The Branches of Artificial Intelligence
About the above–Raj Ramesh is a Chicago-based IT consultant. He has a very compelling way to make technology easy to understand. This is the first of a few videos that, like this book, can help you comprehend something that is actually quite hard.
Before we start with “artificial,” let’s look at “intelligence”
To start this properly, let’s look at the concept of intelligence in general. There are a number of different definitions of intelligence. Here are a couple of the many definitions from Legg and Hutter’s 2007 study “A Collection of Definitions of Artificial Intelligence”.
First:
“The ability to use memory, knowledge, experience, understanding, reasoning, imagination and judgement in order to solve problems and adapt to new situations.”
Second, a definition targeted to AI:
“… the ability to process information properly in a complex environment. The criteria of properness are not predefined and hence not available beforehand. They are acquired as a result of the information processing.”
Each one is right, individually as well as collectively. In fact, they complement each other. Each understands the context of the world around it and it strives to use what it has learned to adapt to new or not predefined situations.
Between the two, the easier to understand is the better. So,
our definition of artificial intelligence is:
“The ability to use memory, knowledge, experience, understanding, reasoning, imagination and judgement in order to solve problems and adapt to new situations.”
Artificial Intelligence: unpacking the suitcase term
Most people think of AI as a single technology. In fact, artificial intelligence is made up of types, branches, applications and tasks. For our purposes we will examine the three types, three major applications and two tasks—which we will call tools. Tasks sound like reminders. Tools are something your mind immediately sees as useful.
Three types of AI: narrow, general and superintelligence
The three main applications of AI
Machine learning is as simple as the spam filter on your email client, that little piece of code that learns what is and is not important. It can also be incredibly sophisticated, like the math needed to guess the next word you’re going to type into the Google search box.
What are companies doing with artificial intelligence?
Instead of focusing on the financial blue sky of the technology, let’s look into what companies and managers are doing with AI.
Here are the results of a 2018 study of 590 companies and their current state with AI:
Adoption plans for Intelligent Automation (IA) technologies in organizations worldwide as of 2018 | Unsure | No plans | Pilot underway | Moving to production | Implemented and scaled-up |
Smart Analytics (including predictive and prescriptive) | 5% | 14% | 28% | 30% | 23% |
Computer Vision | 5% | 23% | 25% | 28% | 19% |
Machine Learning | 5% | 17% | 31% | 30% | 17% |
NLP (extract, interpret, compare & comply, retrieve and recommend) | 6% | 23% | 31% | 24% | 15% |
AI-“Watson” type reasoning apps | 4% | 16% | 36% | 28% | 15% |
Cognitive/smart virtual assistants (chatbots, etc.) | 6% | 20% | 30% | 29% | 15% |
Robotic Process Automation (RPS) | 5% | 20% | 34% | 27% | 13% |
A 2018 study of marketing managers revealed AI showed value as a tool to improve personalization. Highly personalized content, based on analyzing shopping and public internet records, was the clear leader; followed by business intelligence and marketing automation:
Adoption of artificial intelligence (AI) in marketing personalization according to industry professionals worldwide in 2018 | Using or plan to use | No plans to use |
Personalized content, offers, and customer experiences | 80% | 20% |
Auto-generation of content for personalization | 79% | 21% |
Product and content recommendations | 78% | 21% |
Audience identification | 76% | 24% |
Automation of cross-channel personalization | 72% | 29% |
Delivery of customer activity insights | 70% | 30% |
Email campaign automation | 67% | 33% |
Provide predictive customer service | 62% | 38% |
Customer churn/loss prevention | 61% | 39% |
Automation of chatbot, virtual assistant, etc. | 61% | 39% |
Delivery of sentiment analysis from social channels | 59% | 42% |
Customer self-service | 58% | 42% |
Social marketing automation | 55% | 45% |
Auto-generation of customer/technical support scripts | 54% | 46% |
For use in augmented or virtual reality applications | 46% | 54% |
Automation of IoT personalization | 46% | 54% |
A 2017 study of 1028 marketing professionals revealed an interest in broader use cases:
Adoption of specific artificial intelligence (AI) use cases in 2017, by AI adopter and all respondents | All respondents | AI adopters |
Sales and marketing lead scoring | 66% | 83% |
Sales opportunity scoring | 63% | 80% |
Sales forecasting | 61% | 87% |
Customer service case management | 59% | 83% |
Chatbots for customer service or product selection | 47% | 75% |
Cross-selling and upselling | 51% | 68% |
Fraud detection | 57% | 64% |
Credit risk scoring | 55% | 61% |
Email marketing | 74% | 87% |
We’ve looked at marketing, but what about PR? What are communications managers thinking about AI? Unfortunately, there is very little data like this but from the PR vertical. However, Paul Roezer, founder of the Marketing AI Institute and Cision, an enterprise PR company, has some very good takeaways in his video “The Future of AI in PR”:
Need help contextualizing AI? Use a model, Porter preferably.
We’ve looked at a lot so far. But none of it helps you figure out the how and what AI can do for your organization. You need a frame.
A frame can help you understand the nature of the business problem. With that accomplished, it becomes easier to see how AI tools can be brought to bear on the solution. Porter’s Generic Strategy (PGS) for business success is a popular and easy-to-understand frame. The method was created by Harvard Business School Professor Michael Porter back in 1985.
In upcoming chapters, we will look at the tools of AI through Porter’s frame. It will help you start thinking about AI as a possible set of solutions instead of an opaque suitcase term.
According to Porter, “The two basic types of competitive advantage combined with the scope of activities for which a firm seeks to achieve them lead to three generic strategies for achieving above-average performance in an industry: cost leadership, differentiation, and focus. The focus strategy has two variants, cost focus and differentiation focus.”
In other words, to consistently stay ahead of the competition, ask not what AI can do for your company. Ask what problems your company has that AI can help solve. Let’s take a look at each.
Cost Leadership–According to Porter, “Cost leadership is perhaps the clearest of the three generic strategies. In it, a firm sets out to become the low-cost producer in its industry. The firm has a broad scope and serves many industry segments and may even operate in related industries–the firm’s breadth is often important to its cost advantage. The sources of cost advantage are varied and depend on the structure of the industry. They may include the pursuit of economies of scale, proprietary technology, preferential access to raw materials, and other factors.”
The most obvious example in communication is automated media buying. Robotic spiders and machine learning algorithms team up to understand a market for ad bidding and create a strategy to exploit it. Automation can help lower media costs and reduce cost variances. This technology could be added to your business and become part of a lowest-cost provider story.
Differentiation–Porter’s second generic strategy is differentiation. “In a differentiation strategy, a firm seeks to be unique in its industry along some dimensions that are widely valued by buyers.” Differentiation must always lead to creating a durable price premium. According to Porter, “a differentiator cannot ignore its cost position (in relation to the competition), because its premium prices will be nullified by a markedly inferior cost position.”
Netflix is a good example of differentiation. Between the three major streaming services, only Netflix uses sophisticated machine learning processes to understand every click, swipe and keystroke on their UI. They glean tremendous amounts of information from user behavior to serve bespoke content to users each login. They’ve used this information to develop accurate user affinity groups—people who like action movies, or people who like contemporary teen drama and pop music concert video
AI has helped them develop this expertise to the point where they no longer consider traditional demographics for audience research; instead relying on their user affinities. In fact, the hit show “House of Cards” was a result of machines helping humans understand the pent-up affinity for the original British drama, a deep affinity for Kevin Spacey and an affinity to see David Fincher direct Spacey. These are insights demographics and traditional user analytics alone could never provide. Not bad for up to fifteen bucks per month.
Focus–Porter’s third strategy is focus. Focus is when a company “selects a segment or group of segments in the industry and tailors its strategy to serving them to the exclusion of others. By optimizing its strategy for the target segments, the focuser seeks to achieve a competitive advantage in its target segments even though it does not possess a competitive advantage overall.”
There are two variants under the focus strategy. The first is cost focus, where the company seeks a cost advantage over the competition for the target segment. Differentiation focus is where the company seeks an advantage through differentiation in relation to the competition. Again, the key to remember about these variants is that the focus is on the target segment, not the market in general.
Strategies for success with AI
Having a plan for the proper integration of AI into the systems and processes of a company will be something you hear repeatedly. Companies can’t risk botching AI in the same way they did previous IT projects. AI is expensive and hard but creates tremendous value once in place. Three big reasons not to screw it up.
In the video below, Raj Ramesh quickly and skillfully explains why organizations need an AI framework and strategy to chart the right course with new technology.
Before we move on
The issues that people have with AI are legitimate. I think the idea that “machines are coming for our jobs” is due to not knowing what the technology does. What they need to pay attention to is what could happen if we let AI go the way of the public internet.
A big danger is AI following the internet’s curve and descends into the misogynistic, hate-filled, monopolistic practices we see today on the web. The other danger is with biased algorithms. Badly written algos used by Chicago police collects inaccurate data used to make decisions on who to arrest. The Beauty.ai algorithm, created with the noble intent of measuring beauty without relying on color, was doomed from the start. The programmers didn’t include enough examples of beautiful people of color, so the machine eventually ignored women who weren’t fair of complexion.
Unfortunately, it will take legislation and standards at a global scale; much like GDPR. It will have to be monitored closely, and violations have to be punished. And it will have to have some sort of independence from the cycle of politics. Imagine a future Putin, Trump or Steve Bannon with the ability to use even more powerful technology to further their aims.