I wrote several articles on this a few years back, a side of life I don’t share here much. A few ideas:
When business wants to use AI to regulate and observe, where do they think data comes from to train it?
The entire shipping history of FEDEX from their very start could be stored for the equivalent of a very nice dinner in downtown San Francisco.
If every single event - every transaction, every mail, every phone call, proposal, design, bill of materials, repair, support call, hire, training, spreadsheet, PowerPoint - all of it - a company generates were stored in a blockchain, it would make the future staggeringly easier to manage as tools evolve.
You cannot create data from the past, lost, gone forever.
Autoregressive models will be built on such data. Then GAN tools will be used to evolve businesses very rapidly into highly competitive machines.
The mathematics of statistical thermodynamics to manage businesses is already fully developed. It just has insufficient sampling.
I worry that this oversells the value of the raw data compared to enabling interventions. Yes, obviously it should be tracked but aside from things you just need to have records of the real value isn't in mere correlation but in eliciting causal information from that data.
I really like the idea of software that helps enable, encourage and record this kind of information at a company wide level. I mean maybe you can track hires and how new hires correlate with productivity in a group but can your software help implement a 6 month delay on new hires for a random 5% of groups and analyze the effect?
It could, only if the business understood the effect, and of course it depends on the category of the hires. For example it would be basically impossible to understand the direct effect on revenue generation in hiring new sales reps because their business is subjective for the most part. Think of batting averages over six months for rookies who haven't faced any MLB pitchers. These averages will not be consistent. Also the nature of baseball means a handful of hires is about all you get.
On the other hand in manufacturing, if you have a fleet of machines in a factory with maintenance issues, there is a very direct correlation between their uptime and factory output. Hiring certified mechanics to fix broken makes a tractable difference.
But Data Science becomes more useful when there is a massive amount of consistently tracked data at root. If I can track every sort of data related to an organization going back decades, i'm in an ideal position. We are at a point at which, for most organizations, the capacity exists. This capacity didn't exist 20 years ago. So in that regard we are at the beginning of a total information era. It's no longer about tech, so much as it is about governance.
I wrote several articles on this a few years back, a side of life I don’t share here much. A few ideas:
When business wants to use AI to regulate and observe, where do they think data comes from to train it?
The entire shipping history of FEDEX from their very start could be stored for the equivalent of a very nice dinner in downtown San Francisco.
If every single event - every transaction, every mail, every phone call, proposal, design, bill of materials, repair, support call, hire, training, spreadsheet, PowerPoint - all of it - a company generates were stored in a blockchain, it would make the future staggeringly easier to manage as tools evolve.
You cannot create data from the past, lost, gone forever.
Autoregressive models will be built on such data. Then GAN tools will be used to evolve businesses very rapidly into highly competitive machines.
The mathematics of statistical thermodynamics to manage businesses is already fully developed. It just has insufficient sampling.
I worry that this oversells the value of the raw data compared to enabling interventions. Yes, obviously it should be tracked but aside from things you just need to have records of the real value isn't in mere correlation but in eliciting causal information from that data.
I really like the idea of software that helps enable, encourage and record this kind of information at a company wide level. I mean maybe you can track hires and how new hires correlate with productivity in a group but can your software help implement a 6 month delay on new hires for a random 5% of groups and analyze the effect?
It could, only if the business understood the effect, and of course it depends on the category of the hires. For example it would be basically impossible to understand the direct effect on revenue generation in hiring new sales reps because their business is subjective for the most part. Think of batting averages over six months for rookies who haven't faced any MLB pitchers. These averages will not be consistent. Also the nature of baseball means a handful of hires is about all you get.
On the other hand in manufacturing, if you have a fleet of machines in a factory with maintenance issues, there is a very direct correlation between their uptime and factory output. Hiring certified mechanics to fix broken makes a tractable difference.
But Data Science becomes more useful when there is a massive amount of consistently tracked data at root. If I can track every sort of data related to an organization going back decades, i'm in an ideal position. We are at a point at which, for most organizations, the capacity exists. This capacity didn't exist 20 years ago. So in that regard we are at the beginning of a total information era. It's no longer about tech, so much as it is about governance.