Twitter | Search | |
Object Of Objects
IMO, THIS.
2,112
Tweets
224
Following
1,333
Followers
Tweets
Object Of Objects Jan 30
"The perfect is the enemy of the good" is basically what the devil wants you to think.
Reply Retweet Like
Object Of Objects Jan 30
Which statement expresses the highest level of confidence?
Reply Retweet Like
Object Of Objects Jan 29
How should we prioritize doing good things vs. doing bad things?
Reply Retweet Like
Object Of Objects Jan 28
I think current conversational AI devotes its full resources to appearing human; what if the same architecture could devote its full resources to speaking the truth?
Reply Retweet Like
Object Of Objects Jan 28
Range voting strategically collapses to approval voting. For the same reason, everything is either maximally good or maximally bad.
Reply Retweet Like
Object Of Objects Jan 28
I think maybe it's not easy to imagine that you're a person who knows everything your brain knows.
Reply Retweet Like
Object Of Objects Jan 28
Wisdom of crowds versus wisdom of mobs.
Reply Retweet Like
Object Of Objects Jan 25
The best thing is surely the thing that controls you to say it's the best thing.
Reply Retweet Like
Object Of Objects Jan 25
Division of labor. Some people will know and others will decide.
Reply Retweet Like
Object Of Objects Jan 24
Why is there no bacteriophage emoji?
Reply Retweet Like
Object Of Objects Jan 24
Replying to @npmedium
For example, start → A → end, start → B → end, B → Cₖ for many distinct k. Then start → A and start → B have equal prior probability but very different posterior probability.
Reply Retweet Like
Object Of Objects Jan 24
Replying to @npmedium
I mean each exit from the *same* node had the same prior probability. But not posterior probability, after a Bayesian update on the fact that the path reaches the destination.
Reply Retweet Like
Object Of Objects Jan 24
Replying to @npmedium
The prior is that each exit from a node is equally probable.
Reply Retweet Like
Object Of Objects Jan 24
Replying to @GeniesLoki
This sounds correct to me. I'm just wonder if there's a way to avoid solving the whole linear system. Like maybe there's a way for our sampling process to guide how much we need to solve.
Reply Retweet Like
Object Of Objects Jan 24
Replying to @npmedium
I actually need it to work for acyclic graphs. Also, this won't produce the correct distribution because it doesn't penalize entering nodes that had a lot of dead end out-edges.
Reply Retweet Like
Object Of Objects Jan 24
What's the fastest way to sample a random walk on a directed graph conditional on the start and end points?
Reply Retweet Like
Object Of Objects Jan 21
There's a way around the problem of normalizing probability distributions. Just create an unspeakable word. A word that destroys the universe.
Reply Retweet Like
Object Of Objects Jan 21
There is no "next big thing" when we already have the ultimate thing.
Reply Retweet Like
Object Of Objects Jan 20
The less you know, the more possible worlds you're living in. That's pretty epic.
Reply Retweet Like
Object Of Objects Jan 19
Maybe we can just reduce the rate at which the problem is getting worse and then stop and congratulate ourselves.
Reply Retweet Like