Good Article on Why AI Projects Fail
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.
The Downside of Easy (or, the Upside of a Good Challenge)
As a young boy, I was “that kid” who would take everything apart, often leaving a formerly functional alarm clock in a hundred pieces in a shoe box. I loved figuring out how things worked, and how components worked together as a system. When I was 10, I spent one winter completely disassembling and reassembling my Suzuki TM75 motorcycle in my bedroom (my parents must have had so much more patience and understanding than I do as a parent). It was rebuilt by spring and ran like a champ. Beginners luck?
By then I was hooked – I enjoyed working with my hands and fixing things. That was a great skill to have while growing up as it provided income and led to the first company I started at age 18. There was always a fair degree of trial and error involved with learning, but experience and experimentation led to simplification and standardization. That became the hallmark to the programs I wrote, and later the application systems that I designed and developed. It is a trait that has served me well over the years.
Today I still enjoy doing many things myself, especially if I can spend a little bit of time and save hundreds of dollars (which I usually invest in more tools). Finding examples and tutorials on YouTube is usually pretty easy, and after watching a few videos for reference the task is generally easy. There is also a sense of satisfaction to a job well done. And most of all, it is a great distraction to everything else going on that keeps your mind racing at 100 mph.
My wife’s 2011 Nissan Maxima needed a Cabin Air Filter, and instead of paying $80 again to have this done I decided to do it myself. I purchased the filter for $15 and was ready to go. This shouldn’t take more than 5 or 10 minutes. I went to YouTube to find a video but no luck. Then, I started searching various forums for guidance. There were a lot of posts complaining about the cost of replacement, but not much about how to do the work. I finally found a post that showed where the filter door was. I could already feel that sense of accomplishment that I was expecting to have in the next few minutes.
But fate, and apparently a few sadistic Nissan Engineers had other ideas. First, you needed to be a contortionist in order to reach the filter once the door was removed. Then, the old filter was nearly impossible to remove. And then once the old filter was removed I realized that the length of the filter entry slot was approximately 50% of the length of the filter. Man, what a horrible design!
A few fruitless Google searches later I was more intent than ever on making this work. I tried several things and ultimately found a way to fold the filter where it was small enough to get through the door and would fully open once released. A few minutes later I was finally savoring my victory over that hellish filter.
This experience made me recall “the old days.” Back in 1989 I was working for a marketing company as a Systems Analyst and was given the project to create the “Mitsubishi Bucks” salesperson incentive program. Salespeople would earn points for sales, and could later redeem those points on Mitsubishi Electronics products. It was a very popular and successful incentive program.
Creating the forms and reports was straight forward enough, but tracking the points presented a problem. I finally thought about how a banking system would work (remember, no Internet and few books on the topic, so this was reinventing the wheel) and designed my own. It was very exciting and rock solid. Statements could be reproduced at any point in time, and there was an audit trail for all activity.
Next, I needed to create validation processes and a fraud detection system for incoming data. That was rock solid as well, but instead of being a good thing it turned out to be a real headache and cause of frustration.
Salespeople would not always provide complete information, might have sloppy penmanship, or would do other things that were odd but legitimate. Despite that, they expected immediate rewards and having their submissions rejected apparently created more frustration than incentive.
So, I was instructed to turn the dial way back. I let everyone know that while this would minimize rejections it would also increase the potential for fraud, and created a few reports to identify potentially fraudulent activity. It was amazing how creative people could be when trying to cheat the system, as well as how you could identify patterns to more quickly identify that type of activity. By the third month the system was trouble free.
It was a great learning experience from beginning to end. Best of all, it ran for several years once I left – something I know because every month I was still receiving the sample mailing with the new sales promotions and “Spiffs” (sales incentives). This reflection also made me wonder how many things are not being created or improved today because it is too easy to follow an existing template.
We used to align fields and columns in byte order to minimize record size, overload operators, etc. in order to maximize space utilization and maximize performance. Code was optimized for maximum efficiency because memory was scarce and processors slow. Profiling and benchmarking programs brought you to the next level of performance. In a nutshell, you were forced to really understand and become proficient with the technology used out of necessity. Today those concepts have become somewhat of a lost art.
There are many upsides to easy.
- My team sells more and closes deals faster because we make it easy for our customers to buy, implement, and start receiving value on the software we sell.
- Hobbyists like myself are able to accomplish many tasks after watching a short video or two.
- People are willing to try things that they may not have if getting started would not have been so easy.
But, there may also be downsides relative to innovation and continuous improvement simply because easy is often good enough.
What will the impact be to human behavior once Artificial Intelligence (AI) becomes a reality and is in everyday use? It would be great to look ahead 25, 50, or 100 years and see the full impact of emerging technologies, but my guess is that I will see many of the effects in my own lifetime.
Discussions that Seed the Roots of Creativity
A few months ago I purchased Fitbit watches for my children and myself. My goals were twofold. First, I was hoping that they would motivate all of us to be more active. Second, I wanted to foster a sense of competition (including fair play and winning) within my children. Much of their pre-High School experiences focused on “participation,” as many schools feel that competition is bad. Unfortunately, competition is everywhere in life, so if don’t play to win you may not get the opportunity to play at all.
It is fun seeing them push to be the high achiever for the day, and to continually push themselves to do better week-by-week and month-by-month. I believe this creates a wonderful mindset that makes you want to do more, learn more, achieve more, and make an overall greater impact with everything they do. People who do that are also more interesting to spend time with, so that is a bonus.
Recently my 14 year-old son and I went for a long walk at night. It was a cold, windy, and fairly dark night. We live in fairly rural area so it is not uncommon to see and hear various wild animals on a 3-4 mile walk. I’m always looking for opportunities to teach my kids things in a way that is fun and memorable, and in a way that they don’t realize they are being taught. Retention of the concepts is very high when I am able to make it relevant to something we are doing.
That night we started talking about the wind. It was steady with occasional gusts, and at times it changed direction slightly. I pointed out the movement on bushes and taller grass on the side of the road. We discussed direction, and I told him to think about the wind like an invisible arrow, and then explained how those arrows traveled in straight lines or vectors until they met some other object. We discussed which object would “win,” and how the force of one object could impact another object. My plan was to discuss Newton’s three laws of motion.
My son asked if that is why airplanes sometimes appear to be flying at an angle but are going straight. He seemed to be grasping the concept. He then asked me if drones would be smart enough to make those adjustments, which quickly led to me discussing the use potential future of “intelligent” AI-based drones by the military. When he was 9 he wanted to be a Navy SEAL, but once he saw how much work that was he decided that he would rather be Transformer (which I explained was not a real thing). My plan was to use this example to discuss robotics and how you might program a robot to do various tasks, and then move to how it could learn from the past tasks and outcomes. I wanted him to logically break down the actions and think about managing complexity. But, no such luck that night.
His mind jumped to “Terminator” and “I, Robot.” I pointed out that Science Fiction does occasionally become Science Fact, which makes this type of discussion even more interesting. I also pointed out that there is spectrum between the best possible outcome – Utopia, and the worst possible outcome – Dystopia, and asked him what he thought could happen if machines could learn and become smarter on their own.
His response was that things would probably fall somewhere in the middle, but that there would be people at each end trying to pull the technology in their direction. That seemed like a very enlightened estimation. He asked me what I thought and I replied that I agreed with him. I then noted how some really intelligent guys like Stephen Hawking and Elon Musk are worried about the dystopian future and recently published a letter to express their concerns about potential pitfalls of AI (artificial intelligence). This is where the discussion became really interesting…
We discussed why you would want a program or a robot to learn and improve – so that it could continue to become better and more efficient, just like a person. We discussed good and bad, and how difficult it could be to control something that doesn’t have morals or understand social mores (which he felt if this robot was smart enough to learn on its own that it would also learn those things based on observations and interactions). That was an interesting perspective.
I told him about my discussions with his older sister, who wants to become a Physician, about how I believe that robotics, nanotechnology, and pharmacology will become the future of medicine. He and I took the logical next step and thought about a generic but intelligent medicine that identified and fixed problems independently, and then sent the data and lessons learned for others to learn from.
I’m sure that we will have an Internet of Things (IoT) discussion later, but for now I will tie this back to our discussion and Fitbit wearable technology.
After the walk I was thinking about what just happened, and was pleased because it seemed to spark some genuine interest in him. I’m always looking for that perfect recipe for innovation, but it is elusive and so far lacks repeatability. It may be possible to list many of the “ingredients” (intelligence, creativity, curiosity, confidence (to try and accept and learn from failure), multi-disciplinary experiences and expertise) and “measurements” (such as a mix of complementary skills, a mix of roles, and a special environment (i.e., strives to learn and improve, rewards both learning and success but doesn’t penalize failure, and creates a competitive environment that understands that in most cases the team is more important than any one individual)).
That type of environment is magical when you can create it, but it takes so much more than just having people and a place that seem to match the recipe. A critical “activation” component or two is missing. Things like curiosity, creativity, ingenuity, and a bit of fearlessness.
I tend to visualize things, so while I was thinking about this I pictured a tree with multiple “brains” (my mental image looked somewhat like broccoli) that had visible roots. Those roots were creative ideas that went off in various directions. Trees with more roots that were bigger and went deeper would stand out in a forest of regular trees.
Each major branch (brain/person) would have a certain degree of independence, but ultimately everything on the tree worked as a system. To me, this description makes so much more sense than the idea of a recipe, but it still doesn’t bring me closer to being able map the DNA of this imaginary tree.
At the end of our long walk it seemed that I probably learned as much as my son did. We made a connection that will likely lead to more walks and more discussions.
And in a strange way, I can thank the purchase of these Fitbit watches for being the motivation for an activity that led to this amazing discussion. From that perspective alone this was money well spent.
Ideas are sometimes Slippery and Hard to Grasp
I started this blog the goal of it becoming an “idea exchange,” as well a way to pass along lessons learned to help others. Typical guidance for a blog is to focus on one thing only and do it well in order to develop a following. That is especially important if you want to monetize the blog, but that is not and has not been my goal.
One of the things that has surprised me is how different the comments and likes are for each post. Feedback from the last post was even more diverse and surprising than usual. It ranged from comments about “Siri vs Google,” to feedback about Sci-Fi books and movies, to Artificial Intelligence.
I asked a few friends for feedback and received something very insightful (Thanks Jim). He stated that he found the blog interesting, but wasn’t sure what the objective was. He went on to identify several possible goals for the last post. Strangely enough, his comments mirrored the type of feedback that I received. That pointed out an area for improvement to me, and I appreciated that, as well as the wisdom of focusing on one thing. Who knows, maybe in the future…
This also reminded me of a white paper written 12-13 years ago by someone I used to work with. It was about how Bluetooth was going to be the “next big thing.” He had read an IEEE paper or something and saw potential for this new technology. His paper provided the example of your toaster and coffee maker communicating so that your breakfast would be ready when you walk into the kitchen in the morning.
At that time I had a couple of thoughts. Who cared about something that only had a 20-30 foot range when WiFi was becoming popular and had much greater range? In addition, a couple of years earlier I had a tour of the Microsoft “House of the Future,” in which everything was automated and key components communicated with each other. But everything in the house was all hardwired or used WiFi – not Bluetooth. It was easy to dismiss his assertion because it seemed to lack pragmatism, and the value of the idea was difficult to quantify given the use case provided.
Looking back now I view that white paper as having insight (if it were visionary he would have come out with the first Bluetooth speakers, or car interface, or even phone earpiece and gotten rich), but it failed to present use cases that were easy enough to understand yet different enough from what was available at the time to demonstrate the real value of the idea. His expression of idea was not tangible enough and therefore too slippery to be easily grasped and valued.
I’m a huge believer that good ideas sometimes originate where you least expect them. Often those ideas are incremental in nature – seemingly simple and sometimes borderline obvious, often building on some other idea or concept. An idea does not need to be unique in order to be important or valuable, but it does need to be presented in a way that is easy to understand the benefits, differentiation, and value. That is just good communication.
One of the things I miss most from when my consulting company was active was the interaction between a couple of key people (Jason and Peter) and myself. Those guys were very good at taking an idea and helping build it out. This worked well because we had some overlapping expertise and experiences as well as skills and perspectives that were more complementary in nature. That diversity increased the depth and breadth to our efforts to develop and extend those ideas by asking the tough questions early and ensuring that we could convince each other of the value.
Our discussions were creative and highly collaborative as well as a lot of fun. Each of us improved from them, and the outcome was usually something viable from a commercial perspective. As a growing and profitable small business you need to constantly innovate to differentiate yourself from your competition. Our discussions were driven as much by necessity as they were by intellectual curiosity, and I personally believe that this was part of the magic.
So, back to the last post. I view various technologies as building blocks. Some are foundational and others are complementary. To me, the key is not viewing those various technologies as competing with each other. Instead, I look for potential value created by integrating them with each other. That may not always possible and does not always lead to something better, but occasionally it does so to me it is a worthwhile exercise. With regard to voice technology, I do believe that we will see more, better and smarter applications of it – especially as realtime systems become more complex due to the use of an increasing number of specialized component systems and sensors.
While today’s smartphones would not pass the Turing Test or proposed alternatives, they are an improvement over more simplistic voice translation tools available just a few years ago. Advancement requires the tools to understand context in order to make inferences. This brings you closer to machine learning, and big data (when done right) significantly increases that potential.
Ultimately, this all leads back to Artificial Intelligence (at least in my mind). It’s a big leap from a simple voice translation tool to AI, but when viewed as building blocks it is not such a stretch.
Now think about creating an interface (API) that allows one smart device to communicate with another in a manner akin to the collaborative efforts described above with my old team. It’s not simply having a front-end device exchanging keywords or queries with a back-end device. Instead, it is two or more devices and/or systems having a “discussion” about what is being requested, looking at what each component “knows,” asking clarifying questions and making suggestions, and then finally taking that multi-dimensional understanding of the problem to determine what is really needed.
So, possibly not true AI, but a giant leap forward from what we have today. That would help turn the science fiction of the past into science fact in the near future. The better the understanding and inferences by the smart system, the better the results.
I also believe that the unintended consequences of these new smart systems is that as they become more human-like in their approach the more likely they will be to make errors like a human. Hopefully they will be able to back test recommendations to validate and minimize errors. If they are intelligent enough to monitor results and make suggestions about corrective actions when they determine that the recommendation is not having the optimal desired results would make them even “smarter.” Best of all there won’t be an ego creating a distortion filter on the results. Or maybe there will…
A lot of the building blocks required to create these new systems are available today. But, it takes both vision and insight to see that potential, translate ideas from slippery and abstract to tangible and purposeful, and then start building something really cool. As that happens we will see a paradigm shift in how we interact with computers and how they interact with us. That will lead us to the systematic integration that I wrote about in a big data / nanotechnology post.
So, what is the real objective of my blog? To get people thinking about things in a different way, to foster collaboration and partnerships between businesses and educational institutions in order to push the limits of technology, and to foster discussion about what others believe the future of computing and smart devices will look like. I’m confident that I will see these types of systems in my lifetime, and believe in the possibility of a lot of this occurring within the next decade.
What are your thoughts?