Intelligence in the Tech Stack
AI might make a quantum leap, but enormous value can already be unlocked by normal businesses using these models now
Welcome back to the Merchant of Arrakis! Follow me on Twitter @jack_forlines
I’ve been fascinated by the idea of applying new technological frameworks for solving challenges elsewhere. For example, Elon did not invent vehicles, rockets, or internet finance, but rather applied new ideas to established industries ripe for disruption, thus creating new solutions. Where else could this framework be applied?
I read recently that the most significant advances in the next decade won’t just be groundbreaking new technologies that disrupt systems. While such leaps will almost certainly occur eventually (like the internet, blockchain, and artificial general intelligence), the more immediate purpose such advances will serve is revitalizing existing industries. An electric car wasn’t a novel idea, it was just something people thought the technology wasn’t good enough for yet, and so it wasn’t worth pursuing. Musk decided that wasn’t true, and now almost every car manufacturer has plans to introduce electrically-powered vehicles.
The applications of this approach are myriad, and I’m sure many people smarter than I have written excellent books on this. But I think it is worth a discussion in regard to deep learning and neural networks.
The past several decades have seen enormous growth in AI, and exciting technologies like backpropagation, natural language processing, and GANs are already changing the way businesses operate (Cade Metz’ new book Genius Makers is a fantastic brief on this topic). When the AI revolution hit Silicon Valley, all of the big tech firms put the weight of their resources behind building out AI labs. Each has seen enormous success in applying these frameworks to their existing businesses. Gmail can finish sentences for you, and Facebook’s ads are better at targeting consumers’ desires than ever before. These successes can be entirely attributed to applying the power of neural networks to their existing tech stacks.
What is even crazier to me is that these technologies are, in most cases, completely open-source. Anyone who has a basic understanding of Python can watch a YouTube video and apply a machine learning model to their work with just a few lines of code. Which is why I believe significant value creation of AI might not just come from the field’s most brilliant individuals, but rather from an everyday programmer taking a new approach to a problem using a neural network. Google, Facebook, Baidu, and many more companies are spending billions researching AI, but interestingly, they are making the fruits of those labors open for anyone to use.
Financial analysts searching for signals among mountains of data can feed that information to a neural network, creating a model that makes this process exponentially more efficient. Product Managers can apply the same framework to unlock new insight among the metrics data that helps drive their product planning. These are some examples of how AI can already be leveraged, but the mere fact that they are possible serves to underscore the main point: you’d have needed a data center and thousands of GPUs to train a single model to perform a single task, but now anyone can import these models with a few lines of code. Open-source projects like TensorFlow and PyTorch, which are the results of years of research and improvements by the field’s leading minds, are two models that you can use yourself right now.
I don’t pretend to be anywhere close to deeply knowledgeable in this field – I’m a novice programmer with a little curiosity. But this blog’s core purpose is to attempt to bridge complicated ideas to the layman (I do my best), and you don’t have to be technical to recognize this point. It’s obvious to anyone who is even slightly inclined to read a sci-fi novel that AI is going to be big. But I think its worth the time to consider that as the field continues to push the boundaries on how well machines can think, the possibilities for normal businesses to leverage these advances expand exponentially.
Driverless cars, virtual assistants, and ad targeting are just a few of the categories that are already leveraging the technology, but soon these frameworks might be used by any business that accumulates data. Data is everything, and where there is data, there is almost always machine learning model that can be applied to unlock new value. I think that this space will have a far more profound impact than the hype around it already suggests, simply because it can be leveraged by anyone.
How did you like this post? Leave a comment and let me know.
Follow me on Twitter @jack_forlines