PETEGRIF

Machine learning isn’t always the answer.  This is apparently my top five photos as recommended by a system that trawls social media sites and recommends my friends’ photo to me.  It consists of images of two people I have never even heard of, much less met, and three images of a kitten (admittedly held by someone I know), two are duplicates.  I can’t imagine being less interested.

Machine learning isn’t always the answer.  This is apparently my top five photos as recommended by a system that trawls social media sites and recommends my friends’ photo to me.  It consists of images of two people I have never even heard of, much less met, and three images of a kitten (admittedly held by someone I know), two are duplicates.  I can’t imagine being less interested.

The Social Graph is just an Interest Graph

Don’t tell anyone but the social graph is simply a specialized interest graph.  It is the graph of those people in your life that you are interested in.  Is there any good reason to suppose that that graph maps naturally onto other interests in your life?  Yes and no.  Those people in your social graph with whom you have a group relationship by virtue of a shared interest will indeed map well.  But other groups won’t fit so comfortably.  You may share few interests with your family.  Your common interests with college buddies may be no better predicted by your social graph than by demographic data.  There is no way your work buddies would automatically share your interesting in recreating Civil War battles. So if you build a company focused on capturing the social graph, a la Facebook, you may discover that the resultant graph is not as powerful a predictor of interests and friend recommendations are not as influential as you might have imagined.  Hence the need for such sites to mine the data exhaust of interactions and declared interests to deduce interests.  This will always be harder than reading a person’s interests directly of their participation in a community focused around a specific interest.  This may be a community which exists for no other purpose or it may be a sub community within an environment which fosters and supports such sub groups defined by interests. This is why Pinterest, for all its copyright problems, is commercially interesting. It has a hard focus on interests.  Social is not its primary focus.  

Pinterest & Tumblr - copyright, mass infringement & their well chosen enemies

I strongly suspect that Pinterest/Tumblr have a real legal problem wrt copyright.  But this problem only becomes life threatening if they face a determined well resourced foe.  IMHO Pinterest/Tumblr have chosen their potential plaintiffs well and can probably manage the problem.

a) the legal case

It has been argued that Pinterest is protected by fair use or safe harbor.  I’m not convinced.

Fair use - a huge amount of the content they republish is not commented on, not scholarly, not partial or lower quality images or thumbnails, etc.  It is just the unadorned original.  That is not fair use.  Arguments to the effect that someone pinning it shows they like it and that such ‘liking’ is akin to a comment won’t stand up.

Safe harbor - I am not sure that the DMCA safe harbor provisions will protect you if the essential nature of your site or software is to enable copying rather than such copying being an unfortunate side effect of the technology.  P2P companies were shut down for this reason even though they claimed that they were just the conduit.  Furthermore, if the essential nature of your site/service is to provide such copying and your revenues derive from users accessing such facilities, even if your revenues are indirect monetization such as advertising, I suspect that it would be not only possible but convincing to argue that your revenues are indeed a direct financial benefit of the pirated content.  It would be very hard for Tumblr or Pinterest to argue that the extremely hard percentage of copied content on their sites is (a) an unfortunate side effect of their key business, or (b) irrelevant to their financial status - neither of them would have attracted the funding they have without the traffic and such investors clearly believe that traffic can be monetized.  So it’s a tricky area indeed, but IMHO Safe Harbor may not be enough if someone serious gets after them.

Well chosen enemies

There is a big difference between the enemies that P2P made and Pinteret/Tumblr’s potential plaintiffs.  And whilst this is not a legal consideration it is an enormously practical consideration.

Well resourced lawyered up media giants who saw widespread media copying as an existential threat went after P2P and crushed them.  What may purely pragmatically be in Pinterest and Tumblr’s favor is that I don’t see an equivalently resourced opponent who will go after them.  Their opponents are fragmented.  And right now they do not have enough $$ for a litigator to go after them on a class action basis.  

But watch out

If Pinterest and Tumblr explode in monetization and raise huge amounts of $ they become deep pockets and hence attractive targets for litigation.

Pinterest: a discussion of the Social Graph, the Interest Graph, User Experience Design and Startup adoption

Brad Feld recently posted an interesting piece on his blog entitled “Happy Birthday, I’m unfriending you” in which he describes the process he underwent to prune his social graph on the many social networks he participates in.  I found myself making a few contributions to the ensuing discussion and have summed up my key points here.  

Brad’s key point is that different social network encourage different social relationships and that there is no one size fits all model for how to manage your friends on a social network.  ( http://www.feld.com/wp/archives/2012/02/happy-birthday-im-unfriending-you.html#comment-436032025 )  This is absolutely true and is related to the really tricky problem for startups designing products with a strong link to social.  If you put yourself in the position of the product designer one of the trickiest questions to answer is - what is my model for social behavior?  If there is no one model for an individual how can there be for the site?  You can’t support every single possible model. Reading Paul Adams on social groups in the real world doesn’t solve the problem. Many startups are faced with difficult decisions. Friending vs following?  Groups of friends and if so how fine grain?  If groups how prominent do I make the filtering controls?  Does the UI at the point of posting as a producer introduce undesirable friction?  There are tools to ease the pain but even such smart tools as machine learning to create your groups so your group management is minimized don’t make the problem evaporate and what’s left can be a user experience problem. These are all difficult questions even if you are tightly focused on a single well defined demographic and have reason to believe you know what their preferred model will be.  But sadly things aren’t that simple.  Many products overflow such demographic segments.  If you were to design Facebook now you couldn’t afford to be paralyzed but on the other hand, the Facebook community clearly includes many segments with many very different attitudes to social behavior and with very different propensities and competences to manage controls.  Path’s decision to restrict the size of one’s ‘friends’ is an interesting approach to a tough problem.

There are however sites that encourage a huge amount of user interaction and yet are not primarily organized around the social graph.  Pinterest is a hot case in point. Pinterest is not primarily focused on the social graph but rather on the interest graph.  Setting aside for the moment the business model implications of this (which I suspect are favorable for Pinterest), it is very natural for people from differing social groups to share interests just as we all know that our social groups do not exhaust our interests.  Some people that we interact with little and would not figure highly in any network analysis of our social behavior nonetheless may share an important interest.  So with the interest graph the primary matter is not to solve the social group question with its associated UX but rather an ontology of content which can successfully span social groups. This isn’t trivial but IMHO it is more amenable to workable general solution and there is a huge amount of prior art in cataloguing and searching content. So as a rule of thumb my hypothesis is that interest graph focused companies have the easier design challenge.

I also hypothesize that interest graph companies can grow quicker faster because the usefulness of such a site is not entirely dependent on the network effect.  Whilst people like Andrew Chen argue that any product must have intrinsic worth that can be enjoyed for its own sake it is hard to see Facebook through that lens.  FB and other such communities are about social communication and sharing and in such cases the network effect is a drag on adoption.  Unless you can find a way to concentrate your adoption to provide network benefits it is likely to be a slog.  FB focused on Harvard, then other Ivy’s then other schools…  Yelp focused on San Fran, then the next city, then..  In both cases the concentration provided early adopters with more network benefit than would otherwise have been the case.  If you consider Pinterest in this light it is clear that any user can derive real benefit from an easy to use beautiful online pinboard.  They can use it to collect and order content of interest to them and revisit such organized content even if they don’t share it with anyone.  It is more similar to Evernote than Facebook. A person has interests irrespective of their groups!  And products can be built that address a specific interest (eg wine!) or provide a platform to support any interest - Pinterest. The clever thing about Pinterest is that it is the combination of user generated content and the interest graph.  And because it is just pinning it is incredibly easy to use and highly suggestive of interest.  This suggests a very different model for say a wine site than a traditional magazine format where the user is relatively passive and of course that passivity is harder to reliably mine than when the user actively selects the material and pins it.  ’Like’ buttons and the like are, IMHO, a weak substitute in the understandable desire to capture interest.

Social graph, interest graph and website organization

In an earlier post I noted that not all websites should be organized around the social graph even if they could be.  A lemma to this is the difference between the social and interest graphs.  They are by no means identical.  Even if the recommendations of our friends have disproportionate weight with us by no means all our friends share all our interests.  In fact it is perfectly possible that none of my friends share one of my interests - it may be a deep dark secret. Organizing a website by the nature of its content is to organize it by its interest graph.  So perhaps “old fashioned” sites aren’t dinosaurs after all.  The question now is - can one employ something like  the functionality offered by social sites to enhance to experience of content driven sites?     Can one, for example, crowdsource common interests?  Yes you can.  That is what comment boards do.  Lots of people on such boards share an interest but are not friends.  A new twist on this approach is Pinterest.  Can such shared interest communities be better leveraged for user benefit (and commercial gain)?  That is what sites like pinterest are figuring out.

Organizing web content - the social graph isn’t always the answer

The idea is that the really important organizing principle for content is the social graph.  The evidence seems to be that sites built around such a graph are more engaging and more credible.  More engaging because human beings have a high degree of interest in the behavior of their immediate pack (friends) and significant alphas (Hello magazine) and more credible because time is short and we value the opinions of those we already know and trust. This line of thinking is, for example, persuasively discussed by Paul Adams in ‘Grouped.‘ 

That there is value in such insights is indisputable, but does it really mean that all content is best organized in accordance with social relationships?  Is it even possible?

The answer seems to be that it is more possible than you might imagine.  News channels increasingly segment audiences and present the ‘news’ not in accordance with any kind of independent journalistic or meaningful editorial standards but rather to hew to the views of the targeted segment.  And these ‘views’ may never have been directly expressed but rather inferred by the analysis of the big data trail of user behavior.  But, as has been observed by concerned critics, if we present news in a manner designed to appeal to preexisting beliefs we are doing no more than shouting into an echo chamber.  Segmented news that reflects an existing interest graph does nothing to challenge but rather serves only to reinforce existing beliefs and sadly existing prejudices.  Accordingly there are some forms of content that are perhaps best NOT presented in accordance with declared or inferred interests.  And that it may even be dangerous to so organize content.  

This observation turns out to be profoundly important to the way material is in fact organized.  It is lazy to simply go with the new conventional wisdom and throw one’s lot in with social.  There can be no general answer to how content is organized.  In some cases perhaps the social / interest graph is an appropriate metric and in others it may be wholly inappropriate. The point is to reflect on the matter not to just leap into bed with facebook.


#coffeeshopinvasion

#coffeeshopinvasion

N. Cal v S. Cal for startups

There has been much debate on this point.  Fingers have been pointed at the more developed financing infrastructure available in N. Cal as a key factor fostering startups.  And there is certainly truth to this.  Despite the recent explosion in incubators and accelerators in Los Angeles the follow-on financing is immature.  This is important, but is it the most important factor?  IMHO - that depends on what kind of company you are because that determines the kind of people you need and the bottom line is that the ‘staff resource pool’ of these two regions is very different.

The conventional wisdom is that N.Cal is stronger at hard core engineering and S.Cal’s strengths are in content/media.  I am enough of a contrarian to not want to take this for granted so I have spent countless ideas browsing resumes.  I wasn’t trying to prove S.Cal was as strong, as N. Cal on platforms, that was obviously wrong, I just wanted to truly know if despite the pool’s undoubted relative weakness it was nonetheless deep enough in absolute terms. I can now conclude that the conventional wisdom is dead on and there is real staffing risk ‘down south’ if you need to scale hard core engineering.  Whilst it is possible to find such engineers there are !0x as many up north. Conversely if you want people with content skills there is no shortage in Los Angeles.

You may well be thinking that I have painfully arrived at the blindingly obvious.  And you are right.  There is however a wrinkle.  It is not always obvious what kind of company you really are.  Are you an engineering company or are you a media/content/entertainment company?  In the old days this was perhaps easier.  If I am building chips I am engineering.  If making movies I am entertainment.  But with media moving online this is trickier than ever.  IMHO there are plenty of companies who may seem like content companies but whose core competence is in fact engineering.  This is not to say that content is unimportant to them, but the truly hard problems they will face are engineering problems as they scale and if they don’t truly ‘get’ this they are destined to run into major problems.  Technology is changing so fast that without reasonably ready access to the best possible engineers your platform can suddenly become obsolete and in this case the best content in the world isn’t going to save you.

Paul Graham wrote a fascinating piece on this very point which is worth rereading regularly. “What happened to Yahoo?”  Of particular interest is his thoughts on the internal culture of Yahoo and how it progressively devalued engineering.

 http://ils.unc.edu/annotation/publication/Fu_etal_ASIST05.pdf

You do not have to agree with his analysis of the Yahoo experience but the point is well made - it is not always blindingly obvious (especially to senior management who do not themselves necessarily have a tech background!) quite what your core competence is.  Is Youtube an engineering company or a media company?  What do their senior management think?  One thing is blindingly obvious - the technical challenges of operating Youtube at its scale are enormous and ready access to world class engineers is mission critical.

You have to be somewhere that supports your core competence and for engineering that is N.Cal. The more technically demanding your platform/application the truer this is.  The fact that there is less follow-on financing in S.Cal is nothing like so big a problem and N.Cal investors know it.  Conversely, if you need people who truly grok content and the platform is likely to be reasonably stable S.Cal is the biz and if by leveraging this skill pool you build a great product IMHO you will find investors.

People are more important than money.  Without the right people money is worthless.

Visa application process = insanity

Can anyone think of a good reason for including questions like this on an immigration form?  

'Do you intend to engage in acts of terrorism whilst in the United States?'  Y/N

Call me naive but I strongly suspect that few terrorists answer ‘Y.’

Immigration visa for CTO is a nightmare

I am afraid it is time for a rant.

I recently applied for a visa for my co-founder and CTO.  We are funded and he is incredibly well qualified and experienced.  One may have had some dim hope that the thick wad of material we duly sent confirming these facts would have sufficed.  Dream on!  We just received a document entitled “Request for Evidence”  (RFE).  This document makes it luminously clear just how poorly understood our startup world is.  I am sure that if we were General Motors or indeed any other household brand going concern all would go swimmingly.  But for a startup the tedious BS you have to wade through is a mindless time suck.  A substantial part of the RFE was a detailed explanation of how important it was for me to prove beyond reasonable doubt (citing multiple authorities and collecting affidavits willy nilly) that a company such as mine needs a CTO.  It’s too depressing.  I have to go see the dentist now - I have ground my molars to the stumps.  

Bring on the startup visas.  We need them so desperately.

Extensions, native client (NaCl) & the chrome app store

I am a big fan of NaCl and so, dear reader, please take this suggestion as an attempt to be helpful.  Currently the store navigation lists the categories of apps:

So we have Education, Family, Games, Lifestyle, News & Weather, Productivity, Shopping, Social & communication, Utilities, Extensions, Themes. Furthermore, one can search for NaCl apps in the search box.  What’s wrong with the picture?

IMHO there is a marketing error being made here.  ’Extensions’ means nothing to users.  It is not a useful category of app. It is a category distinguished by technology.  Apps should be under categories that are meaningful to users irrespective of how they are implemented.  Calling something an extensions draws attention to the wrong thing.

Similarly, as NaCl makes its mark users should not care about whether an app has been implemented by NaCl.  There is no need to draw attention to the fact.

So I say - abolish the ‘extensions’ category asap and lets keep the focus on things that matter to customers.

Mozilla’s Firefox and the Google deal

Firefox is losing market share to Chrome.  And Mozilla is overwhelmingly financed by Google.  The paid placement deal whereby Mozilla gets paid by Google for hosting its searchbar has come to an end yet no replacement deal has been announced.  There has been speculation that Google will not renew the deal and   will thereby stick a dagger into the heart of a browser competitor.  E.g. http://www.businessinsider.com/and-will-google-now-kill-off-firefox-completely-by-pulling-the-plug-on-its-toolbar-deal-2011-12  It would seem that this would provide an opening for Microsoft to step in a buy search market share.  But there has been no sign of such a deal yet.

A further wrinkle to this has been Asa Dotzler’s assertion that Mozilla does not ‘auction search.’ I had an interesting if inconclusive exchange with Asa about precisely this point on Quora. (attached below).

My suspicion is that google does not need to wipe out Firefox (just winning market share from them organically in a controlled way will do just fine) and they will not give up the deal and hand MS market share.  I suspect that negotiations are ongoing and the deal is a matter of price.

===================

Asa Dotzler Yes, there was an answer. tl;dr “It doesn’t matter if Microsoft tried to outbid Google. Mozilla doesn’t auction search defaults.”
Mike Teasdale The answer is a bit misleading because Mozilla does make revenue from its deal with Google - something like 85% of its revenue has historically come from this source. And according to this story from TechCrunch http://techcrunch.com/2008/08/28… the current three-year deal expires later this year - so the question is well-timed.

My guess is that domination of the browser market is much less important to Microsoft than domination of the search market, so yes, I’d be surprised if they didn’t try to make a deal.
Asa Dotzler Mike, the question was “will Microsoft bid for a search deal with Mozilla” and the answer is that Mozilla hasn’t and won’t open search defaults to a bidding process. That’s not how Mozilla operates.
Pete Griffiths
Asa - your answer is helpful in reminding us of Mozilla’s mission and stating that non-financial mission related considerations would bear on the decision is totally fair but surely it is fair to ask you to be more specific because I have to say I do find your reply so general that it isn’t as helpful as it might be.

How about if I pose the question like this: how would Mozilla’s users and the web community be affected if Bing rather than Google were the default search for Firefox? What difference would it make to anyone? Still more specifically, what difference that bears upon your mission?

Because it seems to me that if there is no convincing answer to that question it is perfectly rational for Mozilla to hold an auction and use the resultant financial gain to advance its goals and mission.
 
Asa Dotzler “How would Mozilla’s users and the web community be affected if Bing rather than Google were the default search for Firefox?”

More of Mozilla’s users would use Bing and fewer of Mozilla’s users would use Google. Some, probably a significant amount, of search market share would shift away from Google and towards Microsoft.

"What difference would it make to anyone?"

Users for whom Bing was a less good solution than Google would have a less good experience. Users for whom Bing was a better solution than Google would have a better experience. Google would lose some users. Bing would gain some users. 

"Still more specifically, what difference that bears upon your mission?"

Mozilla’s mission is not as narrow as you might assume. We have a dual mandate. First, it is to ensure that users are in control of their online lives. We advance user sovereignty through Firefox directly and by driving competition that has the effect of improving other browsers too. Second, we are stewards of a healthy and prosperous Web ecosystem. We care about improving the Web platform for indy Web developers, Web frameworks, large-scale Web development teams, and the businesses and other groups that offer their goods and services on the Web.

Clearly browser competition has been a big win for users and Web developers over the last 5 years or so. I personally believe that meaningful search competition would be good for users and developers too. Search is a fundamental part of the Web and though Mozilla does not have a search service, it definitely has a role to play in making search easier, more capable, and more powerful for everyone.
Pete Griffiths
Asa: I know you are absolutely sincere about the importance of Mozilla’s mission. And I certainly agree that Mozilla has been a huge win for users and web developers. But to be honest I don’t think you answered the questions I posed. I am quite prepared to believer that there are great reasons why it would be a bad thing to do a deal with MS but I don’t think you have made the case at all. And, to repeat myself, if you can’t make the case then the money is surely helpful to help Mozilla pursue the mission that you value so highly. I don’t mean to be a pain here, but I do think clarity on this point is helpful. 

Software IP protection & the weakness of copyright

There are those who oppose software patents and fondly imagine that such protection is already afforded by copyright.  An important recent ruling by the European Court of Justice, the highest court in the European Union, has recently made it luminously plain just how completely mistaken this position is.  Their ruling states categorically that functionality is not protection by copyright, just expression.

http://www.osnews.com/story/25373/EU_Copyright_Doesn_t_Cover_Functionality_Programming_Language

Kindle Fire

On April 20 I wrote a post “Amazon to be Apple’s major tablet competitor.’  It was a short piece and the text is reproduced below.

The power of the Apple offering is its soup to nuts nature.  The entire user experience from browsing to app & content synchronization and management to one click payment has been designed to be seamless to a degree that others can as yet only dream about.

It is this user experience design that is ruling the roost and making Apple the dominant force in tech right now.  So who can challenge them?  Quite a few companies have the ambition, but can they execute?  One who probably can is lower profile than most but IMHO is a serious player.  

Amazon has so many of the required pieces to create a soup to nuts offering.  The hardware is the tip of the iceberg but even here they have demonstrated remarkable competence with the Kindle.  They already have in place the world’s most sophisticated consumer merchandising operation.  They have the credit card information of countless millions and the trust of those consumers.  They have an enormous scaleable IT infrastructure in place.  Can you imagine browsing music, video, apps on an Amazon device and then purchasing with one click and getting fantastic customer service?  It could happen.  I suspect we will see Amazon enter the market this year with a very low cost device and become a huge player.”

Well…pretty much nailed it.

Needless to say a bunch of naysayers have dissed the product.  My memory of Apple’s iphone and ipad being so scorned is very fresh.  Sadly many reviewers seem to have an incorrigible inability to recognize that (a) a new product can define its own market, and that (b) the fact that the product is not perfect on launch does not mean that it won’t get better…and better…and better.  Let’s not lose sight of the fact that Amazon has huge resources.  

For a more intelligent review of the Fire try this:

http://techcrunch.com/2011/11/25/the-kindle-fire-what-is-it-good-for/