Our Blog

“Knowledge is knowing a tomato is a fruit. Wisdom is not putting it in a fruit salad.”

Categories

  • Categories

Search Blog


Actual Artificial Intelligence (AI): is your vendor on the level?

By Stu Elmes



Here’s the deal: 40of European startups that are classified as “AI companies” don’t actually use artificial intelligence (AI) in a way that is “material” to their businesses. This came as a surprise to many. But not to Dr Neil Yager 

“AI has been a hot topic for the last few years, so it’s not surprising that some companies are trying to ride the coattails by falsely claiming they are using AI. Unfortunately, it’s not always immediately obvious which companies are being honest, and which ones are trying to pull a fast one. My advice is to always ask vendors how they are using AI in their solutions. In fact, it’s even more important to ask why they are using AI, as for some tasks other techniques can actually be more robust and effective.” 

With all the hype surrounding AI in 2019, the urge to present one’s non-AI offering as “AI” has apparently proven too tempting for many companies to resist.  

This is a shame, because there are many vendors operating in the AI space with some extremely effective solutions to some very real problems. And with so many bad actors selling sub-par products/platforms built around automation rather than actual AI, it has become increasingly difficult to tell the actual AI vendors from the AI pretenders (until, of course, you start seeing the results).   

For those seeking to implement the awesome power of AI into their business, there are indeed real, practical solutions to be found, if you know what to look for… 



Actual AI and automation: what’s the difference?

One of the biggest confusion points in differentiating actual AI from other technologies comes from a misunderstanding of the process known as “automation”. Automation handles simple, repetitive tasks that are repeated in the exact same way in the exact same order over and over again. These tasks can be completed by machines with relative ease, often more efficiently and cheaply than human workers. This was the impetus of the second industrial revolution of the mid-19th century. 

While many modern age digital tasks can indeed be completed through simple automated processes, other more complex (but still tedious and time-consuming) tasks have outstripped automation’s simple capabilities. 

Enter AI and its awesome potential to learn how to complete these taskswith increasing efficiencyat a speed much faster than any human could hope to achieve. 

Or, as Dr Neil Yager puts it:  

“I’d say one difference between AI and automation is related to the complexity of the solution. If you can specify the rules for a process unambiguously, it can likely be automated. However, if more complex reasoning is required it might require AI. For example, consider a manufacturing line that assembles toys: 

1. If the same parts are used for each toy, the assembly can be automated. You just need to give instructions on how to build it: (1) take part A and part B, (2) screw together, etc. 

2. Consider another assembly line where you don’t know in advance what parts are going to be available, and the goal is to assemble something completely original. This would require AI since complex reasoning, computer vision, object recognition, etc. are necessary.” 

 

Another example of AI is the production of high-performing marketing language at scale. In this case, automation is simply the wrong tool for the job. Sure, one could simply automate a process whereby a bank of words and phrases are bolted together at random, generating a number of variations of a message that come out largely functional. However, the complexity of human languageand how humans respond to that languagerequires further in-depth analysis and reasoning to get the job done right. 

 

What sets AI apart is itability to analyze the performance of the language it generates, gain useful insights from that analysis, and solve the problem of producing language that achieves the desired result at scale on an ongoing basis



Defining “actual AI”

The simple fact is that concise and comprehensive definition of “actual AI” is difficult to pin down. As Dr Neil Yager puts it: 

I usually say AI is computers doing “smart” things. However, that begs the question “what is smart”? As you dig deeper it gets tricky. One problem is that when AI solves a new problem people often say “that’s not smart, so this isn’t actual AI”. So, the goal posts are constantly shifting. I think the key is not to focus on how it is being done, but on the task at hand. If algorithms are performing actions and making decisions that were previously done by humans, I’d call that AI. 

And therein lies the rub. Since all-purpose “general” AI remains little more than a distant pipe dream for even AI’s most brilliant scientific minds at this stage, for the time being we remain limited to the use of task-specific, “narrow” AIthat which can learn to complete one specific task, and complete that one task with startling efficiency and effectiveness. 

But does AI’s current narrow skillset take anything away from its amazing transformative potential? 

Dr Yager thinks not: 

AI can mean lots of different things to different people. I think that’s the source of many disagreementspeople are talking about different things. For example, philosophers are interested in the fundamental nature of intelligence. Are “intelligent machines” theoretically possible? Is there an algorithm that when run will spontaneously develop consciousness? This is a rich and fascinating area. However, it is of little relevance to realworld AI. The AI that is being used every day is a collection of techniques that have been developed over decades that have been designed to solve specific types of problems. One limitation of real-world AI systems is that they are narrow: they are only designed to do a single task. So, some people will say but that’s not true intelligence! I’m flexible and I can do many tasks!, to which I would say good for you, but that was never the goal in the first place. AlphaGo is a great example: they built a system to perform a task (play Go) and it does that very well (better than any human). That demonstrates the power of AI, and whether or not it can be classified as truly intelligent is academic.” 

The question any company considering investing in AI technology needs to think long and hard about is this: Are any of AI’s current narrow skills something that can help my business make more money? 

And for many companies in 2019, the answer has been a resounding “yes it can!”. 



AI skills currently benefitting businesses:

Machine learning 

Machine learning is a sub-field of AI that gives computers the ability to learn from data. This has widereaching implications and applications. One example is recommendation engines, such as Netflix’s “suggested titles for FIRST_NAME” and the “you might also like” section on your favorite ecommerce website. In general, machine learning has been playing an increasingly vital role in how content is fed to consumers. Machine learning’s ability to track performance and behavioral data at scale in real time has ushered in a new, more profitable era for many of the digital age’s biggest brands. With the vast troves of data now available (thanks to humankind’s insatiable thirst for well, everything the internet has to offer), this fascinating branch of AI presents almost limitless possibilities for any business looking to gain insights into how consumers behave.    

 

Computer vision 

Implementing an AI solution capable of processing and interpreting visual data at scale in a split-second may seem like a strange fit for most businesses, but this impressive AI tool has been making waves nonetheless. From stores that can identify items and ring up purchases without the assistance of human cashiers to cars that can drive themselves, the ways in which businesses may implement computer vision in the not-too-distant future are difficult to imagine. 

 

NLP/NLG 

In 2019, content is king. The endless race to stay ahead of the Google-rankings curve, engage audiences, and keep consumers moving along the digital sales funnel has kept veritable armies of content creators, bloggers, and copywriters busy for the better part of two decades now. But the game is changing. Natural language processing (NLP) and natural language generation (NLG) engines are already hard at work building AI-optimized marketing language that regularly outperforms that written by humans. The brands which have moved forward with this game-changing marketing technology are already reaping massive rewards in the form of increased audience engagement, revenues, and overall marketing ROI.   



Identifying actual AI vendors

So, with so many vendors now touting AI solutions to many of modern business’ most pressing issues, how is one to tell the actual AI vendors from the AI pretenders? 

The process is relatively simple, really. 

First, aPhrasee CEO Parry Malm likes to say, Think about the problem you want to solve, and then think about whether or not AI is an appropriate solution to that problem”  

The fact is, AI may not be the appropriate solution for the issue you and your brand happen to be grappling with, and it’s certainly worth keeping this in mind. However, if you’ve decided that AI is a good fit for the particular problem you want to solve, you’re ready to start talking to vendors!  

First, ask them which specific AI technologies they use in their product. Secondly, ask why they use these techniques. For example, Phrasee uses natural language generation (NLG), as this allows the product to generate original, on-brand marketing language at scalePhrasee also uses machine learning in order to adapt the language it produces to the unique tastes and engagement behaviors of different audiences, as well as improving its performance results over time.  

If the vendor you are investigating has an actual AI solution to offer, demonstrating what they are doing (and explaining why) should be no problem at all. 

Finally, do some investigating. Brands which have worked with effective AI vendors are often all too eager to share their experiences publicly and in the press. What does the press have to say about this vendor? Have their customers become their advocates? 

Keep a skeptical mind, ask plenty of questions, and remember that not every vendor who claims to use AI is telling the truth.