Deep learning the brain of an organization

The world is full of beautiful ideas, and people making something with them. While bouncing between pragmatism and idealism, playing with MNIST and CIFAR datasets, figuring out entropy and trying to understand it with information theory, I came across such a marvelous representations of information in words and in visuals that made me wonder in awe at the gift and power of thought put into action; electric signals transforming into neons of imagination.

So how can we create the brain of an organization using deep learning and artificial intelligence. It is not, too big of an idea to get one's neurons firing in all directions as future of computation points to "quantum" leaps in processing and we can let our suppositions soar.

The decisions in an organization are dependent on almost infinite factors. Just the decision makers psychological propensity to make a particular decision seems like something difficult to measure quantitatively, let alone trying to gauge the impact of various departmental interactions, political and vested interests; a summation of all motives. This becomes a question about intelligence and if the AI actually can point to intelligence itself, discerning between desirable vs undesirable attributes in an organization when a medley of attributes result in desirable outcomes lest we assume that the accuracy of the decision is determined by its success or failure alone.  The problem then is determining what human signals can be characterized as positive or negative influences and what initial fixed proportions can be given to each. This may be done without giving a fixed weight to each human feature, but in order to do so, many samples will be required for the same decision with the data on the presence or absence of features effecting the sample decisions.

Human interactions are complex and the emotions and rationale that goes into human buying decisions can be multifarious, yet from a organizations perspective the decisions that affect a sale can be few ( for the sake of simplicity we will only think of a decision at the moment of transaction being based on current circumstances without considering the effect of past experiences with the firm). We can assume that creating the perception for its product or services as the most logical as well as the most pleasing choice is thus the goal of an organization. The variables to entice the buyer to associate the store as the most plausible (I will not say effective here, lets just assume that plausible includes some part effectiveness along with any other complex mechanics happening in buying decision) fulfillment of his need (or desire) for the product or equipment or service can be infinite but the variables that the company effectively controls and manipulates are not that many and can be pretty much based on standard principles of the functions that effect the process.

For designing a brain for organization we need functions of each department that are determined by signals from roles in each of those departments. The results of decisions of each department become inputs to a layer that can then be trained for a specific goal for that stage. The final output of an organization decision, will flow from the first high level input nodes to the deep neural network of functions, with each layer providing their decisions in cascading layers that ultimately yield the output.

Soon almost all qualitative and quantitative components of organizational processes and their outcomes will be recorded, some already are being stored, including human behavioral aspects. That along with increased computing power, supervised and eventually unsupervised learning will result in creating the essence of the organization. I will not be surprised if investors in future ask for such analysis along with lifetime value of customers to determine the future prospects of the firm.

TRA to TAM, Which Technology Will I choose Part 1

The advent of new technologies and the sprouting torrent of startups along with the companies that are dominating everyone's leap of imagination about the promise of a great connected future where human mind will be just one piece of the intelligence puzzle supplemented by machines,  is causing quite a stir in the everyday media. As I take stock of my thoughts, I cannot help but venture into debate with myself about which technologies will become applications that will define a new era. The thoughts led to a desire to find out how we adopt new technologies and what leads people to fall in love with some "tech" while others either wither away in shelves of thrift stores or will eventually be seen in the nostalgic corridors of companies, as also ran the race stimulating conception of an idea, for their budding employees intellectual curiosity to feed upon.

In my search for how to evaluate technology I was motivated towards the theory of reasoned action which pushed me to technology acceptance model and then to the famous diffusion of innovation curve by Everett_Rogers.  If we look at the TAM in its simplest form and its main blocks, I think the media has done a good job of creating perceived usefulness for machine learning and AI. Some good PR along with applications like the self driving cars, the face recognition locks, the voice activated home assistants and self analyzing spreadsheets have prepared the users well for the perceived usefulness of AI. Such is the hype that the perceived usefulness I would say is now anticipatory usefulness for Augmented Reality and Virtual Reality. Where AR/VR does have a hurdle to cross is the perceived ease of use, another important pillar of the TAM. The attitude towards using the technology and the behavioral intention vary from generation to generation. But I don't think anyone needs to prove that the generation that was born with iPhone and Facebook will be more adaptive to always on, non private world, knowing that sharing some information is volitional but life of a digital native is very transparent to data native companies.

Point in making is that except for the perceived ease of use the rest of the pillars of TAM have actually shifted toward faster technology adaption. Along with perceived ease of use there are other variables to consider that play a critical role in forming users perception. In their paper, characteristics of innovation adaption Tornatzky and Klein point out that users differ in their perceptions as they evaluate the benefits through their particular cognitive frame works resulting in subjective conclusions which the individuals believe as their truth. In similar vein Festinger in cognitive dissonance theory argues that the users are predisposed to vote for the technology that they have already adopted therefore after the adoption, look back studies of winning technologies cannot predict the future adoption of another technology. That makes for a very interesting observation, one that I am guilty of as well and that is even before the application of the technology is really available in an application we become vocal advocates of the technology and hence its application becomes unavoidable. This is the trick that successfully works for tech giants with enough media muscle to shift perception. This doesn't work all the time, the case in point will be Google glasses. But will that be the fate of Occulus Rift, Spectacles, Microsoft VR, Google day dream. All signals point to a rapidly rising adaption cycle for many of the new applications on the horizon.

Since I have (not even) barely touched on some of the studies around technology diffusion, its adoption and human psychology, I am but a slave to my desire to explore these further and bound to notate a continuation.

A simple survey for robot adaption

The buzz on the AI is deafening, one need only pick a current issue of any news paper and he will be rewarded with news about some machine with learning abilities. With all the buzz around artificial intelligence it would seem that the world is ready to embrace a walking talking humanoid, with all the goodness of great service minus the emotional baggage to direct their responses, a true stoic, immune to the pangs of hidden spears in human conversation and free from burdens of emotional accommodation. One would think, but generalizing a specialized opinion to the affirmation of masses sometimes have unintended postulations, incorrect most of the time.
To get a gentle feel of how the masses think about it, I set out to perform a small (due to the out of pocket minimums I apply to such endeavors) and simple because I wanted to get a sweeping, yes and no opinion, which we are not going to generalize, but we can take a look and muse in it. If you haven't used, I will recommend trying out Google consumer surveys they are easy to setup and one can do some nice opinion gathering.

Although autonomous cars, voice activated/controlled devices and speech recognition on every smart device is a flag bearer of AI advances, I feel the true test of AI will be the robots.
The first question I asked was simply a perception of people on Robot adaption. Robots as servers is a bit easy I wanted to go bit more bullish on the AI side so I chose a more demanding profession for the robots in my survey, retail sales associate. The first question was

• Autonomous Robots will replace human sales associates in:

Asimo may be a stretch of imagination but I still expected a bit more optimism on the prospects. I also wanted to see if people were mentally prepared for dealing with a robot if they ever encountered one in sales situation hence the question

• I am(will be) comfortable asking questions to a robot in store

Interesting enough very few people were concerned about the look of the robot, my guess is that most have not seen the headless bipedal ostrich. I would think that the best way to adaption will be through familiar robots, especially for children, any one thinking of Anpanman.

My next thought was to figure out the "tech savviness" of my audience so I threw in few question about their online shopping behavior as well as their adaption of voice devices and mobile shopping.

This was a more expected response, the more people bought online (a proxy for their adaption of non human buying behavior) the more receptive they were of robots. Maybe we will see some robots in Amazon go stores.

• Overall what % of your shopping is online (excluding Groceries)?

The same effect was observed for the group of high online buying adapters being more receptive to non human interaction.

• Overall what % of your shopping is online (excluding Groceries)?

There are couple of other ways to look at the picture using the demographic filters of gender and age, or device adaption behavior of using Mobile for shopping or voice activated home devices.  One thing that I found interesting was; of the people in age group 18 to 44 that shopped 75% or more online, not one individual said they will try to find human irrespective of their gender. If I have to guess, I think robots will pop soon, but first as more static answering machines in retail and then to more interactive assistants finally to autonomous helpers. Although I won't mind walking into a store, activate a robot, that walks along my side through the isles, answers my questions, looks up inventory, gives me suggestions, brings up deals etc, I would prefer C3PO as good company but will settle for R2D2 or BB8.

Age of connection will change information flow

The search engines might become irrelevant in the world of connected devices where the devices will talk to each other and the information on each device will be validated by the use of other device owners and by devices themselves through a mix of AI and protocols. There should be a separate protocol for devices to talk (not connect) in terms of human language and the AI that translates and rates it should emerge sooner or later. That protocol with the power of NLP and AI will basically put the power of information truly in the hands of crowds, as the compilation and ranking by few sources will become unnecessary. The race for controlling the device and information echo system will become immaterial if the AI becomes inherent in the devices, regulated by protocol and universally accepted signals for supervised learning (not owned by corporations and big information aggregators) that are neutral and publicly validated; a block chain of AI and information.

Age of information overflow started with the rise of internet, and now it has morphed into age of connections. Connecting individuals to the right information, in right context at point of need almost as fast as the thought comes to them about the need, so that it can be fulfilled. Maybe not at the speed of thought but at the speed of their connectivity.

The power of connections is not in the tribes, When one looks for information, it is almost always an individual providing the information.and more experts jump in to create alternate versions and the crowd participates to tune the information by their comments and ultimately declare a winner through vote or simple affirmations. You may call that crowd a tribe, but the same individual will be member of different tribes, where each one can be the leader of different tribe yet member of many.

I use stack overflow quiet a bit, do I know a single expert on it, maybe, but the chances are the experts I remember are far fewer than the information I have acquired through. Even after knowing a few I don't know the extent of the expertise of the contributor, but considered the nod of the crowd worth validating by using the information.

The power of platform as voice of crowd is bigger validation of the expert opinion then the opinion donor. In few cases the expert is able to transcend the crowd by sheer cult of personality in which case the combination of what and who becomes the knowledge and not just the the piece of information. Who imparts the knowledge becomes irrelevant in which forms of information? Is going to MIT more meaningful than taking all MIT courses online? The educational platforms on internet vary in their size of offering and in their value. The same knowledge disseminated by similar means has different value. But eventually that gap will subside when learning will become the true objective and verification of capability will be more quantitative than branded.

The original question how the ease of information dissemination through internet to the seeker is shaping up the future of experts. The seeker is an important part of the puzzle, because I am interested in a knowledge that needs to be disseminated. Is there a difference in how the knowledge is shared by the type of knowledge shared and the audience of that knowledge; professionals seeking professional advice, normal people seeking professional advice. How knowledge is shared through out history and how the hubs of knowledge are formed.

The libraries were the research vehicle for the past. You could walk into one and do your research on existing knowledge base. Now you can perform the same function, many times faster and instead of knowledge being the hub you are the hub. You can research that knowledge, contribute to it with your comments, become a sharing or explaining resource and that very static piece of information all of a sudden is very dynamic. The finding of the knowledge, giving it your voice and then ability to publish that instantly, that is the modern dilemma. It loads the system with redundant information but it also amplifies the information. You can be right or wrong in your editions but that will eventually get through the digestive system of the internet, munched on by the crowd if at all, spewed to the organic pile of information that gets buried in 100 page down in search results and no one will know.

Granularity and segmentation of information by the user and by the type of information that's, internet. The jazz about personalized content, and targeting is nothing more than connecting the right audience, with the precise information, in the least amount of time. To make that happen the original portal, be it web, be it smart device (imagine all smart devices, cars, watches, phones, homes etc)  where the quest for a piece of knowledge begins matters. Because the interpretation of the intent of person searching the information is the first step to even attempting to get them the right information right away. Words only tell half the story and rest of the context must come from everything else, sensors and artificial intelligence.

Relevant information was always the quest, and although the content creation is reaching monumental proportions the crowd validation along with machine learning is also breaking new grounds. In current digital realm, few big players dominate the information creation because crowds flock to them but eventually the system(any connected device) will be able to find the most relevant information from any available digital source like a certain recipe stored on a connected personal oven of an unknown chef.