Be it the core of their product, or just a component of the apps they use, every organization is adopting machine learning and AI at some level. Most organizations are adopting it an ad hoc fashion, but there are a number of considerations—with significant potential consequences for cost, timing, risk, and reward—that they really should consider together.
That’s why I developed the following framework for organizations planning to adopt machine learning or wanting to take their existing machine learning commitment to the next level.
Part II of The Completely Non-Technical Guide to Machine Learning and AI
My previous post raised the question “what is machine learning and artificial intelligence (AI)?” and answered with a functional definition: computer systems that combine measurements and math to make decisions so complicated that until recently only humans could make them.
Now a new question: “What’s the difference between machine learning and AI?”
Machine learning is usually considered a subcategory of artificial intelligence. It’s artificial, because we’re talking about machines (computers) instead of people. And intelligence, because unlike ordinary computer systems which can only follow exact instructions, with machine learning computers can learn from examples (they can be trained) to do things we associate with intelligent behavior, like translating words from one language to another, recognizing people and animals from photographs, driving cars, and having conversations (via chat bots). Machine learning can not only mimic human behavior, to some extent it attempts to simulate the human brain: neural networks—one of the workhorse methodologies of machine learning—were “inspired by” and named after brain neurons…even though they work in quite a different way.
Probably the most exciting recent example of machine learning “intelligence” is an advanced game program called AlphaZero developed by Google’s DeepMind team. AlphaZero used machine learning to rapidly teach itself how to play the complex strategy games chess, Go, and shogi (a Japanese version of chess). Older computer programs designed to play games like these (for example, IBM’s famous Deep Blue, the first to beat a world champion human chess master) were mostly limited to choosing from pre-designed strategies already spelled out by human experts. But AlphaZero learned to play by playing against itself, without mimicking human players. In a matter of weeks AlphaZero became skilled enough to beat the best human players and the best computer chess programs. Along the way it invented winning strategies that humans hadn’t discovered despite thousands of years of game play.
And unlike previous generation computer games programs which by necessity must specialize in just one strategy game, because it’s learning-driven rather than instruction-driven AlphaZero was able to become expert in multiple games with only a few tweaks.
Having said all of this, machine learning and AI are not completely interchangeable. AI is much broader than just machine learning. (Pro tip: using the term AI to describe machine learning actually annoys some data scientists, who dislike the distractions that come with the hype and fiction surrounding AI.) So, how are they different? Let’s look at a rough diagram:
Machine learning rests at the intersection where data analytics—the computer software for number-crunching, from Excel spreadsheets on up—overlaps AI. Machine learning stands out as a current success story at that intersection. But not everything involving computers that people have described as AI also involves machine learning. Deep Blue, for example, does not use machine learning.
AI is a broad label. Stories and speculation about artificial intelligence go back at least as far as Greek Mythology (Talos, for instance, was supposed to be an “automaton” constructed of brass). In recent times, the 60+ year history of scientific explorations of AI didn’t start with what we now call machine learning, and might not end with it.
By now you’ve noticed an additional term, “AGI”, in a dotted circle on the chart, also at the intersection of data analytics and AI. AGI stands for “artificial general intelligence.” Researchers, futurists, and storytellers focused on AGI are looking forward to a human-like intelligence, an AI capable of self-awareness, rapid, independent learning and the ability to transfer knowledge between situations like humans do. It’s important not to confuse AGI with currently available machine learning. Although an impressive amount of advanced research is in progress to expand the range of AI, AGI doesn’t actually exist at present except in fiction (and arguably we aren’t even close…yet). Many people think that current machine learning techniques such as neural networks, as powerful as they are, can not be used to create an AGI, but rather that as-yet-unperfected or unimagined techniques will be used instead.
However, the present is an excellent time to start talking seriously about AI ethics. For starters, can we allow AI (whether or not it’s an AGI) to make life or death decisions for us without direct oversight—such as decisions made by self-driving cars and automated weapons? This will be the topic of a future post in this series.
Part I: What is Machine Learning? Combining Measurements and Math to Make Predictions
The labels “machine learning” and “artificial intelligence” can be used interchangeably to describe computer systems that make decisions so complicated that until recently only humans could make them. With the right information, machine learning can do things like…
• look at a loan application, and recommend whether a bank should lend the money
• look at movies you’ve watched, and recommend new movies you might enjoy
• look at photos of human cells, and recommend a cancer diagnosis
Machine learning can be applied to just about anything that can be counted/measured, including numbers, words, and pixels in digital photos.
Takeaways from Sam Charrington’s May, 2017 interview with Jennifer Prendki, senior data science manager and principal data scientist for Walmart.com
I am very grateful to Sam Charrington for his TWiML&AI podcast series. So far I have consumed about 70 episodes (~50 hours). Every podcast is reliably fascinating: so many amazing people accomplishing incredible things. It’s energizing! The September 5, 2017 podcast, recorded in May, 2017 at Sam’s Future of Data Summit event, featured his interview with Jennifer Prendki, who at the time was senior data science manager and principal data scientist for Walmart’s online business (she’s since become head of data science at Atlassian). Jennifer provides an instructive window into agile methodology in machine learning, a topic that will become more and more important as machine learning becomes mainstream and production-centric (or “industrialized”, as Sam dubs it). I’ve taken the liberty of capturing key takeaways from her interview in this blog post. (To be clear, I had no part in creating the podcast itself.) If this topic matters to you, please listen to the original podcast – available via iTunes, Google Play, Soundcloud, Stitcher, and YouTube – it’s worth a listen.
Jennifer Prendki was a member of an internal Walmart data science team supporting two other internal teams, the Perceive team and the Guide team, delivering essential components of Walmart.com’s search experience. The Perceive team is responsible for providing autocomplete and spell check to help improve customers’ search queries. The Guide team is responsible for ranking the search results, helping customers find what they are looking for as easily as possible. Continue reading “Lessons in Agile Machine Learning from Walmart”
Summary:Every organization that processes data about any person in the EU must comply with the GDPR. Newly published GDPR Guidelines clarify that whenever an organization makes a decision using machine learning and personal data that has any kind of impact, a human must be able to independently review, explain, and possibly replace that decision using their own independent judgment. Organizations relying on machine learning models in the EU should immediately start planning how they are going to deliver a level of machine model interpretability sufficient for GDPR compliance. They should also examine how to identify whether any groups of people could be unfairly impacted by their machine models, and consider how to proactively avoid such impacts.
In October 2017, new Guidelines were published to clarify the EU’s GDPR (General Data Protection Regulation) with respect to “automated individual decision making.” These Guidelines apply to many machine learning models making decisions affecting EU citizens and member states. (A version of these Guidelines can be downloaded here—for reference, I provide page numbers from that document in this post.)
The purpose of this post is to call attention to how the GDPR, and these Guidelines in particular, may change how organizations choose to develop and deploy machine learning solutions that impact their customers.
What’s so important about interpretability in machine learning?
It’s a poorly kept secret that we lack insight into how complex machine learning models like neural networks make decisions. We see the data that goes in, the math that goes in, and the results that come out. But in the middle, where we want to see a chain of reasoning like a human could give us to explain decisions, there’s only a black box. Neither data scientists nor these complex machine learning models can provide insight into “why” a model chose output A rather than output B.
What does it matter whether we have an understandable explanation for why a machine learning model delivers a specific result? For example, when diagnosing whether or not a patient has cancer, isn’t it enough that the model is accurate, according to rigorous testing? I’ll look deeper into the implications of interpretability in future blog posts. But Continue reading “Understanding Decisions Made By Machine Learning”