.co/UwaqIqOynr

1 0
0 0
7/10/2021
Placeholder
UltimApe@ultimape• about 4 years ago
Replying to @ultimape

Using phase space to model phase spaces is really fun.Once you start thing about optimizing optimization problems, you get into some really fascinating territory.https://t.co/wByer5hVDZ

Placeholder
UltimApe@ultimape• almost 7 years ago

I don't have pedigree or language to pass any qualifications to work for Facebook's AI dept even though all I think about some days is how to solve Contextual Bandit problem by localizing inference to speed up quorum based systems.https://t.co/SU2GM9rioJhttps://t.co/gTYcQKl006

2 0
1 0
7/10/2021
Placeholder
UltimApe@ultimape• about 4 years ago
Replying to @ultimape

AI go FOOOMhttps://t.co/7ptH91LrHc

Placeholder
UltimApe@ultimape• over 4 years ago

"relatively well known in academic circles but only just starting to be exploited by service providers, is that identifying the most appropriate machine learning techniques to model any given data problem is itself a machine learning problem." https://t.co/nqZg3irSwC

2 0
0 0
7/10/2021
Placeholder
UltimApe@ultimape• about 4 years ago
Replying to @ultimape

Robots are learning to become better at tying shoes than I am.https://t.co/8aQqmxb15OMakes me wonder if all I need to do is increase the sensitivity of my c-tactile fibers in my fingers and use that to avoid clumsiness.

0 0
7/10/2021
Placeholder
UltimApe@ultimape• about 4 years ago
Replying to @ultimape

Meta-Knot theory.https://t.co/pTZ5fPHV2l https://t.co/3ADyjvp1ln

Tweet image 1
Placeholder
UltimApe@ultimape• about 5 years ago

Sure I can solve knot theory problems, but I can't fucking tie my shoes half the time. https://t.co/mnnY4TPChv

3 0
0 0
7/10/2021
Placeholder
UltimApe@ultimape• about 4 years ago
Replying to @ultimape

Oooh, thats a neat thing for my brain to pull up.Are we starting to teach robots the basics of cause-and-effect thru embodied cognition? https://t.co/CtWGmnV17H

Placeholder
UltimApe@ultimape• over 8 years ago

Can MIT's cog figure out that it can pull a rope'd toy, but not push it, yet?

0 0
0 0
7/10/2021
Placeholder
UltimApe@ultimape• about 4 years ago
Replying to @ultimape

"Our network architecture produces higher quality results than time-series autoregressive models such as LSTMs as it deals explicitly with the latent variable of motion relating to the phase." https://t.co/FeEIWMy8el

0 0
7/10/2021
Placeholder
UltimApe@ultimape• over 3 years ago
Replying to @ultimape

Centaurhttps://t.co/9duZn7TY1E

Placeholder
eigenrobot@eigenrobot• over 3 years ago

@telmudic @tszzl fuck this actually checks out too metal: huge training sets needs uninterrupted power supply locked in a box literally no genitals moist: just fucking works somehow figures things out from theory and small samples often wrong, never penalized eats anything, shits constantly

26 0
0 0
4/20/2022