Cognika Corporation Blog
The cognitive computing company
Developing next generation technologies at the intersection of semantics, machine-learning, artificial life, social networking and other technologies.
Sunday, April 4, 2010
Trial XL is now GA
Learn more about it here.
Wednesday, December 2, 2009
Pricing Models
1. http://www.sachinagarwal.com/setting-pricing-for-a-startup-the-rule-of-80
2. http://www.neildavidson.com/dontjustrollthedice.html
Please post any other links you might have in the comments.
Sunday, November 29, 2009
Cognika gets covered in India-New England and the Indus Business Journal
Data Visualization
As the classic definition goes: Complexity is a function of the number of components and their interactions in a system, while Complicatedness is a function of the nature of those interactions . Thus a complex system is not necessarily a complicated system, and vice-versa.
We believe that visualization and presentation play a huge part in reducing both complexity and complicatedness, especially when it comes to interactions with data. While that may seem to be a no-brainer, finding the right visualization mechanism for a situation is particularly difficult.
Here is a fabulous "thought-starter" to thinking through what approach might be most appropriate for a given situation.
Source: http://i.imgur.com/YjWta.jpg
Sunday, October 18, 2009
Data Tsunami
The legendary investor Warren Buffett is reported to have said "...it takes all of 20 years to build a reputation and a few minutes to destroy it". The recent turmoil in the markets and the sudden and astonishing collapse of Wall street "titans" such as Bear Sterns, Lehman Brothers etc. has only underscored this aphorism. As the most clichéd of clichés go "hindsight is always 20/20", however there is an extremely serious lesson to be learned from the experience of the last crisis; that the current financial meltdown did not occur due to a lack of information. In fact, it has been argued that investors, regulators were so overwhelmed with data to make any sense. Many question why enforcement was so sorely lacking; while it is very easy to second-guess and blame the so-called regulators - under-staffed and under-resourced that they are - we believe the crux of the problem was that "sense-making" from this tsunami of data was woefully inadequate.
To participate please email us beta@cognika.com
Tuesday, October 13, 2009
Swimming in a sea of data
At Cognika we have been leveraging Hadoop extensively for many of our machine-learning activities. Lately we have developed some image-processing and feature detection algorithms which normally would require significant investments in hardware. However, w have managed to achieve comparable performance using commodity hardware and squeeze out some impressive results. Please post your experiences and/or lessons learned and we are happy to share ours.
Saturday, July 4, 2009
Entrepreneurial Insights
Having dealt with more than a few such poseurs, we think any readers of this blog should take such "big-talkers" with a giant pinch of salt and some. Any serious entrepreneur would never claim to know "all about" startups, because s/he knows that every enterprise is different from the last, and each time they are starting afresh!
Friday, June 12, 2009
Seeking beta users for Trial Accelerator
Trial Accelerator is a state-of-the-art analytics and modeling Tool for accelerating and optimizing Clinical trial design and operations. We are looking for users from a wide spectrum of customers: from big pharma to small biotechs, from academia to researchers etc.
The object is to incorporate multiple perspectives and serve the HCLS community to improve processes across the board. Your time and interest is very much appreciated by the Cognika team.
To sign-up please contact us here
Trial Accelerator Beta Launch!
This is not just a monetary loss for pharma/biotechs - a typical drug takes 10 years to develop, and costs about a billion dollars - but is a much bigger loss for the millions of people suffering from various conditions who could potentially have benefited from a better designed and managed clinical trial process.
Thursday, May 21, 2009
VC funding or not?
Recently, a trusted adviser (and an entrepreneur with a decent exit) mentioned that VCs often have very divergent interests from those of the average entrepreneur. A typical entrepreneur is happy with a 5x or even a 2x exit from a startup. An institutional investor– who is typically invested in several companies simultaneously - is looking for one or two big hits with little heed for the rest. To use a baseball metaphor, an entrepreneur is OK with a double or single, but a VC is looking for a walk-off home-run, and they want everyone in their portfolio to swing for the fences. To them, any startup in their portfolio is expendable in the hope of getting a large payback overall. They want to bet big and win bigger, if some fail “oh well! Too bad”. However to an entrepreneur, their one-and-only bet is their startup on which s/he has invested their blood, sweat and tears.
We think any entrepreneur should consider this proposition very carefully before considering any institutional investment.
Here is yet another fabulous slidedeck from Eric Ries that expounds further about “lean startups”
"Lean" Startups
The company showcased in the presentation IMVU.com proposes an idea we very much subscribe to: that startups must learn by experimentation. We believe “trial & Error” is the ultimate judge of any business-plan. To that we might add they are “useful”, inasmuch one actually learns from the experience.
As the author points out, with lean startups “the problem is unknown and the solution is unknown”, thence the need for a constantly adaptive approach. We posit that virtually all startups are in this situation, even the ones who have a so called well-defined problem-space and solution. We argue there is no such thing as a "well-defined" problem, much less a "well-defined" solution. Most entrepreneurs have a vague, fuzzy idea of what they want to tackle and look at it through foggy glasses. It is only through iteration that the problem gets into focus. As the classic adage goes “all plans are wrong, but all are useful”. The businesses that succeed are the ones who adapt and iterate constantly.
Eric points out startups must “learn to learn”, and do so constantly.
Thursday, April 9, 2009
Fascinating talk on emergence
Sunday, April 5, 2009
Cognika is featured in Xconomy Mobile Innovation Showcase
Cognika's profile is available here.
The beginning of Web 3.0 : Wolfram Alpha
In this post, we discuss Cognika’s ideas on what the next-generation of web can and should be. We believe that next generation systems should address the four “C” s of information retrieval:
1. Contextual
2. Customizable
3. Composite
4. Complete
Unfortunately, current search technologies are far from meeting such requirements. However, a fascinating effort called Wolfram Alpha (of Mathematica and Mathworld fame) seems to be along the lines of we at Cognika have been working on.
Nova Spivak of Twine describes it thus:
“In a nutshell, Wolfram and his team have built what he calls a "computational knowledge engine" for the Web. OK, so what does that really mean? Basically it means that you can ask it factual questions and it computes answers for you.
It doesn't simply return documents that (might) contain the answers, like Google does, and it isn't just a giant database of knowledge, like the Wikipedia. It doesn't simply parse natural language and then use that to retrieve documents, like Powerset, for example.
Instead, Wolfram Alpha actually computes the answers to a wide range of questions -- like questions that have factual answers such as "What is the location of Timbuktu?" or "How many protons are in a hydrogen atom?," "What was the average rainfall in Boston last year?," "What is the 307th digit of Pi?," or "what would 80/20 vision look like?”
We believe while computing the answers to scientific questions as Wolfram Alpha apparently does is amazing and potentially world-changing. However there are far simpler approaches that make information usable for the average user. At Cognika we are developing technologies towards that end, and you can view the Cognika XI demonstrator preview we have created to illustrate what we mean. There are very many lower hanging fruit, which could lead up to tools like Alpha. As described earlier, we are big believers in the concept of emergent semantics, and are positive that such "knowledge" could arise out of much simpler approaches and interactions. Alpha apparently does this by hiring a horde of scientists and researchers to codify the scientific knowledge, but to do this on a web-scale is a different matter altogether. Here the idea of emergent semantics is likely to help develop knowledge on a web-scale.
Having said that, we are very eagerly looking forward to Alpha's due to be release later this year. This promises to be a game-changing tool and move the web to a completely new paradigm of information retrieval. Exciting times lie ahead!
Cognika XI
We have pioneered the paradigm of Zoomable User Interfaces or ZUIs. ZUIs make information-glut usable as “knowledge” in a simple, intuitive fashion. One of the simplest and most intuitive User Interface paradigms in existence is that of a map by zooming in/out for details/abstract. Cognika XI™ offers the user a mechanism to iteratively access and develops a personalized compendium of the information desired. This customizable compendium – which we term a “View” – spans multiple domains, documents and media and self-assembles information contextually, with which the user can iterate.
We illustrate this using an example of searching for a movie, as with most search engines the results are rendered with an initial summary, however the user can then click on a result of interest to zoom-in further to view show-times (based on location) and reviews (from selected sources). On zooming-in further, traffic or public transit information relevant to the closest theater is displayed. The user might then choose to include other facets such as tickets, nearby dining etc. Cognika XI™ also allows for inline expansion of concepts of interest, for example the user might choose to zoom-in further to learn about movies in similar genre by the director, or expand on the plot summary to learn about the context and story. Thus we believe, offers a novel, intuitive mechanism for Human Computer Interaction, especially in information-rich environments.
We shall be launching XI into beta shortly and will also be showcasing this at a few major conferences. Stay tuned for announcements. Here is a sneak preview of the technology:
Tuesday, February 17, 2009
What You Show Is What They Get!
Having been through quite a few sales sessions, we believe that this is likely since technologists are accustomed to interacting in environments with people with similar skills, interests and backgrounds. Selling to customers with microsecond attention spans is worlds apart. The truth of this does not hit like a bucket of cold-water unless one sees much of the audience drifting off to sleep during a sales-pitch.
As the apocryphal saying goes, “perception is reality” and that could not be truer for a startup. Rare is the customer who is interested in the “how” something works, as opposed to “what is in it for me”. A “demo” that does absolutely nothing under the covers, but tells a very good story is a far superior sale than a superb technology that no one understands (read cares about).
While, much of this post might seem rehashed wisdom from other sources, turns out commonsense is very uncommon. An associate recently asked the question “will your mother understand what you do? And if she does…”
We disagree; family, friends etc. are victims of the infamous Hawthorne effect, and anything but proxies for a “customer”. Selling to your “mother” is about as far removed from selling in the real—world, as it is to a fellow researcher.
We have therefore coined a truism for entrepreneurs to keep in mind: What You Show Is What They Get! (WYSIWTG), or its cynical variant "What They See Is What They Get" (WTSIWTG).
Thursday, February 5, 2009
Entrepreneurial Wisdom
Our favorites:
"No means maybe, Yes means maybe":
Some of our most enthusiastic supporters (customers, advisors etc.) today, gave us the cold-shoulder early on. We kept pushing to see what could be done better. An entrepreneur's job is never to take no for an answer. Conversely, a "yes" only means anything until the next customer check, and one should never, ever take anything for granted.
"The best way to get investment is not to need it":
Many startups make the mistake of chasing investors early on hoping to dazzle VCs and angels with the sheer brilliance of their idea and team. Bad news - the odds of that happening are miniscule. As most entrepreneurs would know, focusing on closing sales and getting revenue is the best VC pitch a startup can make. The slide-decks, financials, business-models, team strengths etc. become side-shows, once a startup demonstrates a clear market-need and revenue.
The joke goes:
Q: what is the difference between God and a VC?
A: God doesn't think he is a VC!
As a joke that is funny, but is not intended to be a put-down on VCs. Please remember most VCs are pitched by a huge number of companies, and have to sift through a lot of "noise" to find the best bets. Some entrepreneurs get bitter and disillusioned at being spurned by VCs abruptly.
We think startups best serve themselves by spending all their energy on closing sales and keeping customers happy, and let the investors come to them. Not needing investment is the best pitch an entrepreneur can ever make! That is a pearl of wisdom every entrepreneur should hang onto.
Monday, February 2, 2009
Cognika's new office space!
If you are in the neighborhood, and are interested in learning more about Cognika, please drop by to say hello. You can see where we are located here, here and here.
Sunday, February 1, 2009
"Intelligent" Design
In a previous post we discussed the potential of biomimetics or mimetics in general. What is needed is the ability to connect scientific challenges with potential solutions from learning in nature and a variety of other domains.
Despite valiant efforts by some academicians and some in industry, such a project needs a lot more resources to reach critical-mass. One wishes there was a full-fledged collaborative effort from government, academia and big pharma and other businesses to push for a shared knowledge corpus (a la Wikipedia), but published in a machine-understandable or machine-query-able format that could be leveraged for such cross-domain breakthroughs.
The Semantic Web concept is a a great step in that direction, but we hope baby-steps could be taken by people and organizations just publishing more metadata and better descriptors of their data, on which approaches like the above can be built upon. It certainly need not be a single repository like the Wikipedia, but a just a collection of sources much like the current WWW.
Saturday, January 24, 2009
Recruiting at Startups (follow-up)
Not one of the responses were anything even close to being acceptable. When the obvious was pointed out in no uncertain terms, the applicant threw a hissy-fit: the details are unimportant but needless to say the applicant knew "game over!" .
This example brought up a number of questions in how we evaluate candidates, so we added the following points to evaluate applicants:
2. Prima Donnas Unwelcome: A startup is a very, very rough and tumble game. if one cannot handle being tackled and tackled hard, they are best advised staying away from the field. A fellow entrepreneur pointed out being in a startup is like being a quarterback without a defensive formation, and that is not far from the truth. Any employee with a chip on their shoulder is likely unsuited to work in entrepreneurial organizations.
Working for a startup requires more often than not, the ability to take some hard hits and remain standing. If that doesn't sit well with some.....oh well! Sorry!
Wednesday, January 21, 2009
Is consciousness an emergent phenomenon?
Tuesday, January 20, 2009
Is Search the Future? (Part-1)
A hotly debated topic is the so-called “future of search” or "the next Google". Opinions have ranged from Google’s Marissa Mayer's “search is 90% solved” to Bill Gates’ opinion that search has barely scratched the surface. However, search is increasingly more of a problem than a solution.
Undoubtedly, companies like Google, Yahoo etc. efforts in search have made information accessible to anyone with an internet connection - which is a very laudable achievement - but it has made the challenge of information overload worse than anyone could have imagined. Search, has made information available, but it has not yet begun to make it usable, and there is a night/day difference between information availability and usability. To make information usable, it needs to be coalesced from multiple-sources in a context-specific manner and presented intelligently to achieve a task. (look for specifics in later posts)
Search today is - pardon the expression - a** backwards. Imagine a child asking a parent questions that typically puzzle the infant mind, about “why it rains”, “why birds fly” or “why are leaves green” etc. and now picture the parent handing out a collection of encyclopedias, books, papers, articles etc. in response, and saying “go figure it out”. In many ways, the current paradigm is just as inane. It is left up to the user to make sense of the information from across multiple-sources (and most information does indeed span multiple sources).
A school of thought believes that the holy grail of search is a crisp, succinct response to queries. While this approach might work for questions such as “What is the capital of Peru?” or “Who is the current president of Ukraine?”, this approach would fail for the vast majority of queries – the ones for which there is no ready-made response. Since this involves aggregating information from multiple-sources, reasoning therewith, providing a response that is relevant and contextual, and yet provide a user-definable level of detail.
For example, could such a system answer a subjective question such as “What are the greatest movies of all time?” A response would typically involve user iterations: as ranked by AFI or IMDB or MPAA? Hollywood or International? Specific genre? And so on…
Some might argue the faceted browsing approach would work since a result set could be filtered by selecting/deselecting facets for the above etc. However, that is not quite the end of it. While the content might include Movie Title, Synopsis, Cast and some associated information. This does not allow inclusion of user-definable data points: such as reviews from select critics, availability in local stores, movies with similar plot lines, other movies by the same director, or other movies in the same genre etc.
This information would usually span multiple data-sources, documents, domains and would require repeated searches and manual collation. Now extrapolate this to highly complex domains (such as scientific research, business analysis etc.) and the challenge is multiplied manifold.
At Cognika we believe that a new paradigm needs to be developed for information processing and that search as-is today is a stepping-stone towards this vision. This quote best summarizes Cognika’s vision for the future: “Now this is not the end. It is not even the beginning of the end. But it is, perhaps, the end of the beginning.” – Winston Churchill.
Friday, January 16, 2009
"Biomimetics"? - why not just mimetics?
An example is the idea of TRIZ developed by a scientist in the former Soviet Union Genrich Altschuller. Altschuller postulated that solutions that exist in one domain (example: automobiles) could help solve a challenge in a very different domain (example: medicine). He found such cross-domain solutions using the patent corpus as a source. So biomimetics seems to be a specialized case of TRIZ that uses biological organisms an information source.
AskNature is a website that seems to be a great start at collecting such data and making biomimetics usable. It would be a good next step to release the data into public domain or expose as an API, so it can be integrated into other applications.
Overall we feel that the concept of mimetics should be extended across corpora - biological, patents, journals etc. Only then will the true power of this approach be realized.
Coding errors get 'dangerous'
The BBC also provides some coverage here.
Thursday, January 15, 2009
Emergence in Computer security
Caveat: we are not suggesting uninstalling your current protection tools.
At the core of this, lies the fact that most security software is very reactive i.e. as soon as a new threat is detected, antivirus develop an update against it. By the time the update actually is installed on a user computer – assuming the protection actually works – is often too late. The current deluge of malware calls for a paradigm shift in moving to a proactive, predictive approach.
At Cognika, we are exploring the possibility of applying the concept of emergence to computer security. Artificial Immune Systems (AIS) have been researched for a few decades however such approaches remain researchy. The concept of AIS borrows from the idea of biological immune systems, which constantly evolve to cope with newer and emerging threats.
Biological immune systems are remarkably effective against the countless pathogens that organisms are exposed to during a lifetime. Millennia of evolution have made biological immune systems remarkably effective. Our idea is to adopt some of those ideas to develop an internet immune system akin to bio immune systems. Such research would be at the intersection of evolutionary computing, machine learning, computer security and other fields.
Here are some thoughts along these lines:
- An evolutionary, adaptive system could be developed to cope with the current onslaught of malware: an immune system that is constantly mutating (much like its adversaries) and adapting to newer threats.
- Cohorts of machines form trust networks (Immunets) to warn each other of imminent threats and share mechanisms to thwart their attack. The Immunets could be layered around existing social networks or organizational networks.
- The Immunets employ a combination of evolutionary algorithms and machine learning to develop effective defenses, which do not require manual updates and examination.
- Such systems could be augmented by current security-ware, much as biological immune systems are augmented by medications and treatments.
Monday, January 5, 2009
Cognika is hiring!
More here
Sunday, January 4, 2009
Recruiting at Startups
As they say, Aptitude = (Right) Attitude + P(erseverance) - T(antrums)
There is a great post by Joel Spolsky. The bottom-line is when you are applying for a position in a startup or small company , it is doubtful anyone is interested in your skills in managing 100s of people, or CRM/SAP skills etc. You should focus on your multiplexing skills: your ability to churn out code, while communicating with customers, and yet playing a good sales/marketing person, while offering technical support. Most startups and small-businesses need exactly that kind of "all-rounders".
p.s. Anyone applying to positions at Cognika ought to pay significant attention to these factors.
"Semantic Web" vs "Emergent Semantics"
However, we think while Ontologies and higher-level logical constructs have their place, the vast majority of semantics will be emergent through a combination of machine-learning and user interactions. An exemplar is how the web itself evolved through users linking to pages they thought was relevant etc. - so we think our approach is in the original spirit of the WWW.
Any comments?
Cognika gets some press
Saturday, September 20, 2008
Welcome!
We intended to develop this blog to the advance the idea of "emergent semantics".
Emergent semantics is a mechanism through which meaning arises out of interactions between software agents, humans and metadata. The so-called “folksonomies” Flickr, Del.icio.us are simple examples of such emergent phenomena. The idea is to extend this to the entire web and allow data & metadata to interact in ways currently not possible. For example, we can envisage ways in which IR algorithms can generate relations and allow for serendipitous insights.
While this blog will broadly revolve around the above theme, we reserve the right to post anything we please :)
welcome again!
Labels
- algorithms
- article
- articles
- beta
- biomimetics
- biotechs
- business
- clinical trials
- cognika
- collaboration
- consciousness
- demo
- economy
- emergence
- entrepreneurship
- errors
- evolutionary computing
- financial
- future
- growth
- hadoop
- healthcare
- hiring
- innovation
- insights
- jobs
- mobile
- news
- office space
- pharma
- press
- pricing
- processing
- products
- recruiting
- research
- search
- security
- semantic web
- semantics
- showcase
- site selection
- small-business
- startups
- technology
- threats
- trial
- trial design
- trialx
- trialxl
- vision
- visualization data presentation ideas complexity complicatedness
- web 3.0
- welcome cognika
- xconomy
- zui