Although Thirst is starting out on the Twitter platform, the company is really more about natural language processing technology. The Twitter iPad app is more of a proof of concept around whether its NLP processor works well. Verma says that it’s really difficult to keep up with information shared through Twitter and there has to be a better way of surfacing the most important news. Thirst uses a custom natural language processor to pick out the most important stories around different keywords or subjects like ‘gay marriage’ (because of this past week’s big announcement from President Barack Obama in support of it).
The likes of Amazon and Netflix invest millions of dollars on designing the ”perfect” algorithm to predict what you’re likely to want next. Recommendation algorithms are hugely important within the e-commerce space – the history of your actions on the site will inform what the site suggests to you in future.
Action: You probably don’t have millions to invest in developing recommendation engines or fancy formulas that tell your customer what they should do next. What you can do is look a little bit wider than mass personalisation.

Relevance - The Key to Content Discovery by Scott Gray, May 2, 2012

Actually Relevancer provides “recommendation engines” and “fancy formulas” via a white label service for any website and it doesn’t cost millions.

Both Hammond and Frankel insisted that, while Narrative Science will certainly replace some types of human-generated writing, the stories they’re most excited about are the ones journalists rarely cover. Because of readership expectations, no journalist would write a story with relevance to only one person, or a few—sports writers, for instance, don’t write about Little League games in the first place. That’s why the company’s putting special effort into what they call “audience of one” applications—narratives that bring professional-caliber prose insight where right now we only have confusing data.
Machines can learn what memes will go viral, but only humans can aggregate the content effectively to create a truly social Web experience, according to Jonah Peretti, the co-founder of BuzzFeed, a meme-focused aggregator site.
The Web is moving away from just curating content, to actually performing artificial intelligence operations on that content. … we are experiencing a move away from curatorial tools to algorithmic tools. We are, indeed, creating a type of second brain made possible by basic programming skills. A whole group of sites – Screenr, Google Reader, Diigo, delicious, Instapaper, Evernote, Pinterest, and Social Cast – are becoming part of a massive second brain that is able to establish relationships between all the content that we are creating online with different Web services.

Dominic Basulto, Your Life is an Algorithm, Your Brain is an Operating System, February 23, 2012

Though the title is questionable from our perspective we certainly support the statement that algorithms will be used more and more to figure out relationships between all the content an individual person and her network is consuming, thus freeing the brain from filtering and prioritizing to some degree.

Social is about friends, while interest is about your interests and the two may or may not intersect. … I see interest-based networks as different from social networks (friends versus interest-focused activities) and I consider interest more easily monetizable and more susceptible to the emergence of innovative new applications.
Vinod Khosla, The “Unhyped” New Areas in Internet and Mobile, February 19th, 2012
Writing software that more perfectly emulates the pattern-matching “intelligence” of non-human organisms. Trigger and response — post hoc, ergo propter hoc — is a good start. Better still is a database of past triggers and responses, with the software working probabilistically to determine the best response to the current trigger. Better still is real-time interaction between you and your software, first so you can show it which triggers are significant and which are not, and second so you can show it your ideal response to particular significant triggers. That is to say, software that you can train like you trained your dog.

Nothing that is promoted as being Artificially Intelligent is intelligent in any way — nor even aware as the simplest of organisms is aware. But what AI actually is — canned, pre-programmed human intelligence — is very useful already, and it will only come to be more useful as we get a grip on the idea that we should program software for human beings, rather than always trying to reprogram human beings to fit our software.

Greg Swann, Debunking Artificial Intelligence — while programming your computer to be almost as smart as your dog, February 6th, 2012

Great post on exaggerated expectations on Artificial Intelligence and where a limited scope will still drive significant value.

We live in the App Age and are entering new territory.  The sexy math behind voice or facial recognition, real time translation, or even just assembly of a playlist of music, is no longer the realm of super computers or even desktops. Smart phones will run algorithms, and the data to feed them will also be more fluidly available. Forget Global Warming models: Consumers will pay good money for an algorithm that gathers data and solves everyday problems.

TechCrunch misses the point on personalization


TechCrunch published an article yesterday about the challenges of personalization and why no one has been able to innovate beyond what Amazon did 10 years ago. Leena Rao makes a good effort in trying to understand the challenges, mentioning the need for intent-based data, making sense of social, and privacy concerns. All are true. But the framework with which she’s approaching the problem is wrong.

The right way to look at this is by splitting the world of products into two: products that age and products that don’t.

  • Books retain value over time. A book you wanted to read last year is something you’d still consider buying today (hence, the existence of airport bookstores). Same goes for movies, which is why Netflix beat Blockbuster.
  • Fashion items (shoes, clothing, accessories) do not. Softlines (the retail term for fashion items) are extremely seasonal; items go out of style within months and unsold ones end up on the discount rack.

You’ll notice that successful personalization tech is tightly focused around items in the first category. Books, music, video, kitchen appliances, gardening equipment, (to a lesser extent) electronics - all things that Amazon’s recommendation algorithms are good at. (I would know, I was the product manager for that team). That’s because these products have a long enough shelf life to reach a critical mass of purchase data. You need dense datasets to do personalization right.

Where does personalization suck? The second category. To make it even more difficult, items in this category tend to be ones that you can look at and within half a second decide if you like it or not. They are visual, tactile, sensual. They are also highly individual - a watch that I love is also something you might hate, even if we share the same taste in movies. Hell, I might even love one watch but hate another that almost looks exactly the same. People shop in this category by gut feel and emotion, not by attempting to maximize a list of requirements and system specs. The result is a very sparse dataset with items going out of style too fast for the algorithms to become useful. What you end up with is least common denominator recs (like white socks and undershirts) that completely lack joy and delight.

The solution, like Leena points at, is social, although she gets it slightly wrong. I’ll follow up this post with my thoughts on how social can really make personalization work.

Google Panda in Plain English (Infographic) - Single Grain


The Google Panda visual history laid out in a visual infographic poster. The nine Google Panda updates release dates and key traits and a summary reminder of what are the key things that this automatic filtering algorithm is after.