alexr_rwx: (unreliable narrator)
[personal profile] alexr_rwx
We're in Mountain View again! We're back at basically the same jobs as last time: Lindsey is working on Rust, Mozilla's new programming language, and I'm doing rad stuff on Google Translate.

We've been seeing Martin a fair bit (and Lauryn sometimes too, but not quite as often!), since he's got to come down to the South Bay for work here and again. We spent the morning and afternoon with them; Lauryn was involved with the SF Carnaval parade, so we went up to the Mission and watched the parade; it was pretty wild. Lots of people dancing and wearing crazy costumes and playing music and walking around on stilts. Then we had burritos and coffee and looked and books and hung out for a while! (hooray! :D )

Lindsey and I have been thinking about what we want to do when we're done with gradschool, which will be, y'know, sooner or later. We could try to get the full-time versions of our respective intern jobs; that would be a pretty good result. We could stay in academia and look for postdocs or possibly faculty jobs...

Or we could start a business. Let me lay this business plan on you: we outsource your business's management. You want to have a tech company or a widget manufacturing company or whatever. But you don't want to have management; it's too expensive and boring. This is where we come in. You contract with us, and we provide virtual managers for your company, as many as you need. They have plausible-sounding names. But underneath, your virtual corporate management is actually outsourced to very qualified people somewhere far away where labor costs are lower. And while we're presenting you the interface of a consistent set of virtual managers, the role of Alice (or Ashok, or Andy, whatever you want) is actually being played by some indeterminate number of hot-swappable qualified people, who may be switched out so that we can do better management load-balancing, to provide you better quality-of-service.

And then: we farm out the management decisions to Mechanical Turk, and get them for next to nothing (checking inter-rater reliability so that only sane-ish business decisions get presented to your business by your virtual managers), while still presenting you the same management interface.

Great business plan? Or greatest business plan... of all time?

Date: 2012-05-28 04:26 am (UTC)
lindseykuper: Photo of me outside. (Default)
From: [personal profile] lindseykuper
I just want everyone to know that I had nothing to do with the creation of this "business" "plan".

Date: 2012-05-28 04:28 am (UTC)
ext_110843: (mighty penguin)
From: [identity profile] oniugnip.livejournal.com
It's OK! We can still be equal business partners when this launches. <3

Date: 2012-05-28 04:51 am (UTC)
From: [identity profile] gwillen.livejournal.com
I dunno, it seems pretty compelling.

Date: 2012-05-28 05:17 am (UTC)
From: [identity profile] mindstalk.livejournal.com
I think you need some derivatives and betting markets in there too.

Date: 2012-05-28 05:25 am (UTC)
ext_110843: (cat exoticizes the otter)
From: [identity profile] oniugnip.livejournal.com
I was imagining using the whole process as a source of labeled training data for systems to come up with good business decisions: eventually, we could start falling back to the crowdsourcing only in cases where we can't use classifiers or CBR or whatever to produce management.

Also we could have... futures... on management? I dunno how these things work really.

Date: 2012-05-28 06:46 am (UTC)
lindseykuper: Photo of me outside. (Default)
From: [personal profile] lindseykuper
And eventually we can bootstrap the business and use it to manage itself. Then you and I just sit back and sip champagne in the hot tub while the whole thing carries on self-sustainingly.

Picture it: people on MTurk being called upon to make decisions about how to, say, apportion tasks to people on MTurk.

Date: 2012-05-28 07:19 am (UTC)
ext_110843: (removal of signs)
From: [identity profile] oniugnip.livejournal.com
Oh hecks yeah. How is this going to have any credibility if we don't dogfood it? Importantly, once this is MTurkable, the MTurkers should only be called upon when the machine learner doesn't know what to do, or when it's detected it needs more training data in an interesting part of the space...

Date: 2012-05-28 10:36 pm (UTC)
From: [identity profile] tristmasjedi.livejournal.com
And eventually we can bootstrap the business and use it to manage itself. Then you and I just sit back and sip champagne in the hot tub while the whole thing carries on self-sustainingly.

I think I've seen this theory in some of the more cognitively-motivated areas of AI.

Date: 2012-05-28 05:10 pm (UTC)
ext_1785153: (Default)
From: [identity profile] deepdistraction.livejournal.com
I could see it for some decisions, like finding/maintaining the right mix of B2B services or equipment to buy, or optimizing cash management, or disaster planning. Not sure it would work with hiring decisions or really keeping customers happy. I could really see it for suggestions on future strategies, which would involve keeping good tabs on the competition and on technology.

Date: 2012-05-28 08:26 pm (UTC)
lindseykuper: Photo of me outside. (Default)
From: [personal profile] lindseykuper
suggestions on future strategies

Yep -- that's why the consulting profession exists! Lots of businesses could stand to benefit from outside help with strategizing.

(I'm pretty sure Alex was joking, though...)

Profile

alexr_rwx: (Default)
Alex R

May 2022

S M T W T F S
1234 567
891011121314
15161718192021
22232425262728
293031    

Most Popular Tags

Style Credit

Expand Cut Tags

No cut tags
Page generated Jul. 19th, 2025 11:37 am
Powered by Dreamwidth Studios