AI and ML – Comprehensive Q&A

What is Artificial Intelligence?

AI is most often a computer program which is designed to duplicate intelligent thinking.  Notice the word “duplicate” as opposed to “emulate.”  At its current stage of development, although still primitive compared to the human brain, it has, in some particular situations, replicated human thought processes quite precisely.

How does any AI learn?

There are two methods by which an AI can learn—the most familiar of which is being programmed by humans.  This was the case for IBM’s WATSON AI, designed for the single purpose of winning the TV game show Jeopardy!

We can meticulously define a positive, null, and a negative result for every possible situation that we can imagine.  No matter what happens, however, there are always going to be situations that go beyond our imagination.

Instead, to be effective, we are now using a technique which defines a certain number of “good” results and another batch of “bad” results.  Using this as the most rudimentary guideline, the AI is then set loose in a testing environment where it can try every single possibility with its immense speed.  It tries all of them until it develops an understanding (“experience”) of actions that lead to “good” results, “null” results, and “bad” results.

An example of this is Google’s ALPHA GO AI engine that recently beat the best GO player in the world without any formal training in the game.  GO is many, many times more complicated than chess, with thousands more possible moves.  ALPHA GO observed thousands of games between humans and created millions more in simulations.  There is no “right” move in GO.  There is a large, intuitive element to the game.  The AI had to learn to be intuitive to win.

What is an AI system?

Any system which employs relies upon, or is driven by an AI is, by definition, an AI-System.  Even when provided with incomplete data, a properly trained expert system can still arrive at a human-like decision, generally faster than a person could do so, within its area of expertise.

The ultimate goal is to develop a system that can integrate all properly formed databases, encompassing all human knowledge.  This would eliminate dedicated expert systems, and create a single source that could answer all inquiries using all of our collective knowledge.

Knowing what gardeners know, as well as physicists, beekeepers, chemists, boat crews, laser engineers, dermatologists, miners, and Inuit igloo builders will allow it make new connections that humanity might never have found.


What is an AI player?

Once computer games became playable with multiple players online, it also became necessary to create AI “players” to play in the positions ordinarily filled by real humans in the event there were insufficient participants for the game to function.

As an example, in a game calling for six players on each of two teams (12 in total) the game could generate up to 11 players so that a single player could enjoy the game.  AI-players were automatically dropped as humans joined the game, and recreated if a player left the game in the middle of the proceedings.

The AI-player was designed with (albeit limited) responses that could emulate a human player, making logical, or sometimes “odd” decisions based on observing actual humans playing the game.  It is noteworthy to mention that this was not “true” AI, but rather “intelligent action emulation.”

When was AI invented, and by whom?

AI has roots that go back to ancient Greek mythology where “human-like maidens, forged entirely of metal” by the hand of Hephaestus (aka Vulcan, to the Romans) were described.  Mary Shelly’s Dr. Frankenstein was a 19th-century excursion into AI, followed by others such as Czechoslovakian Karel Capek, and his robot play RUR or Rossum’s Universal Robots.

Real AI, had to wait until 1956 for serious consideration at the Dartmouth Conference, but real AI remained stubbornly elusive even when heavily financed by ARPA (precursor to DARPA).  It made limited progress through the 1970s and 1980s, but the extent of the problem was underestimated, and funding frequently failed.

Nowadays we have success, and people are using rudimentary AI in their everyday devices.  Every AI invention seems to make millions for its creator, so AI is finally “cool,” well-financed, and here-to-stay.

Can AI go wrong?

Of course, just as a wheel can come off of your car if the mechanic doesn’t install it correctly, the lousy programming can compromise an AI.  There are many checks and balances to remind us to do our jobs well as we create new intelligence.

Elon Musk and Stephen Hawking (among others) expressed fear about AIs taking over the world, though this is likely a product of too much Hollywood fear-mongering rather than thoughtful consideration.  Famed science writer and Science Fiction author, Isaac Asimov was more sanguine and created Three Laws to govern AI in his fiction work.

Implementing something similar, as a fundamental filter for all AI actions, is a good starting point, but philosophers abound who are more than willing to warn us that we must consider each step carefully.  If we do it right, we’ll be lauded all through history for a job well done, and if we do it incorrectly, no one will be around to mock us— an excellent reason to make sure we do it right.

Will AI replace human beings?

AIs will be capable of replacing humans in most jobs in the future except those requiring creativity.  If you make beautiful jewelry by embedding quail poop in resin, your job is safe since AIs are incapable of recognizing something that might constitute beauty.  Beauty and “Art” are too ephemeral and individualistic to be “computed.”  Early painting experiments by machine have been equally unrewarding.

Airline piloting, train operation, and even sea-going cargo ships are becoming more and more automated.  It is, for example, theoretically possible for an entire airline flight to take place without the need of the pilot touching the controls from take-off to landing.  Self-driving cars will be here a lot sooner than you expect.

Will artificial intelligence become self-aware?

Human babies are not self-aware until age 18 months (no matter what the proud parents might claim to the contrary).  The standard test, the ability of babies to recognize themselves in a mirror (and after that becoming one of their favorite activities), would be simple to program into a computer, so that would be an inadequate test.

Having a subconscious stream of thought, the ability to “dream,” outside of any programming might be revealing.  Philosophers argue that the only person we can be certain has self-awareness is ourselves, as individuals.

From a technical perspective, if a program has goals or objectives beyond its programming (“I would love to see Paris one day”), then that would indicate self-awareness.

When will we have actual AI?

It is already here and functioning.  Intelligence emulation such as Siri or Alexa in our daily devices is brushing the surface for most people.  The more you interact, the more effective these programs become at predicting most of your requirements and desires.

BIG BLUE from IBM was not AI per se, but rather massive mathematical computing capabilities focused on a single problem (winning at chess); WATSON took on the challenge of beating Jeopardy!, and used a complex associational program to figure out the double entendres, puns, and obscure clues to emulate human thought processes and eventually emerge as the winner; ALPHA GO was essentially self-taught in learning how to play go.  It is artificial intelligence, but still vastly inferior to human intelligence.

When is AI used?

AI is ideally suited to pattern-finding tasks, such as analyzing medical X-rays or MRI scans, or forensic accounting to find illegal money laundering schemes or even the everyday scut work of operating a ChatBot to help clients solve typical technical problems.  All of this tedious or ornery work can be handed-off to the AI, freeing up regular employees to engage in more creative work, such as dealing with customers with sophisticated, multi-phase problems, which will ultimately benefit the company.

Who uses AI?

You might think its use is confined to First and Second World nations, but because of the nature of technology and how it is evolving, there is far less infrastructure required in third world nations, such as thousands of miles of copper wire for communications.

Satellites, microwaves, and cell towers now connect towns and villages, where one or more people in each place may possess a communal smartphone. Many more places have an actual laptop powered by solar panels.  Google’s 113+ language translator allows people to speak in any language that can be translated to English, and receive a response in their language.  AI is accessible to anyone who is connected to the Internet.

Who is developing AI?

Everyone in the computer business is using/enhancing someone else’s AI or developing their own.  The Wall Street Journal calls 2018 “The Year of AI” with plenty of justification.  Mainstream companies are now investing in supercomputers of their own.  Most of the leading medical companies are already using AI on a regular basis, not only for diagnosing patients or interpreting medical imagery but for drug research.

In a recent Ebola virus outbreak, one drug research company set its AI loose in its molecular database, and in a matter of hours had found two new candidates as a treatment for the virus outbreak.  This research would ordinarily have taken two years, and still might not have found this result.

You will witness or discover companies using it for medicine, of course, but also for engineering simulation, for identifying you by your biometrics rather than a pin number or a signature, for research into complex multi-discipline subjects (with vastly improved timelines), as vision, speech, hearing, and physical aides for people with disabilities or infirmities, as well as automation of functions that we previously thought were strictly in the domain of human beings such as vehicle operation (planes, trains, ships, automobiles).  Look for the usual candidates, such as Tesla, Google, Amazon, and so on, but there will be many new players entering the field where each discovery is worth millions of dollars.

Who benefits from artificial intelligence?

You do.  Humanity does.  AI has been lauded as the last invention humanity will ever have to create because once we integrate all our databases, the sum of human knowledge, an AI, whose true, best function is to find obscure relationships and patterns in massive amounts of data, will be able to answer virtually any question we can create.

Scientists have already speculated that we have the answer to space travel faster than the speed of light, or teleporters similar to what you may have seen is Star Trek movies.  Food replicators from the same source may already be possible.  All that remains is to sort through the data, but of course, scientists will continue to research and accumulate new data, too.

How is artificial intelligence changing the world?

Daily tasks that steal time can be delegated to AIs.  It’s already happening to a small extent now but will continue to grow in importance giving us much more leisure time.

Ultimately AI is going to seriously disrupt our social structure and economic model when everyone can feed themselves or have any object they desire by telling an AI to create the desired object.  Feed in the required raw materials and a 3D-printer could make anything you want.

Granted, that may be decades away, but those who work towards that goal now will be the biggest winners, at least until money and commerce become meaningless.  After that, raw material suppliers, scientific researchers, and those who create Art will be the most valued people in society.

What was the first AI?

Some say it was ELIZA, away back in 1966, but even its creator, Joseph Weizenbaum, stated emphatically that it was not, and was never intended to be, an Artificial Intelligence.  ELIZA did pass the Turing Test on several occasions, provided one confined themselves to discussing personal events as if speaking to a Rogerian Psychologist, but it was in no way a real AI.

There are three specific types of AI, the first of which is almost ubiquitous in modern society.  It would be difficult to accurately state which “the first” was, so we’ll leave that up to the historians.  Those three classes are:

The ANI (Artificial Narrow Intelligence) that can handle a given task with no human intervention, like your Siri or Alexa devices ordering you an Uber, or playing a song for you.  IBM’s WATSON might be an example.

The AGI (Artificial General Intelligence) that can do what a human can do such as strategize, plan, learn, use so-called “common knowledge,” and solve problems.  Google’s ALPHA GO is approaching this area but does not qualify.

Last is the ASI (Artificial Super-Intelligence), which would be light-years beyond human intelligence.  If your IQ is 100 (average), its IQ would be higher than 10,000.  This is unimaginable to us, as space flight would be to an earthworm.

Business AI

Can AI be used in the energy industry?

AI could help in many ways in the Energy Industry, including real-time operations, starting up/shutting down plants and moderating output according to need.  It could even coordinate between different grids so that one with an excess could shunt its production to an underpowered grid.

The advantages of keeping a system running compared to the start-up costs or shutdown expense or a different system could be perfectly balanced to make sure each utility was making the best use of its resources.  Maximizing profits and minimizing costs would be the ideal job for an AI.

Can AI be used in Oil and Gas Industry?

AIs can interpret exploration data better than humans, as has already been demonstrated in some test cases.  Drilling dry boreholes is one of the most significant expenses in the oil and gas industry.  An AI could reliably predict where resources were to be found based on field reports and satellite data.  Geologists and chemical engineers are extraordinarily competent at their jobs, finding these hidden resources, but with modern techniques the AIs will be able to do the job better, freeing these scientists up to apply their skills elsewhere.

Can AI be used in Finance?

One of the best uses in Finance for AI is forensic accounting.  Knowing where something came from, who approved it, and where it went, are some of the most fundamental aspects of finance.

AI can protect your Financial System, forbidding unprecedented monetary transfers, detecting intrusions, and most importantly preventing your company from being a money laundering operation.

Can AI be used in Retail?

AI in retail can be used in many ways, such as identifying specific customers by social profile, identifying their needs and desires, and presenting them data directly on their device (smartphone) while they are in the immediate vicinity of something they would likely purchase.  E-retail is much the same, presenting items of interest during a shopping event on your website.  It could be as innocuous as a sidebar with an interesting coupon, to a direct pop-up with a limited-time special, aimed directly at that customer.

Can AI be used in Media?

AI is already used in Media on a regular basis.  The most common example is box-score summaries of sporting events.  You’ve probably read such things where the commentary looks human-crafted, such as “#99 Dickie Aukland, defensive tackle for the [team] intercepted a misthrow, and ran in a 54-yard touchdown, his first points for the year.”  It sounds human, but in all likelihood, it was AI generated from game data.  Humans seldom need to trouble themselves with game summaries anymore, focusing on tasks more worthy of their skills.

Can AI be used in Manufacturing?

AI is an integral part of manufacturing, with systems that study news, political unrest in production areas, weather events, trends in media, and so on, to help determine which products are going to be in demand.  It might be the color of a favorite fabric in the fashion industry or the amount of lithium required to build the batteries you need for your electric car, but an AIs System can pick up the relevant data so that you can assure the supply of your raw materials before the season even begins.

Can AI be used in Insurance?

Historically the insurance industry has relied on actuarial tables to mitigate risk.  This is no longer necessary with AI to interpret actual risk based on someone’s social profile.

In auto insurance, for example of someone who has never been caught street racing might be experiencing an excellent insurance rate for their car.  But it is inevitable that someone with this level of vanity will brag somewhere online about their “skills”, and the AI will find it, assess the risk, and then they can pay $1000 a month.  The perfect driver (like you), on the other hand, presenting practically no risk to the company, can pay $300 per year.

Can AI be used in Healthcare?

AI is already being used in Health Care.  It analyzes medical imagery such as PET scans, MRI, X-rays, and more, finding two “fuzzy patches” on opposite sides of an image that a human might never pick up on, resulting in a much earlier diagnosis of a disease process, resulting in much better outcomes.  Diagnostic Medicine is already benefiting from AI.

In Japan, it is also used to drive robots designed to lift patients from bed to wheelchairs, to the bathtub, or to pick them up from the floor.  This is absolutely essential since Japan is experiencing now what we see as the gray tsunami coming in just a few years to North America, as the numbers of our older population surpass the younger generation and those available to care for them.

Can AI be used in the Legal Profession?

An AI Expert System, equipped with all the legal decisions ever recorded, would make many “helper” professions in traditional Law Offices redundant.  Luckily it will also make it possible for former Legal Assistants and Paralegals to open boutique Law Offices.  They could interrogate an AI Expert System on behalf of clients, giving them all the knowledge necessary for civil cases.  Citing precedents and case law would eliminate the need for lawyers.

Clever lawyers will open up their boutique law offices, staffing them with their people that might otherwise be let go.  This will keep this bread-and-butter income flowing while they concentrate on more complex and higher profile cases.

How will AI redefine management?

It will rid the profession of half of their daily work, which is primarily scheduling and logistics.  Imagine if you had a half day suddenly become available, and all your tedious mind-sapping scut work vanished!  This is the future for managers, where they will focus more on connecting team members and finding new ways for them to cooperate and integrate their labors and tasks.

Judgment is one of the most critical skills, but lost in a sea of paperwork; it is often not utilized to best effect.  Managers in the AI age will be facilitators and creative founts.

Everyday ML

What is Machine Learning?

Machine Learning is a branch of Artificial Intelligence using a series of algorithms that allow an AI to interpret the world—they discover data, categorize it, and then learn from it.  A system that has been trained to recognize images of birds can look at an image, determine that it is a bird, and then go on to identify the species.

If it is presented with a picture of a bulldozer, but told it is a bird, it will find the closest match in its database or “knowledge” (perhaps a canary of similar color) and make its pronouncement.  Over the course of training, with appropriate feedback, such a specialist system using thousands or millions of images will gain “experience” and eventually be able to say “That is not a bird” even if it cannot identify that it is a bulldozer.

What is the difference between machine learning and deep learning?

ML algorithms discover data, categorize it, and then learn from it; Deep Learning algorithms are the same, except they are particularly good at “feature detection” which gives them a much higher level of discrimination when interpreting data.  An ML algorithm could look at images of a canary in a cage, a pigeon on a building ledge, and an eagle in flight and respond “Bird,” but a DL algorithm would respond “canary,” “pigeon,” and “eagle.”

This is the same mechanism that humans use to learn, where objects are identified and classified because humans are experts at Deep Learning.  We learn to discriminate between birds “flying underwater” to catch fish, and fish leaping from the water to escape predators.  We are seldom confused by Exocoetida (flying fish) and Cormorants (swimming birds).

What is machine learning all about?

ML is all about discovery, sorting, and learning from data.  A single or small sample is not very useful, but a large sample provides many points of comparison.  An Artificial Intelligence using it can distinguish between many types of data.  The information can be in the form of still or moving images, tables of numbers, or actual words in documents.  Just about anything a human can interpret can be examined and interpreted by a machine.

Is NLP machine learning?

No, neither Natural Language Processing or Natural Language Programming (depending on context) are examples of Machine Learning.  NLP, in either form, is based on Natural Language, meaning that algorithms have been trained to interpret human speech for a specific purpose.  You use NL Programming when you tell your Electronic Assistant to “order an Uber Cab for 4:30”, or to “wake me up in 40 minutes”.

When the Assistant interprets your words to turn them into instructions, it is using NL Processing to accomplish that task.

What is the Difference between machine learning and data analytics?

ML is essentially an automated system for building models of the world (or discrete bodies of data).  By looking at thousands of images of “round” things, it can learn that a featureless two-dimensional circle drawn on a piece of paper and a three-dimensional globe of our planet are both considered to be “round” objects.

Data Analytics, on the other hand, is simply Machine Learning to the nth power, analyzing vast collections of data, collating, organizing, categorizing and sorting it into a series of relationships.  This is what happens to your Big Data.

Business ML

How long does it take to do a POC for a machine learning project?

This is the sort of project that will take 3-4 months, provided you avoid the usual mistakes.  POCs are useless if you don’t pick a real Use Case—make sure yours is substantive.  Make sure you define what you want as a result—not just wait to “see what happens”.  Right-size your team; don’t let every Tom, Dick, and Jeanette hop on board because they’re interested—and that means don’t leave your team too sparse, either!  Don’t let the project drag out—get on with it—and make sure you put it into production.  This isn’t a theory, so make sure it responds to real-world conditions.

How to introduce AI in an organization?

AI is a multi-disciplinary task, requiring many different skills.  You can go with the easiest implementation, which is a supplier’s straight out of the box solution with all sorts of instructions and support; you can buy a partial solution and write your code to tie it into your current system; you can hire some young genius to write, develop, and implement unique code exclusively for you.

The solution you choose is going to depend on your business model and how you conduct business.  Each one has its advantages. Pre-boxed is easy; partial solutions allow you to integrate with existing systems (CRM, HR, ERP, etc.), and unique solutions have their advantages.  Give us a call, and we’ll help you choose the best solution for your particular scenario.

How to build an AI Lab in an organization?

Developing AI in an organization isn’t a one-and-done task.  There is a lot of interdependence.  According to Baidu the tech-monster from China, their most significant success came with interactive quizzes posing real-life problems and solutions in an A)B)C)D) answer format.  Pen & Paper in this modern age?  Yup.  It made people think about their responses, and Baidu ended up with a great AI Lab.  Real Life questions—Real Life answers—Full Buy-in by employees.

How long to do a typical implementation of AI?

Management needs to be aware that AI isn’t some magical sauce you can sprinkle on your business to make it awesome.  If your business isn’t “intelligent” soon, it will cease to exist because it can no longer compete.  The implementation is going to take 4-8 months.

The process is bottom-up, not top-down.  Analyze what you are already good at, implement AI to streamline those processes, and then expand it to other areas where the ROI is less obvious, but tangible.  Management decisions to “become AI compatible before the end of the next Quarter” almost always result in monetary loss and project failure.

Know your strengths—fortify them—expand from there.

Does AI have immediate ROI?

Absolutely, but the C-suite is often looking for a tangible monetary return in the first Quarter “to justify the expenditure,” instead of appreciating the value of customers being thrilled with new, improved services.  If you don’t care about happy customers, you don’t care about recommendations that will see your business volume swell, with no additional revenue-seeking costs.

Clients aware that you’re enhancing your business, and becoming more responsive, are worth their weight in gold when it comes to word of mouth.  People listen to personal recommendations and act upon them—much more so that they respond to additional advertising.

How can I convince my executive team to fund AI projects?

The most important factor is to be credible—not a penny pincher doing everything at the lowest possible cost—who always manages to be highly effective.  Next, you must have a clear plan for getting from point A to point B, using a real business Use Case that they are going to be able to relate to.  You’ll have to show benefits, risks, and mitigation strategies, so they know you’ve looked at the whole picture.  Finally, you need a POC demonstration, so solicit those executives with whom you’ve established rapport, and make sure you have their buy-in before it comes to presentation time.

Can machine learning help with data quality?

There is an adage from the early computing days: GIGO, which means Garbage In = Garbage Out.  Cleansing data for use in Analytics is essential, but with the sheer quantity of data we are now creating every day, there is simply no possibility that a human being can keep up with it.  An ML implementation run by an AI most certainly can keep up with the rate of data creation.

The only requirement is properly formed algorithms to parse the data correctly.  Cleaning data could take weeks or months in a large organization when done manually.  Old data is nearly as bad as wrong data, and decisions based upon it are just as wrong.  AI ML is probably the only practical way to keep up with it.

Can machine learning help HR?

HR is more than just hiring & firing—yes, AI/ML will zip through 500 résumés and present you a short-list of skilled candidates for a particular opening faster than any human being could manage, but there is more to it than that.  During the actual interview, such a system can let an interviewer know (in real-time!) when they are going off-track or let personal bias interfere with their assessment (discretely, on a screen, so safe for online or “live” interviews).  It can also handle e-mail communications with candidates, and schedule interviews for those selected.

Do you want to understand your churn rate better?  AI/ML will help you predict attrition, too, something which took the majority of businesses by surprise when it came time for the Baby Boomers to retire.  It can free up hours of time spent organizing on-boarding, too, creating customized programs for new employees, assigning desk space based on who they need to interact with, and preprogramming a blank laptop with the tools they need to do their jobs.  It has a place in HR.

Can machine learning help Operations?

ML is extremely capable of predicting outcomes based on available data.  This simplifies many aspects of traditional Operations.  It doesn’t need training per se, but instead uses algorithms for iterative observation to teach itself from the data.  It can spot inefficiencies and streamline procedures; it can manage inventory (JiT is a good choice), SCM, ERP, and many other aspects, freeing up Operations staff to handle much more creative tasks; it gives back the most valuable resource, time, by absorbing unrewarding and tedious work.

Can machine learning help in sales and marketing?

AI/ML can customize sales strategies for individual customers, or customers by group, occupation, or geography.  Sending generic e-mail solicitations loses a customer’s respect.  If (for example) they always buy 50 bag lots of concrete mix, two skids of drywall, and bricks by the ton, sending them ads for garden bug-zappers, throw rugs, or indoor lighting sales makes no sense.  If every solicitation is relevant to them, they look forward to your e-mails because they’re always useful and timely.

In the B2B arena, it can track social media and business news to see that your customer is opening a new office or plant, and predict what additional resources you can supply to them as a result.  It can help you offer discounts for up-sizing purchase orders, or exploring new products.

Can machine learning help in Accounting?

Instead of situational analysis looking for a particular result, a company’s entire ledger could be scanned into an AI/ML algorithm and all sorts of insights and revelations would appear.  It could be interrogated for specifics, or present results based on the most common inquiries.  In either case, most of the scut work of Forensic Accounting becomes automated, leaving the accountant free to assess the results rather than spending hours “checking the numbers.”

Putting the entire ledger in would allow the AI/ML algorithm to identify common transactions, and then isolate all the “abnormal transactions” so they were ready for immediate analysis.  This pattern-identification is the ideal work for AI.

Can machine learning help in R&D?

It already has, in innumerable cases.  The best example is of a medical chemical giant, faced with an Ebola outbreak, who turned their AI loose in their molecular database.  It found two new drug treatment candidates in just a couple of hours which, under ordinary circumstances, would have taken years to accomplish.  It was ready for clinical trials in days, instead of years.  This is the future of R&D.

Can machine learning help in Purchasing?

Purchasing is all about predicting customer behaviors, wants, and desires.  Knowing that “Aqua will be the new Black this season” could put a fashion design house miles ahead of its competition—and the same applies to any business, including yours.

Social Media and other publically available information show trends, and in what direction the public’s desire is evolving.  AI/ML is ideally suited to sorting through thousands of tweets, likes, and other indicators, spotting subtle trends that humans haven’t recognized yet.  It’s the closest thing we have to predicting the future.  What are people buying or desiring?  Now you will know.

Can machine learning help in Production?

Perfect Scheduling alone could save companies a fortune, but integrating downtime into a schedule when it is most necessary but will have the smallest impact on production is priceless.  AI/ML could also improve product quality by noting that certain excessive production speeds results in more defect identifications or returns.  It can schedule production for optimal delivery, the smallest warehousing requirements, and manage JiT implementation to free up more capital for better cash flow.

Information in minutes instead of months can positively affect Finance, Operations, R&D, and Supply Chains.  Instantly knowing the status of WIP (Work In Progress) makes life easier for Sales and Logistics, too.

What is the difference between data analysis and data analytics?

Data Analysis is a broad term describing the collection and analysis of data so that results can be presented to management, usually to enhance business decision making.  Data Analytics is just a portion of Data Analysis, which is focused on using specific tools and techniques to acquire information.

What is data science and analytics?

Data Science is distinct from Analytics because DS extracts insights and knowledge from data with custom-designed systems using algorithms, processes and scientific methods.  Analytics uses that data and knowledge to create reports to enhance business operations and strategies.  Data Science is like a Library; Analytics is like a Book.

What is RPA?

Robotic Process Automation is a technique that often utilizes sophisticated techniques such as AI/ML to train a robot to perform a boring, repeatable task.  One of the most familiar examples is the automotive body welder robots that make identical welds on hundreds of identical units per day.

Difference between RPA (robotic process automation) and ML?

In a physical process (see next question for the software process), Robotic Process Automation can be trained by mechanical memory recording.  For example, when a human being moves a device (say a welder) in a particular fashion to accomplish a task, each movement is fed back by spatial sensors, recorded, and the task is repeated many times so that subtle variations can be homogenized into one “ideal” action.  Now a computer can replay the action “set” whenever it is required, and the device powered by actuators duplicates the previous actions exactly.

Such a process taught instead by an AI/ML program observing the task (and even receiving additional input from the same spatial sensors) might replay the actions millions of times, finding faster, more efficient movements that just weren’t possible with a human body in the way.

What is an RPA platform?

RPA platforms are software robots reliant on AI/ML technology which can perform dull, repetitive tasks formerly done by humans.  Whereas once a programmer had to create a step-by-step set of instructions for a computer to automate a task, ML can observe that task being done, dozens or hundreds of times, build its own set of instructions, and then perform that task right within the Graphical User interface (GUI).

What is Blue Prism RPA?

Blue Prism is the name of a UK company that led the development of RPA software in the early 2000s.  Their objective was to automate data-entry processes that consumed large amounts of human manual activity, which were at high risk of error.  They aimed to provide a digital workforce to the often overlooked back office functions of a business.  It is a coding-free way to automate a process using a drag-and-drop interface.

Is RPA artificial intelligence?

RPA’s only purpose is to automate a rule-based task.  That can be accomplished in many ways, including using AI or ML to enhance the learning.  AI, on the other hand, is programming that emulates, mimics, or ultimately, actually replicates human mental processes.

What is deep learning?

Deep Learning describes the ability of a program to identify relationships and to understand associations, but more importantly, to then apply that understanding to entirely different situations.

What is neural learning?

Imagine you are given one million photographs that may or may not have cars in each one.  Going through them you might eventually notice that some shared characteristics and, even without knowing what a “car” was, the preponderance of similarities might guide you to the notion that “car” might describe these similar aspects.

If some helpful person then started identifying some of the cars as cars, pretty soon you would be able to locate almost every instance of a car in the images.  That is Neural Learning.

What is a deep neural network?

Typically a DNN possesses many sub-layers that (for example) a Computer Vision image will pass through for the AI to determine what the image is.

The same technique is used for Voice Recognition where layers may represent sex (timber), region (dialect), pitch, tone, emotive content, tension, and so on.  Without this variety of layers, a Bronx/Yiddish/New York accent might be unintelligible to a Southern Tennessee trained voice recognition system.


What is the difference between AI and ML?

At its most basic, AI is merely programming that attempts to replicate human mental processes in hardware.  It uses the experience to extrapolate future data, similar to how human beings assess data.

ML is a specific type of AI that does not require manual training.  Instead, it interprets incoming data of a particular kind and uses statistical analysis to draw conclusions about the data.

It operates in one of two modes: Supervised and Unsupervised.

Supervised mode means it gets feedback from humans that “Yes, it did assess a situation correctly,” or “No it did not assess a situation properly.”

Unsupervised mode uses an iterative approach (Deep Learning) to draw conclusions.  With millions of examples, they learn very well and can be stunningly accurate.  They are most often used for much more complex tasks than Supervised ML systems.

Who are the top service providers for AI and ML?

There were almost 3,000 AI companies in the United States in 2017, 230 in Canada, 700 in China, 365 in the UK, and ±200 each in India, Israel, and Germany (among many others).  That is an extensive collection, and it will be chockful of familiar names like Amazon, Apple, Baidu, Google, IBM, and Intel.

Those companies may be anxious to supply tools (often “free”) for you to implement AI, which can be useful if you possess the skills.  If we’re honest with ourselves, however, they’re not going to be interested in direct support for your plans, hopes, dreams, and ambitions.

MegaCorps are great for almost unlimited research money and pushing the boundaries of development—something more cash-constrained enterprises can’t do—but their size also doesn’t allow them to get down in the dirt with you to help plant the seeds that will grow your company.


We at WildFire would like to suggest our company as the one that is vitally interested in your success; as the one that will work with you to solve your challenges and help you to advance to the forefront in your industry; as the ones that will help you to become leaders rather than followers…

Give us a call at 908.758.1244 or use our contact form above to request a consultation.  We’re here to help, and we’d love to hear from you!