What is Digg?Sponsored by Travelzoo
Take Advantage of Ridiculously Low Holiday Airfares view!
travelzoo.com - Flights $52 and up for Thanksgiving, Christmas & New Year. But move on it now.
30 Comments
- inactive, on 10/12/2007, -0/+15Tea; Earl Grey; Hot.
- aahpandasrun, on 10/12/2007, -2/+12Make sure you keep the Difference Engine far away, or else BOOOOOOOM
http://en.wikipedia.org/wiki/Difference_engine - dssstrkl, on 10/12/2007, -1/+9Booage to all software patents!!! Hopefully Google will not be evil and only use this patent to defend themselves against ambulance chasers.
- inactive, on 10/12/2007, -0/+6"a similarity engine generates compact representations of objects called 'sketches'. These sketches, which contain weighting according to placement vectors, can then be compared for similarities."
Well that explains everything. - zoom1928, on 10/12/2007, -1/+4It reads much like my patent on fingerprint matching that uses vectors to efficiently match fingerprints. The direction of the vector is the direction of the point of interest on the finger, and the length of the vector is the size of the point of interest. To make it easy to compare, the vectors are converted to a 2d matrix where the 3rd dimension is the length. You compare vectors with an error-covariance matrix to see if two readings are similar enough to be considered the same fingerprint. These methods have been used for at least 30 years. Google's patent (and mine) are completely bogus.
- lnxaddct, on 10/12/2007, -0/+3mrFreeze,
If you are interested in the specifics of the patent, Google had a talk about this hashing and sketching process back in November (although patents and stuff aren't mentioned... its just an academic discussion). They put it up on Google video, as they do all of their tech talks. It is about an hour long and is a little heavy in math in some areas, in particular as it relates to computer science, just so you know. I enjoyed it, the link is: http://video.google.com/videoplay?docid=-2473927019456727523 - MrBobDobolina, on 10/12/2007, -0/+2So... are we getting closer to an image seraching tool yet?
- spyrochaete, on 10/12/2007, -0/+2Just balance it out with an infinite improbability drive.
-(-) = + = sperm whale - AeonTorpor, on 10/12/2007, -0/+2Well, technically, wouldn't the similarity engine be completed by using the difference engine? So if i apply the ever-so long reaching arm of math-which-makes-no-sense, -(-)=+.... so... ... SHUT UP!
- lnxaddct, on 10/12/2007, -0/+1Many companies establish a patent portfolio for defensive purposes only. Microsoft is another big company that patents everything, but has never actively taken anyone to court specifically for a patent violation. They collect these huge number of patents so that when some other company threatens to sue them, the can simply write a letter back saying "You violate the following 20 patents of ours. If you'd like to proceed with legal action against us, we'll of course proceed with ours."
- jake13jake, on 10/12/2007, -0/+1Any information recognition patent is no good for the advancement of research into AI.
- lnxaddct, on 10/12/2007, -0/+1usually I'm against patents, but this is not your typical similarity matching. Google is taking a pretty novel approach here, in a higher comment I posted a link to the video of a talk they did back in November on this algorithm (and a few others). People have been doing similarity matching pretty much the same way you described for a long time without any huge advancements. Google kind of changed all this with the new algorithm, they needed to in order to allow them to effectively test similarities between billions of objects in milliseconds. It seems dumb to post the video twice in this thread, but here: http://video.google.com/videoplay?docid=-2473927019456727523 They start talking about similarity matching in the last 1/3 or so of the video.
- heyy, on 10/12/2007, -1/+2hey man google is in your closet waiting to chop off your head
- polyGone, on 10/12/2007, -0/+1I may be wrong here, but wouldn't it be more efficiant to make a non-similar engine? kinda like our immune system....ummm nevermind about that.......:)
- norris, on 10/12/2007, -0/+1This is a pretty logical step from collaborative filtering. I doubt it would hold up in court.
- jstohler, on 10/12/2007, -0/+1Another stoopid patent. Every major search engine uses some form of similarity analysis, so granting the concept to one of them is pointless. Patent the specifics, not the general idea.
- iloveliberals, on 10/12/2007, -0/+1For all those people who believe Google must patent this process to "protect itself," please explain how, exactly, this protects Google, and from what.
And don't say it protects Google from another company ripping off their process, patenting it themselves, then using the patent to blackmail Google. Google doesn't need a patent to protect against this scenario; it merely needs to publish to establish prior art.
Face it, kids: Google's like any other company, using patents and any other means to stifle competition. - Saiyanz, on 10/12/2007, -1/+2Next google patent they learn to scan your brain for relevant info stored away. But seriously with all the trouble that has hit google recently they need some good pr about now, maybe this will help them out.
- inactive, on 10/12/2007, -1/+1One the patent is issued other companies with vaguely related existing patents can't wait for google to make money from it and then go after them for 'damages'.
- Kythas, on 10/12/2007, -0/+0When I first saw the headline, I misread it to read 'singularity engine'. Now THAT would have been interesting.
- BRamBo, on 10/12/2007, -0/+0All that the method you describe and the google patent have in common are similarities...
Tho i have to agree the patent is pretty bogus, but i guess its a great way to cover your ass for upcoming concurrents... What more will google own ? - bierce, on 10/12/2007, -0/+0Google ceased to be a special case a while ago. "Do no evil" is not a sustainable business model (profit + growth) and G is not an exception. They are playing with themselves.
- dritix, on 10/12/2007, -0/+0Did anyone even read the patent?
Direct link: http://patft.uspto.gov/netacgi/nph-Parser?u=%2Fnetahtml%2Fsrchnum.htm&Sect1=PTO1&Sect2=HITOFF&p=1&r=1&l=50&f=G&d=PALL&s1=7158961.PN.&OS=PN/7158961&RS=PN/7158961
In claim 1 they just patented vector multiplication!
So much for do no evil. - plana, on 10/12/2007, -0/+0@dritix False.
What is claimed is a method for vectorizing graphics. I had this idea long ago- and guess what, emailed it to google :-) I wonder if they have my same idea in mind:
My idea goes as follows: user would draw a sketch and have it vectorized. Vectors are weighted and then matched against a database (wow sounds like what google patented, but mine is in the form of a user application).
Similar applications of this have been created before using image databases and using other conventional image scans using computer graphics algorithms of raster images, and then matching these using bitmasks for similar features. This new method (well, its been done before but google is patenting it) first vectorizes the images and assigns the vectors weight values. Create a database of vectors is faster to search than a database of raster masks, and is much smaller. - BullTaco, on 10/12/2007, -0/+0Other competitors can study the patent and patent improvements to it.
All competitors will then have access to and build upon the latest advances (accelerating innovation).
All competitors can sue each other and then figure out that they would be better off cross licensing.
Fifteen year old genius can read related patents and figure out an improvement and patent it.
Fifteen year old genius can then attempt to sell his patent for pennies to trolls since he can't pay enforcement costs and big companies ignore his licensing requests.
Trolls can then decide against buying patent because they can purchase trivial patents (trivial patents cost less to enforce) that are enforceable against many weaklings instead of formidable Google.
Fifteen year old genius can then learn his lesson, attend law school, graduate and excel at devising securities regulation loopholes. - drules101, on 10/12/2007, -2/+1im a bit confused. though this is interesting it is long and readers on the go wont want to read. ( by the way i love all the coments)
- BigSlacker, on 10/12/2007, -1/+0Uuuuuh...It protects themselves from spending the millions to develop a new similarity system and then have someone rip it off before they can make any money off it. Duhhh.
- BigSlacker, on 10/12/2007, -1/+0Typical knee jerk reaction out of ignorance. The problem with obvious software patents does not apply to a general problem with patenting algorithms.
- s73ve, on 10/12/2007, -3/+1LOL?
- Derelict267, on 10/12/2007, -3/+1p-placement vectors? oh noes.
© Digg Inc. 2009
— Content created and posted by Digg users is