Today I ran across this article that was very good as it focused on lessons learned, which potentially helps everyone interested in these topics. It contained a good mix of problems at a non-technical level.
Below is the link to the article, as well as commentary on the Top 3 items listed from my perspective.
The article starts by discussing how the “problem” being evaluated was misstated using technical terms. It led me to believe that at least some of these efforts are conducted “in a vacuum.” That was a surprise given the cost and strategic importance of getting these early-adopter AI projects right.
In Sales and Marketing you start the question, “What problem are we trying to solve?” and evolve that to, “How would customers or prospects describe this problem in their own words?” Without that understanding, you can neither initially vet the solution nor quickly qualify the need for your solution when speaking with those customers or prospects. That leaves a lot of room for error when transitioning from strategy to execution.
Increased collaboration with Business would likely have helped. This was touched on at the end of the article under “Cultural challenges,” but the importance seemed to be downplayed. Lessons learned are valuable – especially when you are able to learn from the mistakes of others. To me, this should have been called out early as a major lesson learned.
This second area had to do with the perspective of the data, whether that was the angle of the subject in photographs (overhead from a drone vs horizontal from the shoreline) or the type of customer data evaluated (such as from a single source) used to train the ML algorithm.
That was interesting because it appears that assumptions may have played a part in overlooking other aspects of the problem, or that the teams may have been overly confident about obtaining the correct results using the data available. In the examples cited those teams did figure those problems out and took corrective action. A follow-on article describing the process used to make their root cause determination in each case would be very interesting.
As an aside, from my perspective, this is why Explainable AI is so important. There are times that you just don’t know what you don’t know (the unknown unknowns). Being able to understand why and on what the AI is basing its decisions should help with providing better quality curated data up-front, as well as being able to identify potential drifts in the wrong direction while it is still early enough to make corrections without impacting deadlines or deliverables.
This didn’t surprise me but should be a cause for concern as advances are made at faster rates and potentially less validation is made as organizations race to be first to market with some AI-based competitive advantage. The last paragraph under ‘Training data bias’ stated that based on a PWC survey, “only 25 percent of respondents said they would prioritize the ethical implications of an AI solution before implementing it.”
The discussion about the value of unstructured data was very interesting, especially when you consider:
- The potential for NLU (natural language understanding) products in conjunction with ML and AI.
- This is a great NLU-pipeline diagram from North Side Inc. in Canada, one of the pioneers in this space.
- The importance of semantic data analysis relative to any ML effort.
- The incredible value that products like MarkLogic’s database or Franz’s AllegroGraph provide over standard Analytics Database products.
- I personally believe that the biggest exception to assertion this will be from GPU databases (like OmniSci) that easily handle streaming data, can accomplish extreme computational feats well beyond those of traditional CPU based products, and have geospatial capabilities that provide an additional dimension of insight to the problem being solved.
Update: This is a link to a related article that discusses trends in areas of implementation, important considerations, and the potential ROI of AI projects: https://www.fastcompany.com/90387050/reduce-the-hype-and-find-a-plan-how-to-adopt-an-ai-strategy
This is definitely an exciting space that will experience significant growth over the next 3-5 years. The more information, experiences, and lessons learned shared the better it will be for everyone.
Recently I was helping one of my children research a topic for a school paper. She was doing well, but the results she was getting were overly broad. So, I taught her some “Google-Fu,” explaining how you can structure queries in ways that yield better results. She replied that search engines should be smarter than that. I explained that sometimes the problem is that search engines look at your past searches and customize results as an attempt to appear smarter or to motivate someone to do or believe something.
Unfortunately, those results can be skewed and potentially lead someone in the wrong direction. It was a good reminder that getting the best results from search engines often requires a bit of skill and query planning, as well as occasional third-party validation.
Then the other day I saw this commercial from Motel 6 (“Gas Station Trouble”) where a man has problems getting good results from his smart phone. That reminded me of seeing someone speak to their phone, getting frustrated by the responses received. His questions went something like this:
“Siri, I want to take my wife to dinner tonight, someplace that is not too far away, and not too late. And she likes to have a view while eating so please look for something with a nice view. Oh, and we don’t want Italian food because we just had that last night.”
Just as amazing as the question being asked was watching him ask it over and over again in the exact same way, each time becoming even more frustrated. I asked myself, “Are smartphones making us dumber?” Instead of contemplating that question I began to think about what future smart interfaces would or could be like.
I grew up watching Sci-Fi computer interfaces like “Computer” on Star Trek (1966), “HAL” on 2001 : A Space Odyssey (1968), “KITT” from Knight Rider (1982), and “Samantha” from Her (2013). These interfaces had a few things in common:
- They responded to verbal commands;
- They were interactive – not just providing answers, but also asking qualifying questions and allowing for interrupts to drill-down or enhance the search (e.g., with pictures or questions that resembled verbal Venn diagrams);
- They often provided suggestions for alternate queries based on intuition. That would have been helpful for the gentleman trying to find a restaurant.
Despite having 50 years of science fiction examples, we are still a long way off from realizing that goal of a truly intelligent interface. Like many new technologies, they were originally envisioned by science fiction writers long before they appeared in science.
There seems to be a spectrum of common beliefs about modern interfaces. On one end there are products that make visualization easy, facilitating understanding, refinement and drill-down of data sets. Tableau is a great example of this type of easy to use interface. At the other end of the spectrum the emphasis is on back-end systems – robust computer systems that digest huge volumes of data and return the results to complex queries within seconds. Several other vendors offer powerful analytics platforms. In reality, you really need a strong front-end and back-end if you want to achieve the full potential of either.
But, there is so much more potential…
I predict that within the next 3 – 5 years we will see business and consumer interface examples (powered by Natural Language Processing, or NLP) that are closer to the verbal interfaces from those familiar Sci-Fi shows (albeit with limited capabilities and no flashing lights).
Within the next 10 years I believe we will have computer interfaces that intuit our needs and facilitate the generation of correct answers quickly and easily. While this is unlikely to be at the level of “The world’s first intelligent Operating System” envisioned in the movie “Her,” and probably won’t even be able to read lips like “HAL,” it should be much more like HAL and KITT than like Siri (from Apple) or Cortana (from Microsoft).
Siri was groundbreaking consumer technology when it was introduced. Cortana seems to have taken a small leap ahead. While I have not mentioned Google Now, it is somewhat of a latecomer to this consumer smart interface party, and in my opinion is behind both Siri and Cortana.
So, what will this future smart interface do? It will need to be very powerful, harnessing a natural language interface on the front-end with an extremely flexible and robust analytics interface on the back-end. The language interface will need to take a standard question (in multiple languages and dialects) – just as if you were asking a person, deconstruct it using Natural Language Processing, and develop the proper query based on the available data. That is important but only gets you so far.
Data will come from many sources – things that we consider today with relational, object, graph, and NoSQL databases. There will be structured and unstructured data that must be joined and filtered quickly and accurately. In addition, context will be more important than ever. Pictures and videos could be scanned for facial recognition, location (via geotagging), and in the case of videos analyze speech. Relationships will be identified and inferred based on a variety of sources, using both data and metadata. Sensors will collect data from almost everything we do and (someday) wear, which will provide both content and context.
The use of Stylometry will identify outside content likely related to the people involved in the query and provide further context about interests, activities, and even biases. This is how future interfaces will truly understand (not just interpret), intuit (so it can determine what you really want to know), and then present results that may be far more accurate than we are used to today. Because the interface is interactive in nature it will provide the ability to organize and analyze subsets of data quickly and easily.
So, where do I think that this technology will originate? I believe that it will be adapted from video game technology. Video games have consistently pushed the envelope over the years, helping drive the need for higher bandwidth I/O capabilities in devices and networks, better and faster graphics capabilities, and larger and faster storage (which ultimately led to flash memory and even Hadoop). Animation has become very lifelike and games are becoming more responsive to audio commands. It is not a stretch of the imagination to believe that this is where the next generation of smart interfaces will be found (instead of from the evolution of current smart interfaces).
Someday it may no longer be possible to “tweak” results through the use or omission of keywords, quotation marks, and flags. Additionally, it may no longer be necessary to understand special query languages (SQL, NoSQL, SPARQL, etc.) and syntax. We won’t have to worry as much about incorrect joins, spurious correlations and biased result sets. Instead, we will be given the answers we need – even if we don’t realize that this was what we needed in the first place. At that point computer systems may appear nearly omniscient.
When this happens parents will no longer need to teach their children “Google-Fu.” Those are going be interesting times indeed.
Back in early 2011 myself and 15 other members of the Executive team at Ingres were taking a bet on the future of our company. We knew that we needed to do something big and bold, and decided to build what we thought the standard data platform would be in 5-7 years. A small minority of the people on that team did not believe this was possible and left, while the rest of us focused on making that happen. There were three strategic acquisitions to fill-in the gaps on our Big Data platform. Today (as Actian) we have nearly achieved our goal. It was a leap of faith back then, but our vision turned out to be spot-on and our gamble is paying off today.
Every day my mailbox is filled with stories, seminars, white papers, etc. about Big Data. While it feels like this is becoming more mainstream, it is interesting to read and hear the various comments on the subject. They range from, “It’s not real” and “It’s irrelevant” to “It can be transformational for your business” to “Without big data there would be no <insert company name here>.”
What I continue to find amazing is hearing comments about big data being optional. It’s not – that genie has already been let out of the bottle. There are incredible opportunities for those companies that understand and embrace the potential. I like to tell people that big data can be their unfair advantage in business. Is that really the case? Let’s explore that assertion and find out.
We live in the age of the “Internet of Things.” Data about nearly everything is everywhere, and the tools to correlate that data to gain understanding of so many things (activities, relationships, likes and dislikes, etc.) With smart devices that enable mobile computing we have the extra dimension of location. And, with new technologies such as Graph Databases (based on SPARQL), graphic interfaces to analyze that data (such as Sigma), and identification technology such as Stylometry, it is getting easier than ever to identify and correlate that data.
We are generating increasingly larger and larger volumes of data about everything we do and everything going on around us, and tools are evolving to make sense of that data better and faster than ever. Those organizations that perform the best analysis, get the answers fastest, and act on that insight quickly are more likely to win than the organizations that look at a smaller slice of the world or adopt a “wait and see” posture. So, that seems like a significant advantage in my book. But, is it an unfair advantage?
First, let’s keep in mind that big data is really just another tool. Like most tools it has the potential for misuse and abuse. And, whether a particular application is viewed as “good” or “bad” is dependent on the goals and perspective of the entity using the tool (which may be the polar opposite view of the groups of people targeted by those people or organizations). So, I will not attempt to make judgments about the various use cases, but rather present a few use cases and let you decide.
Scenario 1 – Sales Organization: What if you could not only understand what you were being told a prospect company needs, but also had a way to validate and refine that understanding? That’s half the battle in sales (budget, integration, and support / politics are other key hurdles). Data that helped you understand not only the actions of that organization (customers and industries, sales and purchases, gains and losses, etc.), but also the goals, interests and biases of the stakeholders and decision makers. This could provide a holistic view of the environment and allow you to provide a highly targeted offering, with messaging tailored to each individual. That is possible, and I’ll explain soon.
Scenario 2 – Hiring Organization: As a hiring manager there are many questions that cannot be asked. While I’m not an attorney, I would bet that State and Federal laws have not kept pace with technology. And, while those laws vary state by state, there are likely loopholes that allow for use of public records. Moreover, implied data that is not officially taken into consideration could color the judgment of a hiring manager or organization. For instance, if you wanted to “get a feeling” if a candidate might fit-in with the team or the culture of the organization, or have interests and views that are aligned with or contrary to your own, you could look for personal internet activity that would provide a more accurate picture of that person’s interests.
Scenario 3 – Teacher / Professor: There are already sites in use to search for plagiarism in written documents, but what if you had a way to make an accurate determination about whether an original work was created by your student? There are people who, for a price, will do the work and write a paper for a student. So, what if you could not only determine that the paper was not written by your student, but also determine who the likely author was?
Do some of these things seem impossible, or at least implausible? Personally, I don’t believe so. Let’s start with the typical data that our credit card companies, banks, search engines, and social network sites already have related to us. Add to that the identified information that is available for purchase from marketing companies and various government agencies. That alone can provide a pretty comprehensive view of us. But, there is so much more that’s available.
Think about the potential of gathering information from intelligent devices that are accessible through the Internet, or your alarm and video monitoring system, etc. These are intended to be private data sources, but one thing history has taught us is that anything accessible is subject to unauthorized access and use (just think about the numerous recent credit card hacking incidents).
Even de-identified data (medical / health / prescription / insurance claim data is one major example), which receives much less protection and can often be purchased, could be correlated with a reasonably high degree of confidence to gain understanding on other “private” aspects of your life. The key is to look for connections (websites, IP addresses, locations, businesses, people), things that are logically related (such as illnesses / treatments / prescriptions), and then make as accurate of an identification as possible (stylometry looks at things like sentence complexity, function words, co-location of words, misspellings and misuse of words, etc. and will likely someday take into consideration things like idea density). It is nearly impossible to remain anonymous in the Age of Big Data.
There has been a paradigm shift when it comes to the practical application of data analysis, and the companies that understand this and embrace it will likely perform better than those who don’t. There are new ethical considerations that arise from this technology, and likely new laws and regulations as well. But for now, the race is on!
I have a Corvette that I like to work on for fun and relaxation. It gives me an excuse to learn something new and an opportunity to hone my troubleshooting skills. It can be a fun way to spend a few hours on a weekend.
A few weekends ago I was looking for a few parts for a small project. This was spur of the moment and really didn’t need to be done now (as the car will be stored soon for the winter). I found the parts I needed from a single company, but then something strange happened.
This website had my address, knew the two parts that I wanted, but failed to make the process easy and almost lost a sale. I needed to manually check five different store locations to see if they had both parts. In this case two of the five did. One store was about 5 miles from my house and the other about 20 miles away.
Just think how helpful it would have been for this website to use the data available (i.e., inventory and locations) and present me with the two options or better yet default me to the closest store and note the other store as an option. Using spatial features this would be extremely easy to implement. It’s the equivalent to the “Easy Button” that one office supply uses in their commercials.
Now, take this example one step further. The website makes things quick and easy, leaving me with a very pleasant shopping experience. It could then recommend related items (it did, but by that time I had wasted more time than necessary and was questioning whether or not I should start that project that day). The website could have also created a simple package offer to try to increase my shopping cart value.
All simple things that would generate more money through increased sales and larger sales. It would seem that this would be very easy to justify from both a business and technical perspective, assuming the company is even aware of this issue.
I frequently tell my team that, “People buy easy.” Help them understand what they need to accomplish their goals, price it fairly, demonstrate the value, and they make the rest of the sales process easy to complete. This makes happy customers and leads to referrals. It just makes good business sense to do this.
So, while geospatial technology might not be the solution to all problems, this is a specific use case where it would. The power of computing systems and applications today is that there is so much that can be done so fast, often for reasonably low investment costs in technology. But the first step getting there is to ask yourself, “How could we be making this process easier for our customers?”
A little extra effort and insight can have a huge payoff.
In an earlier post I mentioned that one of the big benefits of geospatial technology is its ability to show connections between complex and often disparate data sets. As you work with Big Data you tend to see the value of these multi-layered and often multi-dimensional perspectives of a trend or event. While that can lead to incredible results, it can also lead to spurious correlations of data.
First, let me state that I am not a Data Scientist or Statistician, and there are definitely people far more expert on this topic than myself. But, if you are like the majority of companies out there experimenting with geospatial and big data it is likely that your company doesn’t have these experts on-staff. So, a little awareness, understanding, and caution can go a long way in this type of scenario.
Before we dig into that more, let’s think about what your goal is:
- Do you want to be able to identify and understand a particular trend – reinforcing actions and/or behavior? –OR–
- Do you want to understand what triggers a specific event – initiating a specific behavior?
Both are important, but they are both different. My personal focus has been on identification of trends so that you can leverage or exploit them for commercial gain. While that may sound a big ominous, it is really what business is all about.
There is a popular saying that goes, “Correlation does not imply causation.” A common example is that for a large fire you may see a large number of fire trucks. There is a correlation, but it does not imply that fire trucks cause fires. Now, extending this analogy, let’s assume that in a major city the probability of multi-tenant buildings starting on fire is relatively high. Since they are a big city, it is likely that most of those apartments or condos have WiFi hotspots. A spurious correlation would be to imply that WiFi hotspots cause fires.
As you can see, there is definitely potential to misunderstand the results of correlated data. More logical analysis would lead you to see the relationships between the type of building (multi-tenant residential housing) and technology (WiFi) or income (middle-class or higher). Taking the next step to understand the findings, rather than accepting them at face value, is very important.
Once you have what looks to be an interesting correlation there are many fun and interesting things you can do to validate, refine, or refute your hypothesis. It is likely that even without high-caliber data experts and specialists you will be able to identify correlations and trends that can provide you and your company with a competitive advantage. Don’t let the potential complexity become an excuse for not getting started, because as you can see above it is possible to gain insight and create value with a little effort and simple analysis.