How companies are going wrong with AI
wanting AI isn’t enough. You got to figure out where it fits into your business puzzle. It’s pretty pointless to buy some fancy AI thing and then scratch your head trying to find out what it’s good for.
We are living in an era where AI and machine learning have become the new buzz word. They’re like the cool kids in high school that everyone wants to be associated with. Marketing strategies further contribute to the hype. A product labelled “AI-powered”, possesses an irresistible allure — the label that magically transforms any ordinary devices into cutting-edge marvels. Whether it’s smartphones boasting AI-driven cameras or refrigerators with AI temperature control, the market is awash with these buzzwords.
However, let’s pause for a moment and ask — how well do we understand AI-ML, or more importantly, what is the difference between the two!
The Mismatched Marriage: AI and ML
Artificial Intelligence, or AI, has been a buzzword since, well, forever. It’s the concept of creating machines that can simulate human intelligence. Think of it as the overarching umbrella that covers everything from your cute digital voice assistant to the futuristic idea of robots with consciousness.
Zoom in a little, and there is ML — the favourite child of AI.
Machine learning — less precisely but more popularly called artificial intelligence (AI) is a subset that involves teaching machines how to learn from data so they can make decisions or predictions. Basically, it’s the magic sauce that powers recommendation systems like Netflix telling you what to binge-watch next or those eerily accurate targeted ads on your social media feed.
“A scientific field is best defined by the central question it studies. The field of Machine Learning seeks to answer the question: How can we build computer systems that automatically improve with experience, and what are the fundamental laws that govern all learning processes?” — Tom M. Mitchell wrote; in case you prefer a more precise and concise definition.
Artificial intelligence, on the other hand, is vast in scope.
“Artificial intelligence is the science and engineering of making computers behave in ways that, until recently, we thought required human intelligence.” Andrew Moore Former-Dean of the School of Computer Science at Carnegie Mellon University.
So, there you have it. AI and machine learning — the tech world’s hottest power couple every leader wants to integrate in their business. Many of them want to be the first to use AI in their respective industry too. But just as anyone, companies need to be careful not to rush into AI.
If you look closely, it becomes obvious that there are some obvious reasons why most companies go wrong in integrating AI.
1. AI Washing
This is an annoying phenomenon where a company claims to be AI driven whereas in reality it is far stretched. A more common example is data analysis, where a company “thinks” that it’s using AI to analyse the data, but in reality, they might just be utilizing rudimentary data analysis techniques to make something more “intelligent”.
A true AI is something that can learn from the data you feed in, take decisions on its own and gets smarter with right type of data it consumes over time.
Not every automation is an AI. MMC Ventures found in their research that a substantial 40% of AI startups in EU lacked meaningful incorporation of AI technology into their operations. This messes with the whole cool image of AI because there’s this big gap between what they brag about and what they’re actually doing. It’s not just a hit on their reputation, but it also confuses everyone about what AI even means in the market.
2. The shiny object syndrome
It is the typical problem of buying the latest gadget without knowing what you’re going to use it for. Some companies do that with AI. They jump on the AI hype train just because it’s trendy, without really knowing how it can actually help them. No wonder, Gartner reported in 2020 that 85% of Machine learning projects fails.
But here’s the deal: wanting AI isn’t enough. You got to figure out where it fits into your business puzzle. It’s pretty pointless to buy some fancy AI thing and then scratch your head trying to find out what it’s good for.
To really make the most out of AI, companies need to stop getting distracted by all the glitz and glamour and start using AI to solve real problems. It’s like going on a purposeful journey to make AI work for you.
3. K — Deficiency & Unprepared Data
In this digital era, most companies generate a lot of data. But huge volume of data does not simply mean that you can throw an ML model into it and pull-out insights from those data. Expecting to get some deep insight from data which are not ML ready is what popularly called as Knowledge deficiency. Just because you have a lot of it, does not mean that they are useful.
Garbage In = Garbage Out
Data must be relevant, accurate, and structured in a way that AI can understand. Using messy, incomplete or biasness in data can lead to catastrophic failures.
Despite all these, it’s hard to deny the incredible progress we’ve made in AI and machine learning. These technologies are infiltrating every corner of our lives, often without us even realizing it. And honestly, some of the applications are mind-blowing. Take financial sector for example, where AI-powered algorithms crunch numbers faster than a calculator on steroids.
But here’s the thing: as we dive into the AI world, we’ve got to clear up some common myths. These myths sound great, but they mess up how we understand what AI can and can’t do. That’s what leads to stuff like “AI washing.”
Myth 1: It delivers 100% accuracy.
No doubt, machines deliver higher accuracy then us humans. But to be more specific, unlike other digital technology or automations, AI only make predictions. And predictions are bound to go wrong. The accuracy of AI models heavily depends on the quality and relevance of the data they are trained on.
If an AI model is trained on data that doesn’t reflect the real-world situations it’s intended to be used in, its accuracy can get compromised. There are two main reasons for that, concept drift and covariate shift.
Simply put, Concept drift means that the underlying assumptions of a trained model become outdated as new data is generated. This can lead to a degradation in the model’s performance because it’s making predictions based on an understanding that no longer holds true.
For example, consider an AI model that predicts stock prices. If the market dynamics change due to economic shifts, geopolitical events, or technological advancements, the model’s predictions might become less accurate over time. Concept drift requires continuous monitoring of the model’s performance and regular updates to its training data and algorithms to adapt to the changing conditions.
Covariate shift occurs when the data used to train the model differs from the data it encounters during deployment. This can lead to a degradation in model performance because the model’s assumptions about the relationships between input features and the target variable no longer hold.
For example, let’s take a model that predicts whether it will rain or be sunny based on the temperature and humidity pretty accurately. Now, you decide to use this model in a different city where the weather patterns are a bit different. In this new city, the temperatures can be much hotter, and the humidity levels might vary in unusual ways compared to where you collected the initial data.
This is a case of covariate shift. Although the relationship between those characteristics remains same, outcome (sunny or rainy) might not be as accurate as the previous location. This is because the characteristics of the inputs (temperature and humidity) have changed due to the change in location.
Myth 2: AI gets better & better over time on its own.
AI algorithms do have the potential to improve over time. However, this improvement isn’t guaranteed, or more importantly not automatic. Improving an AI model requires continuous training with relevant and updated data. Also, the rate of improvement might slow down as the model approaches its performance limits. Moreover, AI models may struggle when presented with data that is significantly different from what they were trained on, highlighting the importance of adapting models to new situations.
Think of it like a program that translates languages. At the start, it’s not great, but it learns from mistakes as more people use it. However, you have to keep giving it new words and sentences to learn from. It gets better over time, but it slows down as it gets really good. Still, if you show it new stuff, it might struggle because it only knows what it learned. So, while it can improve, it’s crucial to keep teaching it new things for different situations.
Myth 3: AI Plug & play.
Deploying AI models is often more complex than simply plugging them in. While some pre-trained models might work well out of the box for specific tasks, they still require integration, optimization, and sometimes even fine-tuning to perform optimally in a particular context. Additionally, understanding the limitations, biases, and potential ethical concerns of the AI system is crucial.
Myth 4: AI can replace humans or think like humans.
AI models don’t truly “think” like humans. They process data using patterns and statistical analysis, which vastly differ from human cognition. AI can be powerful and efficient in specific tasks, but it’s highly dependent on the quality and relevance of the data it’s trained on. If the input data is flawed, biased, or incomplete, the AI’s output will reflect those issues.
A classic example of this is apple card being “sexist” against women applying for credit. The model was trained on historical data where men used to earn more than women. Due to this biasness in data, the model ended up predicting that women are less creditworthy than men!
There are n-number of examples like this. To harness the full potential of AI, companies must escape the grasp of the common myths and embark on an intentional journey toward leveraging AI as a solution-driven enabler. “Knowledge is power.” With careful research, clear-cut goals, and data primed for machine learning, companies can ride the AI wave to smooth integration.
As Thomas Edison put it, “Opportunity is missed by most people because it is dressed in overalls and looks like work.” So, let’s roll up those sleeves, defeat the K- deficiency and make strides toward a smarter future.
