tag:blogger.com,1999:blog-2425290326823263574.post6504811479818128769..comments2022-12-04T18:48:06.405-08:00Comments on Krazy Glew's Blog: New models for Industrial Research (in reply to: The death of Intel Labs and what it means for industrial research)Andy "Krazy" Glewhttp://www.blogger.com/profile/08442494949914217568noreply@blogger.comBlogger4125tag:blogger.com,1999:blog-2425290326823263574.post-10696603093189592892011-04-26T16:50:33.180-07:002011-04-26T16:50:33.180-07:00The lablets' mission was never to do architect...The lablets' mission was never to do architecture research but rather to focus on complementary areas -- OS, networking, software (targeting SOSP, Sigcomm, PLDI, NSDI -- not ISCA). This is also the community Matt Welsh comes from. I suspect that if your research interests aligned with those areas, you wouldn't be unaware of their research contribution, nor suggesting the lablets weren't "effective" at least in terms of generating high quality research output. You may be right that the lablets struggled on tech-transfer but keep in mind that it's much harder to take a software innovation to an Intel business unit than a hardware one. MRL was designed to be much more closely aligned with Intel's BU interests. I suspect the lablets closure reflects as much on Intel's continued inability to branch out from their core expertise rather than the quality of research that emerged from the lablets.<br /><br />It also sounds like your main goal is to point out that you invented lots of things before other folks did. Feeling under-recognized? :)Anonymousnoreply@blogger.comtag:blogger.com,1999:blog-2425290326823263574.post-19505308627887178442011-04-10T11:28:18.261-07:002011-04-10T11:28:18.261-07:00Who moved where?
If they moved on to products in ...Who moved where?<br /><br />If they moved on to products in Intel, then it may have been worth Intel's investment (although if they were bringing more ideas like Itanium, not so).<br /><br />If they moved on to competing companies like Nvidia or AMD or ARM or IBM, then it may have been a waste of Intel's money. (Although if they brought said Itanium-like ideas, maybe that would be a good thing.)<br /><br />If they moved on to academia... unkown and uncertain value.<br /><br />--<br /><br />Note that I am not opposed to a post-doc place. I'm just saying this from a company centered ROI point of view.<br /><br />Doing stuff just for the good of it is good, but is not what company's do.Andy "Krazy" Glewhttps://www.blogger.com/profile/08442494949914217568noreply@blogger.comtag:blogger.com,1999:blog-2425290326823263574.post-52110804911206417192011-04-10T11:23:43.672-07:002011-04-10T11:23:43.672-07:00Matt Welsh's post is at:
http://matt-welsh.bl...Matt Welsh's post is at:<br /><br />http://matt-welsh.blogspot.com/2011/04/death-of-intel-labs-and-what-it-means.html<br /><br />It also happens to be linked from the title of my post - apparently that is where Google's blogger puts the reference link, unfortunately.Andy "Krazy" Glewhttps://www.blogger.com/profile/08442494949914217568noreply@blogger.comtag:blogger.com,1999:blog-2425290326823263574.post-86556736065304240442011-04-09T17:44:47.061-07:002011-04-09T17:44:47.061-07:00"[T]he best technology transfer is by transfe..."[T]he best technology transfer is by transferring people."<br /><br />By that measure, I know the Berkeley Intel lab (or lablet) was successful. The folks I knew who went there moved on with more skills. I always saw it as a post-doc place while I was a grad student, and I suspect the faculty saw it similarly.<br /><br />Industry/academic labs never last forever. Both sides face different problems that change via different forces at different rates. Both the Berkeley and CMU labs housed good people as post-docs, produced (or continued) good software, and produced useful data. Perhaps not great for hardware, but good for software and networking. For example, PlanetLab helped gather information on virtualization before it was widely deployed by companies.<br /><br />I haven't read the post you cited (link?), but I see the Intel labs as a successful model. They stayed small enough (IIRC) to be shut down when the different sides' needs grew apart. It's not like Berkeley, CMU, etc. no longer are collaborating with Intel; they just no longer share some management and financial meetings. Same thing happens with NSF centers. The supercomputing centers are a pain because they're too big to be shut down without passing along much pain (SDSC).<br /><br />BTW, have you seen any deep work coming out of Google's 20%? I haven't, which makes me think along the same lines as you: concentration doesn't come in k% increments.Anonymousnoreply@blogger.com