But the two are the same, aren’t they?
Some time back, a few friends of mine built a sleek sentiment analysis tool. The tool accesses a stream of tweets around certain keywords, and it can tell you the tone, direction and sentiment of the tweets — along with the impacts these tweets or tweet-storms are having. Such systems have existed in the past as well, but yeah — I have had a look at their system, and I got to say, its much more powerful, robust and accurate in its analysis than any of the other tools I have seen in past. Neat, right? Maybe. If only they can answer one question for me — So??
I like products, in general. Products are simple. Like a Mayfly, they have one singular purpose, and good products perform that function really really well.
Like a Mayfly, products MUST have a really well defined and quite distinct function they are there to perform, and it has got to be the first thing that you understand as soon as you look at the product. If your product is missing that, then you need to rethink the product model.
Getting back to the sentiment analysis system my friends had built, my problem with it wasn’t with the efficacy or the accuracy of it, but with the fact that I couldn’t figure out how does it help me, as a business?
So, I started with exactly that. I asked them this question, and they gave me what I can call a satisfactory answer. But now, I had a new problem statement for them. Why did they need to explain it to me? Should it not have been apparent from browsing through the system itself?
You would have noticed that I keep on calling it a system. Because that’s what it is, in my opinion. It is a system, an application of technology, amalgamation of data science, machine learning and content feeds. It is definitely NOT A PRODUCT! And if you don’t have a product, in all probability you won’t have the consumers either.
Products, as I mentioned earlier, need to be intuitive in nature. Products are meant to deliver value. They are meant to solve bottlenecks. And products need to carry all the business value they can deliver right on their sleeves. That’s what makes a product usable. That’s what drives the adoption of a product. That’s what makes it mainstream.
In my opinion, their system missed out on those. It stopped possibly as low as one step short of being a product. Maybe it was just one, but it was probably the most crucial one.
THE PROBLEM WITH ALL THINGS TECH
This — what I have just highlighted — is possibly the biggest problem with any tech system, including those based on AI and machine learning. We need to understand and respect the fact that having a sound technology behind our ‘quite valuable’ systems does not necessarily make it a product.
Look at any of the products you use for your businesses — be it google analytics, Facebook ad manager, Google adwords. What makes them so meaningful is how robust they are and how much value they add to our businesses. But what makes us use them so widely is the simplicity and ease of use behind each of them. Would you use Facebook ad manager if the audience targeting it offers just didn’t work? Probably not; and you would certainly be furious about how broken the system is. But would so many people have been using the audience segmentation tool if it wasn’t intuitive and simple enough to use? Definitely not!
Presenting a product in a useable and intuitive fashion is almost as important as how well the product does what it is supposed to do.
ACADEMIC AND BUSINESS APPLICATIONS ARE DIFFERENT
For a large part of the time-period for which AI has been in play, it was restricted to the field of academia. People would be publishing research papers about it, breathtaking computer simulations were created. A lot of it was therefore intended for consumption of the academic audience.
Now, AI has moved from being purely academic to finding applications in the business world. And that completely changes the audience profile. The product needs to be simple and intuitive enough for everyone to use. The value the product imparts should directly be aligned to the business objectives. (If you are not sure of what the business objectives of any business is, start from the basics — increasing revenue, decreasing cost, decreasing cost/head, decreasing cost/second.)
Big businesses are generating terrabytes of data every hour of every day, and the need to handle this data more efficiently and present it in a useable fashion is the reason behind this tectonic shift in AI and machine learning from academia to the world of business.
Realising the inevitability of the rise of applications of machine learning in business processes, big tech companies even started releasing an increasing number of open source packages to help startups and individual developers. Now developers can build deep learning based products faster, without having to invest millions of dollars in research as was done by the IBMs of the world.
The result? We are seeing the rise of more and more startups claiming to be building products centered around deep learning. But as long as they fail to understand the difference between a tech product and a tech package/system, many of these startups will fail to see the adoption they so deservedly are worthy of.
So, what are you upto? Are you working on some kickass technology, or are you building a product people would line up for? I really hope you understand the difference between the two now.
That’s it for today; see you tomorrow!
I am Abhishek. I am here... there.... Everywhere...