On Google’s recent Q3 earnings call, Google’s CEO, Sundar Pichai said that one “transformative” technology is causing Google to rethink “how we’re doing everything.”

Read that again. There’s a single technology that’s causing Google to rethink they way it does everything.

And it’s not just Google.

The same technology is in the process of transforming many of the biggest names in tech — Facebook, Amazon, Netflix, UBER, Twitter — not to mention smaller, up-and-coming startups.

Entrepreneur and thought leader Peter Diamandis say that it will “do more to improve healthcare than all the biological sciences combined” and will generate large amounts of wealth and abundance.

Billionaire venture capitalist Vinod Khosla agrees, saying that over the next 50 years, it will drive abundance, transform industries, and impact almost every part of society.

What is it?

Machine learning.

What is machine learning

You’ve probably started to read about about machine learning in the popular press and technology press. Media outlets like TechCrunch, New York Times, and Forbes have been writing about it over recent months and years.

That said, it’s not always discussed under the name “machine learning.” In some outlets, you’ll hear about “statistical learning.” (This is less common in the popular news. “Statistical learning is more commonly used in academic circles.) To be fair, there are minor differences between statistical learning and machine learning, but as far as the common ML practitioner is concerned, they are basically the same thing.

Machine learning is also discussed indirectly. You might read articles about closely related fields, like machine intelligence, artificial intelligence, predictive analytics, and automation. To be clear, these are not all identical to machine learning, but they strongly related. Machine learning is a key component of all of these areas.

This is important to keep in mind: machine learning is an area of study in and of itself, but in an applied setting, it has wide-ranging applications that cut across various parts of business and software.

Machine learning, a quick and dirty definition

So what is machine learning?

Machine learning is using computation and algorithms to enable computers to learn from and make predictions about data (source: wikipedia article on machine learning).

From the standpoint of a practitioner, machine learning is is a set of tools for writing programs that input data and output predictions or decisions.

 

Software that learns: the key to unlocking the value of Big Data

Currently, most software does not adapt. It does not learn. It has no intelligence and only limited ability to change how it works to meet the needs of the user.

In contrast, programs that incorporate machine learning can adapt. They can learn over time, and improve.

The process that machine learning algorithms use to “learn” is analogous to how a human learns through examples. To be clear, this is a simplification, but an instructive one.

When a human learns by example, they observe the examples, make generalizations about the examples, and then apply those inferences and generalizations to new observed data.

What’s critical in the process is the data. Data is required to provide training examples from which to learn.

data-center_machine-learning-at-google-blog-post_600x398

 
Although this is a large simplification, machine learning (and specifically, supervised learning) works in a very similar way: observe training examples, make generalizations, and then apply those generalizations to new observed data.

The main takeaway here, is that the key component in getting machine learning algorithms to learn and operate properly is data. In fact, some of the most powerful machine learning techniques not only require data, but very large amounts of data.

This is critical for understanding why Google is “rethinking everything.”

Software that predicts and adapts

Once trained, the machine learning algorithm (and again, specifically, a “supervised learning” algorithm) can make predictions.

We may not recognize this, but much of what human beings do involves a prediction of some sort. Simple, everyday tasks like picking up an object involve a variety of predictions, like “how far away is the object” and “how much force do I need to apply” to pick it up. These predictions happen below our level of conscious awareness, but they do occur.

Similarly, in the task of driving a car, humans make predictions about terrain conditions and about the outcomes of various directional changes.

In more complicated cognitive tasks, even decisions can be boiled down predictions; when we make decisions we’re choosing the best of several options, where each option has it’s own predicted outcome. For example, when we decide to watch one movie over another, we are, in some sense, implicitly making predictions about our satisfaction with the available options, and choosing the best one.

And it’s certainly not just inconsequential decisions like choosing the best movie. Effectively, every important decision we make is underpinned by some type of prediction. We make decisions about how to spend money, how to invest, and how to allocate our time, and how to select people as friends and team-members. Good decisions (and therefore, good predictions) are required for a variety of tasks with important consequences.

Good predictions are valuable

If you look closely, you can see that in a wide variety of areas, from driving a car, to choosing how to allocate resources, good predictions are extraordinarily valuable.

And increasingly, where there’s a prediction to be made, there’s a machine learning algorithm that can automate it.

To the extent that these predictions can be automated and optimized, machine learning algorithms – powered by data – can produce tremendous value.

So, to recap where we’re at, machine learning is powered by data. Good data is required for training machine learning algorithms. But at the same time, the machine learning algorithms, provide tremendous value by make and automate decisions.

In some sense, machine learning unlocks value from data.

What that also means, is that machine learning will become more valuable as data volumes increase.

Again, this helps us understand what’s going on with Google. It helps us understand why Google is “rethinking everything” because of machine learning.

Google is sitting on more data (and better data), than almost anyone in the world. Data is necessary for training machine learning algorithms, which in turn enable adaptive software. And it’s this adaptive, intelligent software that produces massive value.

 

Why machine learning will become even more valuable in the next decade

Having said that, for companies like Google, it’s not only about the data that they currently have, but the data that they will have in the future.

The estimates vary, but common estimates state that the amount of data that humanity generates is doubling about every two years.

Given this exponential increase in data, by 2020, it’s estimated that the world will have 44 Trillion gigabytes of data.

Unfortunately, our human intuition breaks down when trying to conceptualize volumes this large (and rates of change this fast). What we can say, in very simplistic terms, is that data is growing very, very fast and reaching huge volumes.

Sensors, exponential data, and connecting the physical & digital worlds

This explosion of data is being driven not only by the internet itself, but also the the emerging connection between the virtual and physical worlds.

That sounds a little abstract, so let me unpack it.

Sensors are becoming smaller and less expensive. As they become smaller and less expensive, companies are adding a wide variety of sensors to physical objects. The current example of this is mobile phones. Mobile phones are instrumented with a variety of sensors that enable the phones to collect data.

Just like in mobile phones, as sensors get smaller we will be able to add more sensors to everyday objects. We’ll be able to instrument more and more of the physical world. Effectively, this is what people mean when they say “the Internet of Things.” The IoT isn’t so much an internet of things, but an internet of sensors, data, actuators, and intelligent software.

Setting aside the large topic of the IoT (that’s a different article) the point here is that data is exploding, and will continue to explode as sensors miniaturize and we begin to connect the physical world with the digital world.

And again, the value of this data (and really, the value of the IoT) will largely come from machine learning and intelligent software.

Software is eating the world, and machine learning is eating software

In some sense, this is the ultimate meaning of Marc Andreessen’s thesis that “software is eating the world.”

Back in 2011, the famous entrepreneur and venture capitalist wrote an article for the Wall Street Journal making the bold claim that “Software is Eating the World.”

Looking backwards, and seeing the rise of companies like AirBnB, Uber, not to mentioned the continued success of Amazon and Netflix (as they dominate their industries and outcompete their former brick-and-mortar competitors), it seems that Andreessen was absolutely correct. Software is being built into the DNA of the most successful companies of our time. Software is critical to the success of the companies who are disrupting old industries, and winning.

But a large part of what makes the software at these companies so valuable is machine learning. In many of the most successful software-driven companies, machine learning is part of the “secret” that drives their success.

Take Amazon. Like nearly all businesses, what drives revenue at Amazon is very simple: increased customers, increased order sizes, and repeat buys. At every step of this process, Amazon uses machine learning to optimize. For example, to increase order sizes, Amazon uses machine learning to recommend “similar items” as you browse and make your purchase. It also uses similar techniques to target past customers with new “recommended” offers.

If we look closely, Amazon is really just a data driven marketing machine that sells books. (well, to be clear, Amazon sells everything.)

Similarly, Netflix (who uses very similar machine learning techniques) is a data driven marketing machine that sells movie rentals.

In both cases, data is the fuel, and machine learning is the critical part of the “engine” of those machines. And software sort of wires it all together.

Yes, these are software-driven businesses. And we can say that they are data-driven businesses. But what entrepreneurs and software engineers alike need to understand, is that in a very critical way, these are machine learning businesses.

And that’s only the the initial part of the story.

As we instrument the world, and collect larger amounts of data from everything, we can build software for the physical world that’s powered by that data; software that interacts with the world (i.e., robotics) and also software that digitizes and optimizes formerly physical services (like Uber and AirBnB).

As we collect more data from the world, we’ll be able to use machine learning to create better software that predicts, decides, optimizes and adapts. We’ll be able to build more valuable software.

Yes: software is eating the world, but machine learning is eating software.

 

Why software engineers and entrepreneurs need machine learning

This then is why Google is transforming everything it does because of machine learning. Machine learning will allow them to unlock the value of big data.

In any industry that Google (or Alphabet) is in – search, mobile, the IoT, robotics, and even healthcare – it will allow them to create software that unlocks the value of their massive datasets.

Machine learning will likely create massive amounts of wealth for companies like Google and will have a large impact on almost every part of society: transportation, healthcare, communications, marketing, to name a few.

To quote billionaire investor Vinod Khosla (one of the heavyweights of the venture capital world): “I think the impact of machine learning on society will be larger than the impact of mobile … Almost any area I look at, machine learning will have a large impact.”

Ultimately, the pervasive impact of machine learning on business (not to mention, society itself) is causing investors like Khosla to invest.

So Google is rethinking everything because of machine learning. Investors like Khosla are investing heavily, claiming that machine learning will have a large impact on almost everything.

Here’s my claim: if you’re a software developer or an entrepreneur, you need to learn and leverage machine learning.

It’s not just a transformative technology that is causing Google to rethink everything. It’s a transformative technology that will cause software developers and entrepreneurs to rethink everything (whether they want to or not).

Software businesses will increasingly disrupt traditional industries. And intelligent software will increasingly win out over “dumb” software.

Applications that learn, adapt, optimize, and automate will increasingly win.

Moreover, data-driven businesses that leverage machine learning and develop systems to optimize and automate marketing, hiring, and operations are going to out-compete those that don’t.

Google understands this. Investors like Khosla understand this. You need to understand it too.

 

If you want to learn more, sign up now

Here at Sharp Sight Labs, we’re going to be writing quite a bit more about machine learning in coming months.

We have content planned for machine learning in entrepreneurship, machine learning in R, more general machine learning tutorials, and articles that explain the relationship between machine learning and the broader data-science skill set

I have two requests:
1. If you want to learn more about machine learning, then sign up now for the email list. That is the best way to make sure you receive all of this content.

2. Leave a comment explaining your largest challenge with machine learning. What are you struggling with? What sort of content would help you?