Real progress in artifical intelligence


Finally, however, in the last decade ­Hinton and other researchers made some fundamental conceptual breakthroughs. In 2006, Hinton developed a more efficient way to teach individual layers of neurons. The first layer learns primitive features, like an edge in an image or the tiniest unit of speech sound. It does this by finding combinations of digitized pixels or sound waves that occur more often than they should by chance. Once that layer accurately recognizes those features, they’re fed to the next layer, which trains itself to recognize more complex features, like a corner or a combination of speech sounds. The process is repeated in successive layers until the system can reliably recognize phonemes or objects.

via New Techniques from Google and Ray Kurzweil Are Taking Artificial Intelligence to Another Level | MIT Technology Review.

A primer on probabilistic computing

Inference, particularly with large data sets, and disparate solution criteria, is one of the tougher challenges of current computing models.  Probabilistic computing may unlock an alternative approach to tackling complex problems by enabling systems to infer solutions that lie outside the current linear computational models:

Probabilistic programming languages are in the spotlight. This is due to the announcement of a new DARPA program to support their fundamental research. But what is probabilistic programming? What can we expect from this research? Will this effort pay off? How long will it take?

A probabilistic programming language is a high-level language that makes it easy for a developer to define probability models and then “solve” these models automatically.

via What is probabilistic programming? – O’Reilly Radar.

Bonus: The video at the bottom of the linked blog post serves as an excellent overview of where this technology is headed

Phil Windley on understanding the architecture of Personal Clouds

Just under the radar, there’s been a lot of activity in the ProjectVRM space of late.  Various clusters of work are underway in the VRM space, including identity research and personal data store development.  On the latter, Phil Windley has an excellent post explaining the framework in which personal clouds should operate by referencing the tried, trusted and true technologies around the IMAP protocol:

In short, email was designed with the architecture of the Internet in mind. Email is decentralized and protocol-mediated. Email is open—not necessarily open-source—but open in that anyone can build clients and servers that speak IMAP and SMTP. As a result, email maximizes freedom and control for the user and minimizes the chance of disruption. The features and benefits that email provides are exactly the same as those we want for personal clouds. Designed right, any application built on a personal cloud would provide similar functionality.

Web 2.0 has given us a model that is exactly the opposite of email. The model encourages user data to be stored in separate silos. You cannot easily migrate from one service provider to another. And when a service provider goes away, you are abandoned and marooned. You are not in control. Of course, it doesn’t help that this is all in the service provider’s best interest. They make money from the fact that the predominant model for building online applications leaves their users powerless.

via IMAP as the Proto Personal Cloud.

There’s lots of activity underway in this space.  I’ll have my own thoughts in several subsequent posts.

Applying a data lens to India

Conducting a census of India is a monumental task.  The last such undertaking happened in 2011 [wikipedia].  While the raw data reveal, well, raw statistics, delving deeper into census data is a fascinating exercise.  On that note, I recently wandered upon a new weblog that is devoted to extracting insights from India’s last census.  The author, an anonymous reporter based in Delhi, has pulled some fascinating revelations.  One recent post looks at an approach to identify the clusters of wealth by districts across India:

We could start with the fact that only around 42,800 people in the country admit to an income of Rs 10 million or more to the income tax department. But almost everyone, the finance minister included, thinks that figure is laughably low. Here I want to talk about a more er…inclusive definition of the privileged.

Take a look at the map below. It maps the proportion of households in each district, who told census-takers that they own all of the following – a TV set, a phone, a computer and a vehicle (scooter/motorcycle or car). That number, for the country as a whole, is 4.6% (roughly 11 million households).

I leave you to draw your own conclusions about what it means to be ‘privileged’ in this country. I also leave you with this question: If the census takers had asked each one of these households, what ‘class’ of society they thought they belonged to, or where they fit in within the income distribution, what do you think their response would have been ( and by ‘their’, I also mean ‘our’)?

Post: We are the 5%

In another set of posts, the author dives into older census data to reveal that per capita income divergence between India and the west (especially the US) peaked at an unexpected time, the close of the ’70s:

In 1979, the difference between American and Indian per capita incomes peaked and India began a period of catch-up with not just the US, but with the West in general which continues today.  But that year also set in motion another divergence – between India and China which also continues.

Posts: 1979: When a Historic Shift went unnoticed, This is not about China or India

Each post on Data Stories provides another useful lens to apply on India’s census data.  I know I’ll be following Data Stories to see what else the census data reveal.

An excellent post on big data and the customer experience…

An excellent post on big data and the customer experience over at the Harvard Business Review blog. Of note:

Expand the Value You Create for Customers

Improving the customer experience is a fine idea. But companies often take it to extremes. It’s always a good idea to look for new ways to create value for customers. But focusing only on doing so through your product or service is entirely one-dimensional. The hard reality is that your product or service, however great it is — however much it helps your customers get a job done or provide an enjoyable experience — is likely just not that important to their lives in the grand scheme of things.

via The Big Goal Behind All that Customer Data – Bill Lee – Harvard Business Review.

Task manager and collaboration tool Trello turns a year old today

Having tried nearly every to-list, task manager over the years, I think I’ve finally found one that works for me with Trello.  I have stuck to it as my go-to application for managing a wide array of both personal activities and collaboration across groups.  Many of the current generation of team collaboration/task management tools provide great flexibility, but the user experience curve is still too high to quickly bring a disparate group of tasks and people together.  This is where Trello shines; it is a deceivingly simple application that provides significant horsepower behind the scenes.  Trello uses a skeumorphic approach to managing activities, relying on a time-tested approach of ‘boards’ that contain columns of movable ‘cards’.  I’ve seen people online refer to this approach as being similar to the Japanese Kanban process used in manufacturing, which I suppose is the inspiration for the application.  I can’t really do justice to how the application works here, so I suggest you visit their home page and take a tour.  

The main browser-based application (there are iPhone and Android companion apps) displays some of the best web coding I’ve seen.  Fog Creek Software, the developer of Trello, is providing enterprise-class horsepower with a consumer level user experience, which is not an easy feat. 

Earlier this year, Joel Spolsky, CEO of Fog Creek Software, wrote that Trello was designed to be used by a wide variety of people:

The biggest difference you’ll notice … is that Trello is a totally horizontal product.

Horizontal means that it can be used by people from all walks of life. Word processors and web browsers are horizontal. The software your dentist uses to torture you with drills is vertical.

Vertical software is much easier to pull off and make money with, and it’s a good choice for your first startup. Here are two key reasons:

  • It’s easier to find customers. If you make dentist software, you know which conventions to go to and which magazines to advertise in. All you have to do is find dentists.
  • The margins are better. Your users are professionals at work and it makes sense for them to give you money if you can solve their problems.

Making a major horizontal product that’s useful in any walk of life is almost impossible to pull off. You can’t charge very much, because you’re competing with other horizontal products that can amortize their development costs across a huge number of users. It’s high risk, high reward: not suitable for a young bootstrapped startup, but not a bad idea for a second or third product from a mature and stable company like Fog Creek.

via Joel on Software

Fog Creek is aggressively developing the application, and has recently updated the Trello iPhone companion app (I can’t wait to see a native iPad app!).  So, a year in with a horizontal product, Trello marks the milestone with a great statistic: 

You’ve made 717,337 accounts. We hit 500,000 in July, so it’s going even faster these days.

via It’s Trello’s Cake Day!

Congratulations to the Trello team for a successful year!

One of these days I may post about my own workflows using Trello, but in the meantime I encourage you check out the application for yourself.  

On spreadsheets, big data, and GoodData’s Bashes

During a recent conversation I had with Mark Angel [founder of Knova Software and most recently the CTO at Kana], he was quick to point out that the ‘spreadsheet’ stage of cloud computing had yet to arrive. His point was that most of the computational horsepower of the cloud was still largely relegated to the technical elite inside organizations, and end-users had limited options on how cloud data were interpreted. Data, therefore, are frequently interpreted out of context, and far removed from the impacted business process. Nearly a generation ago, spreadsheets altered the corporate landscape by empowering end-users to manipulate data based on their expertise, unleashing an entirely new way of extracting meaningful insights from data. To Mark’s point, for many enterprises, that stage of cloud computing has yet to arrive.

The best opportunity for this ‘spreadsheet’ stage to take hold is in the white-hot field of big data. While capturing and storing data has never been cheaper, the opportunity to extend the ability to interpret this data to vast armies of knowledge workers has been limited. It was in that context that I found this morning’s announcement by GoodData of their Bashes to be an interesting development:

We call our apps “Bashes” — for business mash-ups — because they combine the best elements of consumer apps with modern, enterprise-class technologies. That means consumer apps’ clean and intuitive user interface, ease of use and device independence, with cloud-based business technologies that collect and manage structured and unstructured data from hundreds of sources. With Bashes, businesses can discern meaning from all the data flooding in from emails, social media, enterprise software and cloud apps.

Filling in Big Data’s Missing Link: Making Big Data Pay for Itself @romanstanek

Clearly there’s an opportunity to give today’s knowledge worker a spreadsheet-like environment to mash-up disparate data sets on the fly. It looks like GoodData’s positioning their platform, and Bashes, to be one of the spreadsheets for this generation’s knowledge worker.

David Kay on knowledge sharing challenges

David Kay is one of the best knowledge management consultants (if not best consultants) I’ve known.  As anyone who knows him will attest, David’s one of the few consultants that ‘gets it’.  Real change in organizations, particularly with knowledge management, is not as much about technology implementation as it is about process transformation.  And, process transformation rarely happens unless there is an organizational culture that is amenable to change.

Over at his blog, David has an excellent post about warning signs he’s seen in corporate cultures over the last decade of consulting.  While he’s focused on the knowledge sharing process, the post could apply to just about any corporate transformation effort.  Representing the technology vendor’s point of view in many instances, I found myself nodding with each of his eleven points, especially point number eleven:


11. A lousy work environment, food service, and coffee.  Look, we’re not all going to work in the Googleplex with free gourmet lunches and company-branded ice cream treats.  But we spend a lot of time at work, and our mental state there matters, and our heads are influenced by our environment.  (There’s a reason they spent so much time building cathedrals in the Middle Ages.) If I go to an office building that’s dingy, dreary, sterile, and cut off from natural sunlight, I know something.  If the coffee service comes out of 1950s-style glass carafes and hotplates with generic pre-ground beans in foil packets, I know something.  If people resign themselves to the depressing burger-and-fries or meatloaf options at the cafeteria, I know something.  And if the company hasn’t spent the money for decent computers, double monitors, comfortable ergonomic chairs, and IT that works, I really know something.  I know the company doesn’t really care about the employees, no matter what they say, and it’s going to be wickedly hard to get the team excited about taking on a new challenge.

via  DB Kay & Associates 

The cafeteria comment made me laugh as I was reminded of a large company cafeteria I visited where there was a five cent up-charge to use plastic utensils (for EACH utensil).  You can only imagine what the morale level was at that now defunct large company.

Check out David’s post, I’m certain it’ll remind you of places you’ve seen as well.