Sign in or Join FriendFeed
FriendFeed is the easiest way to share online. Learn more »
Thomas Hawk
Microsoft Officially Buys Powerset -
I've yet to be impressed by Powerset. What am I missing here? Why did MSFT pay $100 million for this? Seems like grasping at straws more than anything to me. I firmly believe the future of search lays more with companies like FF where they can eventually turn on powerful social search functionality than something like Powerset, which can't even effectively search wikipedia today. - Thomas Hawk from Bookmarklet
Sarah Meyers video interview with Powerset CEO - Thomas Hawk
I just don't get how money follows things that clearly do not work. Back when Riya was supposed to be doing facial recognition technology it just simply did not work. I tested it personally. It didn't do what it was supposed to do and still VCs dumped truckloads of money into the thing on a whim. Same goes for Powerset. Their search sucks. What am I missing? Will a public co. like MSFT really pay $100 million for simply an "idea"? - Thomas Hawk
hmmm... I have an idea!! - Damien Franco
wtf is Powerset may i ask? - Outsanity
Powerset search for "Thomas Hawk" Google search for "Thomas Hawk" Google's results are vastly superior. - Thomas Hawk
It's my understanding that the "big deal" with Powerset is that they use "natural language". So how come when I type "How do I clean a sensor?" into Google I get such better results than Powerset - Thomas Hawk
IIRC, is a public "proof of concept" site that demonstrates powerset's semantic query capabilities, based on a couple of small datasets. It doesn't actually index the interwebs yet. - Chris Hollander
Powerset has only indexed Wikipedia. Google indexes everything. That would explain most of the difference right there. The point of the semantic search is to (basically) help you answer something. What Google and Live do right now is more of a keyword search. I, for one, am *super* curious to see what happens when something like Powerset hooks into a full index. - Jordan Hofker
Chris, but the "proof of concept" returns terrible search results. Maybe I'm just not getting it. Someone enlighten me. Can anyone come up with an example of better search results using Powerset vs. Google? I don't mean to be a skeptic here, but beyond the hype, why is Powerset worth $100 million? - Thomas Hawk
hmmm... so Powerset is a enhanced Wiki search from what i see. i guess not getting Yahoo really upset Microsoft - Outsanity
They are trying to keep up with Google and don't know how. I don't think MSFT knows what they are doing and they are going to be on the same path as Yahoo (over a longer period) if they don't watch themselves. - Justin Korn
$100 million "leap of faith" seems like a big one to me. Smells more like desperation if you ask me. - Thomas Hawk
100m is not a big leap of faith at all for MSFT. They've paid more for less - Michael Gartenberg
Michael, was this a sensible purchase? - Thomas Hawk
The reality is that we're probably 3-5 years away from a decent semantic search. Powerset did a proof-of-concept using a homogenous set of data (wikipedia) and gets so-so results. Semantic analysis today only works in very niche areas where the user will invest time to tune the engine to match the content. So, what Microsoft bought was a lab experiment. I think that, more than anything, they wanted to keep it out of Google's hands for fear that Google might be able to leverage it more quickly. But, it does nothing to strengthen MSFT's position in the search market in the near-term but $100M is not a huge amount for MSFT to spend to keep it away from Google. I could also see MSFT able to leverage some of Powerset's capabilities within their FAST enterprise search platform. Perhaps that's where they'll make back their investment. - Barry Graubart from twhirl
$100m isn't much when you think of this purely as an R&D investment. They get dozens of engineers familiar with semantic search to work on Live Search - Jamie
@Barry We might still be 10-30+ yrs away from Semantic search. I worked on it in 1992-93 with Professor Michael Dyer and "big" data sets of the time were 188-195 MB. See TREC-1 from 2000. His 1982 Ph.D. was "In-Depth Understanding: A Computer Model of Integrated Processing for Narrative Comprehension" - Mitchell Tsai
@Thomas Hawk: Why you say semantic search doesn't work? I've seen and made it work wonderfully plenty of times. The only disadvantage is the need to recognize the user more deeply than current search solutions, that means it will work well with users who are already have profiles on the system. - Amit Morson
The internet swamped "semantic" search, and Bayesian Net statistical approaches won. See Eric Horvitz - Research Area Manager, Adaptive Systems & Interaction Group, Microsoft Research Microsoft raided UCLA and hired some of Judea Pearl's best people. MSFT has the cash to spend $1 billion on a product launch, and chase "pie in the sky dreams". I hope they succeed. - Mitchell Tsai
Incidentally, I got my first "F" in Professor Dyer's class. I insisted that a statistical approach would be better. About 10 years later, we had a fun chat about that... :-) - Mitchell Tsai
This makes a LOT more sense for now (than the price offered for Yahoo). It's a good step. - Charlie Anzman