AI for business: deep learning

As a broad concept, AI is useless in the business world. Its applications and potential are theoretically vast, but real, tangible use cases have been hard to come by. In fact, while it seems like everyone is talking about AI, only 20 percent of companies claim to have a deployed AI technology in a core part of their business, or at scale. And this is by AI’s broadest definition, meaning that a customer-service chatbot on the homepage counts.

But slowly, that’s changing. Companies with deep pockets are dreaming up ways to apply machine learning, a subset of AI, to greater business challenges. With machine learning (or data-driven AI), a model (the brain of the AI technology) is trained by the input of vast datasets. By studying a large volume of data to detect patterns and learn from them, machines can do things like:

  • Predict likely outcomes or actions
  • Identify unknown patterns or relationships
  • Detect out-of-the-norm behavior patterns that signal issues
  • Understand text, images and sounds

Naturally, businesses across many industries are looking to create their own AI use cases using increasingly abundant big data. Unique applications are starting to emerge. Ride-sharing companies Uber and Lyft use machine learning to predict the arrival of our car and decide what price to charge us. Banks use it to decipher fraudulent checks. Airlines use it to fly planes (the original self-driving vehicles).

In the process of navigating to these words on your screen, you almost certainly used AI
— Gautam Narula, Tech Emergence

Blue-sky machine-learning dreams include things like better prediction of supply and demand, optimized retail trend forecasting, monitoring of the individual human health condition — even early detection of disease that might otherwise be missed by a medical doctor not looking for certain signs and symptoms.

These are real-world applications that won’t just drive better business outcomes. They’re in line to better humankind. It’s very exciting. And the key to unlocking AI’s potential lies in a subset of AI called deep learning, which can handle both structured and unstructured data. Deep learning, in fact, is the main reason that AI has now entered the lexicon of mainstream business. Based on artificial neural networks, deep learning is inspired by how the human brain itself operates.

Why deep learning is different

  1. This technology can handle many different data types simultaneously - numbers, images, text, etc
  2. It's scalable to the amount of data and processing power: increases in data and processing lead to better results
  3. It's good at solving problems that require different levels of abstraction, which is something humans are very good at and computers have traditionally been very bad at
  4. And it's extremely resilient when unexpected changes occur - things like changes to data input or corruption to data

Today’s deep learning models are big, powerful and scalable. By making sense of data, deep neural networks can find patterns in extremely complicated datasets — things like images, speech samples, audio and video. Most of the intelligence that a rarified data scientist once had to build manually can now be handled automatically by the model, and the results are typically much more accurate, and faster. Models are now much more robust and capable, and that’s good news, because data scientists are hard to find. Within “the industry”, it’s common knowledge that there’s both a high demand for talented, experienced data scientists and drastic shortage of such talent.

The bad news? You may not need a data scientist to implement a deep learning project, but you do need an AI expert. And there are even fewer of those. Even if by some miracle you do manage to land and afford an eminent AI expert for your team, the challenge doesn’t end there.

The even worse news? There are a lot more deep-learning obstacles in your way than just the lack of a good AI expert on your team.

Let me introduce you to a construct I like to call “The 7 Levels of AI Purgatory”. This is the endless suffering most AI seekers find themselves in when they launch into an ambitious deep-learning project.

References

  1. 01/ Everyday Examples of AI  — Tech Emergence
  2. 02/ How Artificial Intelligence Can Deliver Real Value to Companies  — McKinsey
Get started for free