Does Google do "research"? | Hacker News Reader

Does Google do "research"?

matt-welsh.blogspot.com

70 pointshiteshiitk19 comments

Reading this article confirms my tremendous respect for Google for:

1) Creating a culture where writing unit tests, fixing bugs and the like are not viewed dirty work or an overhead, something that scientists (or even engineers) are above

2) At the mean time, being a place where creation of new technology (even if not strictly related to the primary product i.e., not just information retrieval algorithms) is not viewed as "academic" or a risk to be avoided at all costs.

There are companies which do (1) well and companies that do (2) well; they have created wealth and made products I enjoy and use. Only a handful managed to do both: they changed far more than their industry.

I was spoiled as my internship January last year with Google was my first introduction to the world of corporate IT. Google as a culture really encourages engineers to follow best practices - "Testing on the Toilet"[1] was always a highlight.

As you say, the emphasis on unit tests is amazing and something not found in a lot of professional companies. When you tell someone you haven't committed your code yet because you're writing unit tests they don't tell you to hurry - it's not considered something you can rush.

Additionally some of the research that has come out of Google Research [2] is stunning and I suggest if people are interested they can keep tabs on new publications via their feed [3]. A lot of their tools and products really are pushing the boundaries of the field they're in, both in complexity and scale, so the "reports from the front line" offer a unique insight.

[1] - http://googletesting.blogspot.com/2008/08/tott-100-and-count... [2] - http://research.google.com/pubs/papers.html [3] - http://research.google.com/pubs/atom.xml

The danger with integrating research as part of day-to-day development is that you might not end up with groundbreaking or disruptive innovation. The problem is that forces which are controlling day-to-day work are essentially controlled by value network of the company (social and technical resources within and between businesses) and constrained by resources, values, and processes of the company. So, there is a danger of missing inventions which will replace the current (profitable) way do doing things (i.e., start Groupon kind of business model, or Twitter, or Facebook).

However, I also 100% agree that having pure "research labs" is not effective: in many cases some great idea from research arm are just forgotten.

Clayton Christensen wrote a few books on this topic (and I am a very big fan of Clayton Christensen's work).

I would like to also mention that Google is not the only company which has this approach: it seems to me that majority of companies in Silicon Valley operate like this (I don't think Facebook has "research labs", but they do invent things).

> forces which are controlling day-to-day work are essentially controlled by value network of the company (social and technical resources within and between businesses) and constrained by resources, values, and processes of the company.

Google's 20% Time could be considered "research" or R&D not constrained by business requirements. 20% Time seems like a clever way to harness engineers' downtime for work that may benefit the company.

But none of the Google engineers I've met had a 20% Time project. They said product teams (like Android) had much less flexibility for 20% Time than people working on google.com properties.

Two disconnected thoughts that came to mind reading this...

1. I agree with Google's approach. Microsoft spends $10 billion a year on R&D yet I can't think of a revolutionary product they've come up with (Kinect was bought from the outside)

2. I notice the word "patent" is never mentioned in this piece. A big reason companies like Microsoft and Intel have research labs is to patent protect themselves in the future (wasn't there an article on HN just a few days ago that mentioned Google's anemic patent portfolio?)

As were Google earth, gmail, google docs, wave etc

GMail was definitely internally done (by Paul Buchheit). I'm pretty sure Wave as too.

But yes, maps/earth, docs, voice, etc. were acquisitions.

I agree mostly, though your assessment of Microsoft Research is not completely true. Don't forget they came up with the Songsmith :D

I read repeatedly that much of Kinect's software was derived from their in-house R&D. Is that not true?

It is true. Microsoft bought from PrimeSense the IR depth sensor technology, but all the body recognition/tracking software was developed by Microsoft Research.

I don't know specifics about Kinect, but I do know that MSR has done a ton of work on computer vision and its applications over the last ten or fifteen years, and I would be extremely surprised if that work didn't inform the efforts of whoever did invent Kinect.

The computer vision aspects to the Kinect (and there are a lot, and very well done), were headed up by Andrew Blake at MS Research Cambridge, who's one of the foremost vision scientists in the world.

http://research.microsoft.com/en-us/um/people/ablake/

Most recently they have built the machine learning technology inside the Xbox Kinect human motion tracking system.

MS's approach to research is fundamentally different than Google's. Google's approach is about creating better Google products. Microsoft's approach (which is really Bill Gates's dream) is to advance science.

Both are good goals. But if you ask a researcher from MS how important it is to get in a future version of Windows, they may well shrug their shoulders. Their feeling is if they can advance technology then all boats rise, which is good for Microsoft.

I must say when I'm actually doing research, and not just dev work for product Version Next, I don't want to do unit tests and a whole bunch of other cruft. My code isn't meant to be shipped to customers. Its meant to test a hypothesis.

MSFT research is much more like research of great American companies of the past, Bell, AT+T etc - I'm sure this is what they were aiming for.

This is what scares me about Microsoft Research. Since Google's research adds direct value to what it's doing, it is quite unlikely to disintegrate.

I suspct if left completely up to Sinofsky or Ballmer MSR would shut down (or be substantially smaller). I think as long as Gates is alive and/or has large influence, it will be a relatively big part of MS.

I agree. For example, Leslie Lamport, who invented many distributed algorithms that are used at Google (eg. Paxos is used in Chubby and in Megastore) works at Microsoft Research. Watching or reading interviews with him, I get the feeling that he is a "pure [computer] scientist", probably not that interested in getting his stuff into the next product.

The appeal of Google for many researchers is that you can test your hypothesis much more broadly when actual people use it than when it's confined to a research lab. Having a billion+ actual users generates tons of data which can itself spawn new research directions and inform new hypotheses.

Isn't what Google practicing a kind of applied research? Where MS Research is doing basic research which advances the science.

Reply on news.ycombinator.com