1,439 karma · joined February 19, 2011
Previously: founder @ LeadGenius (YCS11), PhD @ Berkeley.
Computer scientist. AAAI HCOMP is my home conference.
http://www.crowdbotics.com
talk to me: anand [at] crowdbotics.com
We've ruled out a number of techniques and also have shown that any solution must fall into a category avoiding a number of conditions (natural proof, algebrization barrier). Note also Ryan Williams' result separating ACC0 and NEXP. All of that is progress as I see it.
It's simply that construction of practical approximation algorithms or heuristics that work on large instance is (in general) much easier than separating proofs for complexity classes.
As someone who once struggled extensively to use AMT, I think the need here is genuine and I'm glad to see someone doing it right.
LeadGenius (YC S11) is the operating system for outbound sales, merging data curation and machine learning with crowd computing at massive scale. Our customers are some of the biggest sales & marketing teams in the country. Our not-so-secret social mission is to attack poverty by creating lots of fairly-paid jobs in data curation. You'll be working with a team of smart, good-hearted people, with awesome benefits, and we'll relocate you to the Bay Area if you're anywhere else in the country.
Interview process is a phone screens followed by a half-day onsite.
It's quite possible to teach sophisticated computer science without computers at all, so the first question is what the most useful skillset will be. Basic frontend engineering is probably the most in-demand skill today and allows individuals to see their product on the web right away. Young inmates may also be interested in coding as a path towards entrepreneurship and this can be bundled into a curriculum.
You can code with just a text editor and view results in a web browser without internet access.
Here are two apps that may prove useful as text editors:
1) https://itunes.apple.com/us/app/diet-coda/id500906297?mt=8
2) https://itunes.apple.com/us/app/codetogo/id382677229?mt=8
It may also be possible to take the iPads out and upload the content to the web yourself, provided staff review & clear the content. This can lead to effective outcomes.
-Anand
Finding a purely polynomial algorithm for linear programming or establishing that none exists is a major open question for linear programming, and indeed, for all of mathematics. It's not a Millenium problem, but it's the next best thing [2].
[1] This paper has an even better result, finding a polylogarithmic simplex variant. http://www-personal.umich.edu/~romanv/papers/simplex.pdf
[2] It's Smale problem number 9. http://en.wikipedia.org/wiki/Smale's_problems
You're right that challenge work does a reasonable job of screening tasks from search engines, though!
The reason was that I've seen this kind of thing go awry when the company interferes to try and make itself look good (ie: http://www.fastcompany.com/1773610/column-was-crowdsourced-s...). Our writing pool and management system is solid, so there was little to worry about.
The other reason that this worked well is that Dylan wrote from a set of notes that outlined the story he wanted to tell, rather than letting them write something totally from scratch.
The most effective way to learn is to go out and find the people who should be using the product, and talk to them.
We might benefit from introducing some specialized vertical pages, and we'll provide some more color on who's doing the research work.
The part of the process that's masked is research and strategizing around likely customer sources, which varies significantly per business category. Users might definitely prefer to pay per lead, but that's much harder for us to model from a cost perspective – some users might be easy to target, others might be harder, and some might be impossible to uncover at all.
Here's a more sophisticated use case: suppose a crowdfunding service is trying to construct a list of accredited investors in Los Angeles eligible for solicitation in particular domains of interest via the JOBS act. These lists don't exist in databases today -- generally the research element is the more complex part.
We try to help people determine their ROI for lead generation out of the box, but LTV is something that a company has to figure out for itself!