HNHacker News
TopNewBestAskShowJobs

poppingtonic

1,279 karma · joined August 31, 2013

Artist, author & programmer. Incremental learning buff. Among many other things, obviously.

[ my public key: https://keybase.io/poppingtonic; my proof: https://keybase.io/poppingtonic/sigs/bkWTUUgvF-hPLoKKg5asnSdZul2o9KdCmvSaQNLmm-s ]

submissionscomments
poppingtonic··on ANTLR Mega Tutorial
I used ANTLR to write a Python parser for the SNOMED expression language last year, and testing it was one of the weirder parts of the experience. I was up and running in a few days, which was largely thanks to the ANTLR book. I love this project. It made doing what I did a lot more fun than I thought it would be. Hand-rolling an ABNF parser from scratch would be a nice hobby project, but not when one has a deadline.
poppingtonic··on Prophet: forecasting at scale
This is very interesting. Forecasters who participate in the Good Judgment Project, such as myself, will find this useful.
poppingtonic··on Switch OS Walkthrough [video]
It makes me think "It's happening!". It's designed to make the player more excited to finish the next step of setup. This is a very interesting reward mechanism.
poppingtonic··on An Introduction to Counterfactual Regret Minimization
Yes, I looked it up and found this source quite useful for getting a high-level idea of what the CMU team accomplished.
poppingtonic··on That Old-Time Hucksterism: The Oddest Doohickeys of Industrial-Age Entrepreneurs
The funniest thing on that page is the "Frequently bought together" section. A moustache-twirling coffee mug?
poppingtonic··on #define CTO OpenAI
Has anyone at OpenAI tried implementing Quantilizers [1]?

[1] https://intelligence.org/files/QuantilizersSaferAlternative....

poppingtonic··on OpenAI Universe
This is true, but some are more complex, which is also due to the choice of tool. For example, [1], made by Chris King, using Adobe Premiere. This is a time-lapse of the process that he used to make parts of [2]. Notice the pattern that emerges when sequential images start lining up.

[1] https://youtu.be/M7Hr83OI-rs

[2] https://youtu.be/1-rFV_d6RH8

poppingtonic··on OpenAI Universe
This is amazing! I was thinking of this problem when I saw a friend making a stop-motion video. The steps are super repetitive and I asked him, "maybe an DeepMind Atari-style RL agent can learn how to do this?" But I didn't want to do what DeepMind did to emulate Atari games with an Adobe editing tool. This is an experiment that I can now run.
poppingtonic··on Tutorial: Metacompilers Part 1
Yes, I googled it during the talk.
poppingtonic··on Hard Tech Startups
I watched Chad Rigetti's talk (https://www.youtube.com/watch?v=GzMvG8UO6Eg) this morning, which was really exciting.
poppingtonic··on Ask HN: How long is your daily commute?
By public transportation: 45 minutes each way ± 10 for vehicle delays, since the system I use is inefficient. Good time to read and listen to podcasts, though.
poppingtonic··on Ask HN: What are the cool life hacks for Developer?
Hi, dreeves! The Complice integration is the best thing yet! I'm wondering how many other kinds of such services can be added like this.
poppingtonic··on Ask HN: What are the cool life hacks for Developer?
Push the responsibility one step away from yourself. Since the goal needs data from you, then you can connect the goal to something like github, RescueTime, TagTime, Complice, etc. Or design your own service for your own personal needs [please do this if nothing exists]. Example with GitHub: if, to lie to Beeminder through GitHub, you have to make a fake commit, push and pull request, then you're already part of the way to doing what you want! Just s/"fake commit"/"what you actually want to accomplish", and as they say, Bob's your uncle.
poppingtonic··on Ask HN: What are the cool life hacks for Developer?
Use this perl project called TagTime <https://github.com/dreeves/TagTime> to track your own time. Description from GitHub: To determine how you spend your time, TagTime literally randomly samples you. At random times it pops up and asks what you're doing right at that moment. You answer with tags.

Use Beeminder <beeminder.com>, a commitment device with a sting. Make a commitment, and give them a price (in USD) for failing to meet that commitment. They take your money if you don't do what you said you'd do. They have APIs for many services such as RescueTime and TagTime, and an Android app.

Note: If you're the kind of smart person who always finds a way out of doing the things you claim to care about, then if these two tools don't get you to do those things, you'll at least think deeply about, and know if you really cared about them.

You either get value (success in the things you care about), or information (learn what you don't care about, that you thought you cared about). Then you can use the value of that information to reorder your preferences in an optimal way.

poppingtonic··on Alan Kay has agreed to do an AMA today
There's a recursive sense in which "training" "System 1" involves assimilating more abstractions, through practice and spaced repetition, such as deferring to the equations of motion when thinking about what happens when one throws a ball in the air. Going as far as providing useful interfaces to otherwise difficult cognitive terrain (a la Mathematica) is still part of this subproject. The process of assimilating new abstractions well enough that they become part of one's intuition (even noisily) is a function of time and intense focus. What do you see as a way to aggregate the knowledge complex and teach further generations of humans what the EEA couldn't, fast enough that they can solve the environmental challenges ahead? What's HARC's goal for going about this?
poppingtonic··on Alan Kay has agreed to do an AMA today
Alan, Thank you for doing this AMA!

I would love to hear your thoughts on how to "train" "System 1", in order to make "System 2" more powerful. Not necessarily here, due to the time factor, but if you find some time to think more deeply on this, please let me know and we can think through this together.

poppingtonic··on Alan Kay has agreed to do an AMA today
1. Spaced repetition can make the recalling - and thus the thinking and pondering - easier. It can certainly make one more consilient, given the right choice of "other things to study" e.g. biology or social psychology, as you've mentioned in an earlier comment. 2. It takes quite a bit of training for a reader to detect bias in their own cognition, particularly the "cognition" that happens when they're reading someone else's thoughts.

What to do about System 1, though? Truly interactive research/communication documents, as described by Bret Victor, should be a great help, to my mind, but what do you think could be beyond that?

poppingtonic··on Two-hundred-terabyte math proof is largest ever
Hooray for proof search techniques. How much data would the Godel-numbering process generate, though? I'm assuming such a search would use this, unless I'm mistaken about my mathematical logic.
poppingtonic··on The Busy Beaver Game
I'm surprised it's that simple. Then again, simplicity is usually surprising.
poppingtonic··on The Busy Beaver Game
That's what he means. We need a formal system more powerful than ZFC, i.e. new mathematics. If a formal system can compute BB(7910), then it is more powerful than ZFC. I'm not sure anyone knows how to build that yet.
poppingtonic··on CakeML: A verified implementation of ML
One of the research organizations using CakeML is MIRI (Machine Intelligence Research Institute), which is doing research in machine self-reference. They recently used HOL, an automated theorem prover, to demonstrate a formal correspondence between the model of HOL within HOL (“inner HOL”) and HOL itself (“outer HOL”). Informally speaking, Fallenstein and Kumar show that one can always build an interpretation of terms in inner HOL such that they have the same meaning as terms in outer HOL. The authors then show that if statements of a certain kind are provable in HOL’s model of itself, they are true in (outer) HOL. See more here: https://intelligence.org/2015/12/04/new-paper-proof-producin...
poppingtonic··on Please scan my towel
Oh, what a hoopy frood.
poppingtonic··on AlphaGo beats Lee Sedol again in match 2 of 5
There's an interesting angle to this phrase "intelligent species opening up a new way of looking at the world", which is that we (humans) designed go as a game - a subset of the real world we interact with. Go is "reality" to alphago. The superset of all possible sense data it could have, in principle. Whatever "chunks" AlphaGo uses, if it does use them, all of its policies are built only from subsets of the sense data that is the interactions (self-plays) and inferences from past games. There's nothing outside the game to bring into its decision process. With humans, however, our policies are noisy and are rife with what, for lack of a better term, I would call leaky abstractions.
poppingtonic··on Should AI Be Open?
'If you were to come up with a sort of objective zoological IQ based on amount of evolutionary work required to reach a certain level, complexity of brain structures, etc, you might put nematodes at 1, cows at 90, chimps at 99, homo erectus at 99.9, and modern humans at 100. The difference between 99.9 and 100 is the difference between “frequently eaten by lions” and “has to pass anti-poaching laws to prevent all lions from being wiped out”.'

[EDITED: the intended quote is below. the quote above is the next paragraph of OP, which is only slightly less relevant than the intended one]

'Why should we expect this to happen? Multiple reasons. The first is that it happened before. It took evolution twenty million years to go from cows with sharp horns to hominids with sharp spears; it took only a few tens of thousands of years to go from hominids with sharp spears to moderns with nuclear weapons. Almost all of the practically interesting differences in intelligence occur within a tiny window that you could blink and miss.'

Yudkowsky's position paper on this idea explains this in more detail: http://intelligence.org/files/IEM.pdf

poppingtonic··on Human-level concept learning through probabilistic program induction
http://github.com/brendenlake/visual-turing-tests, http://github.com/brendenlake/omniglot and http://github.com/brendenlake/BPL
poppingtonic··on Human-level concept learning through probabilistic program induction
Bayesian Program Learning is probably the most powerful of the technologies that come with the use of fully probabilistic programming languages. This is a subset of a technique known as "probabilistic generative modelling", in which one designs a model based on inference algorithms like MCMC or Particle Cascade, that generates samples based on some simple rule. You can learn more from a pretty good online book called Probabilistic Models of Cognition: https://probmods.org/, also here: http://www.robots.ox.ac.uk/~fwood/anglican/ and here: http://probabilistic-programming.org/wiki/Home.
poppingtonic··on Modafinil for cognitive neuroenhancement: a systematic review
[Gwern](https://news.ycombinator.com/user?id=gwern) has done some interesting self-medication experiments [here](http://www.gwern.net/Modafinil), for Modafinil and [here](http://www.gwern.net/Modafinil), for nicotine. I haven't seen similar reproductions or analyses by other scientists, so peer review (or functional analog) is definitely needed here.
poppingtonic··on William Gibson: “What would my superpower be? Redistribution of wealth”
This makes such an interesting case. There a number of concepts so original and unique that they haven't been included "as a single word" in the conversations that are likely to determine what is considered relevant in world culture. There are cultural patterns that we simply don't have words for. Unless we create a truly universally granular language, that is: `Q: What is your favourite word? A: “Thing”, apparently. The harder I’m writing, the more imprecise my spoken language becomes.`
poppingtonic··on Scheme Bricks
Livecoding is awesome, and one of the coolest resources when I started getting interested in it was pawfal.org and Fluxus. What an interesting way to represent s-expressions.
poppingtonic··on Ask HN: What are some startups working on big problems in the developing world?
Bitpesa: http://bitpesa.co

Based in London and Nairobi, they've built a Bitcoin-based service that enables migrants & expats to send money quickly and safely back to their relatives (called remittances in FinTech parlance). I'm not directly associated with them, but we've met many times at Nairobi's monthly Bitcoin meetup.

← PreviousPage 2 of 3Next →