Jessica Guynn has an excellent interview with Google’s Marissa Mayer today about Google’s first ten years (today is arguably Google’s tenth birthday). Good stuff in there – Marissa talks about Google’s accomplishments in search and advertising, and looks forward to a future where cloud computing becomes pervasive. Marissa also says she hopes to still be at the company in another ten years.
But one thing caught my eye. Marissa says search is “90 to 95%” solved:
Search is an unsolved problem. We have a good 90 to 95% of the solution, but there is a lot to go in the remaining 10%. How do we monetize new forms of content as they come online such as video, maps and books. How do we help content providers transition their businesses online and build healthy businesses.
Here’s the thing. I don’t think search is even close to being solved yet. In a May 25 post I talked about how early I think we are in search, and why a competitive search market is so important to make sure innovation keeps happening:
Innovation In Search Has Just Begun
I simply cannot believe that just a little over a decade into the commercial Internet, Tim O’Reilly is willing to say that the search war is over. Did he not read his good friend John Battelle’s book, The Search? He’s not the only expert out there who thinks the war is over – Danny Sullivan argued as much on the Gillmor Gang last week. But I simply cannot believe that this is all we can expect in terms of search innovation.
There are so many areas on search that remain to be conquered. Semantic search. Real language/AI search. The deep web. Media search. Today search basically returns web documents. What I want is for search to complete tasks for me. We’re no where near that today.
We are just getting started in search. To think that search has reached its pinnacle today is like saying aircraft were perfected before World War I. And if just one company were to carry on in aircraft innovation at that point, I doubt we’d have jetliners whisking us around the world today.
Innovation does not occur at a rapid pace without competition. If Google or any company were to control search exclusively, we could expect to see little happen in search technology or business models over even the medium and long term.
Sure, the odd startup or two would still come along and try to shake things up. But search is infrastructure intensive – the cost and difficulty of indexing the web and building a business in an established market requires resources that most new startups can’t realistically access. And if the market consolidates further, competing will become that much harder. There’s a reason monopolies get broken up by governments – market forces can’t generally undo them.
If search was 90% solved, Google could look at a picture of me standing by the Eiffel Tower and know, without textual metadata, what’s there. It could return results for a Barack Obama query that include all the videos he’s in, again without relying on tags or other textual metadata. Natural language. Deep web searches. Semantic search. All of these problems are unsolved.
This is not the long term solution to image search.
But anyway, Happy Birthday Google. You’ve done a lot in ten years. Just don’t give up on search yet.
Google may have largely solved the problem of search based on keywords, but where all current search models fall down is the ability to make subtle connections between related concepts without the use of keywords – like the human mind does.
To facilitate this, the next generation of search engine will allow every web user to make direct links between any two piece of web content. This will enable each piece of content to be mapped in relation to every other piece of content, providing a vital context that has never before existed. The perfect search engine will be part man, part machine so to speak.
This is what we have began doing at rowdii.com – users can view any web page while simultaniously viewing related links added by other system users. Anyone can rate the quality of links provided or add their own links to a particulat piece of content.
Once context between every piece of content exists, search engines will be able to do a much better job of intuitively generating useful results. And Google are not even close to reaching this point.
Hamish Robertson
Well… If you take a broader vision..
Scope really is 90% solved…
I mean. First you needed stuff to search on… Then you needed a really fast way to get it to people, then you needed capital to actually get after the problem, then you needed people to generate enough data for it to actually work …
Oh yah, and for the final 10%.. you need a workable scalable search algorithm.
Context is king (especially in search)
Wonder how many ‘The Search’ books this link sold on Amazon… you guys should join Associates…
Nice write up Michael and I certainly agree with your sentiment that search has not been “solved”. It reminds me of the infamous quote by Charles Duell “Everything that can be invented has been invented” – US Commissioner of Patents in, 1899.
As a former Googler I am surprised to hear Marissa say that. Whether a marketer or technologist is irrelevant, the search experience remains poor for many searches.
For example, take travel and the search for a holiday. I defy anyone to have a good experience – its a time wasting and frustrating exercise:
- Ads appearing by my search results that do not relate to the destination requested (think broad match for generic holidays)
- web sites “down for maintenance” yet showing in the search results
- web sites incompatible with the browser I am using showing in the search results
- review sites being listed that when clicked on only show “be the first to write a review”
I look forward to the day that a search engine can understand all of these and leaves out the crap so I can retrieve information quickly and efficiently.
Although search engines are light years ahead of what has gone before, the user experience has not really changed since the hey days of Alta Vista. The only radical attempt I have seen to change this is http://www.tafiti.com...
Brian
We would be interested to know also what Larry & Sergey really think about this issue. We think Search is only a 5% problem solved.
Moore’s law doesn’t apply to search and search (as it exists today) is 90% complete. Yeah, there are better algorighms out there, but they earn’t going to change search results that much. For example, the term “dog” already pulls the most relevant search results, so (short of changing the UI) there isn’t much to be improved. What I DO see changing is the WAY people search. Tag Clouds, for example, help us find stuff better. Eventually, someone will reinvent search and turn it on its head…
As for flight, how much has it changed in the past 30 – 40 years? Not much. Commercial aircraft still go about the same speed they did years ago.
“If search was 90% solved, Google could look at a picture of me standing by the Eiffel Tower and know, without textual metadata, what’s there.” Uh.. when did contextual image processing become the same thing as search? The problem is, we don’t know what “done” means, and tend to carry on very fluffy subjective discussions with faulty or overly-specific assumptions of “done.”
Another problem with this discussion is that it’s very easy for lay-people (especially technical lay-people: those with some technical knowledge but no in-depth expertise in the search space) to think they understand the search problem because they’ve used a search engine before. And this makes them think they have intelligent things to say about the answer, or even the scope of the problem.
This sounds like Microsoft’s attitude of “we’ve reached the limits of innovation” towards browsers (circa 2003) before Firefox and Safari entered the market and started taking share. I have trouble believing that search is even 10% solved, so I’m pretty baffled by Mayer’s comments. With an attitude like that, don’t be surprised if Google get’s disrupted by one of the many search startups gunning for Google’s mantle as the reigning king of search.
http://news.cne..._3-1011859.html
There are a number of searches that are far from being solved. Of course, many have to do with data collection. Search algorithms aren’t difficult if you have the data. Here are some examples:
- List all the public tennis courts within 5 miles of my house, including which ones are lighted (and the hours of operation).
– List all of the restaurants that will deliver food to my home after 12 midnight.
– List the closest public facility (restaurant, library, etc.) that is currently open and has Wi-Fi access within 3 miles of my current location.
Perhaps it look really big the 90% but she is one of the experts of the field, so I will take her words more seriously…
Listen up Google, Search is not 90% solved. It ain’t even 50% solved.
The business model of Search and Ads maybe 90% solved.
But the business model of AI Search and Live Online Purchasing – hasn’t even begun.
What about Vertical search like http://www.kayak.com and http://www.roost.com? These sites have vert-specific user interfaces that vastly improve the users experience.
Way off… Next generation RIA applications that build on frameworks like Adobe AIR, Silverlight and GWT / Javascript can’t be indexed by search engines today.
How about multilingual search? That’s definitely not solved. Natural language search? Not solved in any way. Any I agree with the comment above that human interaction hasn’t changed since 1995, there’s a lot of work to be done there.
Good day! problems with synthroid
Wow, it’s real big, to claim that 90% search is solved. Personally, I feel that it’s not even 10% or 20% solved. How many times we can find something from just 1 or 2 search. many times, it will be trial and error until we get it right.
—-
HDB for Rent