Last week I was watching The Big Short (sidenote — a brilliant movie!) which depicts the events leading up to the global financial crisis from the perspective of a few people who realised the bubble was about to burst.
What was interesting to me was that much of the financial instruments that ultimately led to the downfall of the system weren’t actually created as evil. But by cutting corners and people without the knowledge or incentive to do their homework eventually they ended up ticking time bombs. It’s a compound effect where small choices in the moment don’t seem significant but over time exponentially grow to become a massive capital (or debt).
Stock market bubbles don’t grow out of thin air. They have a solid basis in reality, but reality as distorted by a misconception.
I also realised that our startup Lead Wizards is in an increasingly hot space, that of Artificial Intelligence. Now this isn’t a bad thing per-se. It means investors, competitors and potential partners take us seriously which opens up capital and opportunities that might otherwise have not. It means it’s relatively easy for us to gain media coverage or SEO traffic.
But there’s a flip-side…as the discussion around AI becomes more mainstream we’re going to see a lot of unfounded, incomplete or wrong “facts” on the matter. So in this post I wanted to address some of the misconceptions I’ve come across and show what “AI” really means and to make sure that we’re not over-inflating the promise of AI companies based on wrong ideas of what it can and can’t do.
In the next post on this subject I want to give a honest look behind the scenes of a lot of AI companies (including ours) and how their “fake it till you make it” culture helps AI startups thrive in the long run.
But let’s start with a few AI misconceptions.
First of all, let’s stop using the term AI for everything! The “Hollywood” AI or Artificial Intelligence is a true super-human brain running on silicon. A brain that can think, reason, dream and most importantly improve itself. And using this term often paints the wrong picture. Because most of the “AI” that we see around us today is actually Machine Learning. Mathematical expressions operating on data to obtain “statistical” insights…nothing spooky!
Even advanced systems like the IBM Watson machine which won Jeopardy aren’t really AI. If we look on the IBM website we see the following description of what they actually do:
“IBM Watson is a technology platform that uses natural language processing and machine learning to reveal insights from large amounts of unstructured data”
In other words, they are a set of algorithms whose parameters have been trained to give structured data (so think detecting years, names and nouns and putting them in a database and linking these concepts together) from a chunk of unstructured text. Now it is true that in some Machine Learning fields such as neural networks do draw a lot of inspiration from how the brain actually works. But calling that AI would be the same as calling an airplane a bird. Yes the airplane has been optimised to be great at flight and transporting load and is in fact better than a bird. Yet if we ask it to survive on it’s own it would fail miserably (in fact it wouldn’t even register the request :P ) What’s important to note here though is that if there is a symbiosis between man and machine great things can be achieved! By leveraging the scalable, continuous and emotionless qualities of our machine friends and combining them with our emotions, imagination and versatility.
Now we are moving towards true AI at a rapid pace! Wether that’s a good or bad thing is another discussion entirely but the fact that Elon Musk has set foot in the AI research field should give some indication as too the big changes ahead of us. But remember; not yet! Just because somebody runs an “AI” company doesn’t mean they’re going to destroy the world. Nor does it mean that they can help you turn lead into gold, which brings me to my next misconception…
Now another misconception I heard is something along the lines of “We’ve been saving data for years and have now used this great machine learning algorithm on our big data and we have concluded that <fill in mumbojumbo here>”.
There’s two things wrong with this sentence. First of all it’s the notion that just because you have a lot of data there is useful data. You might have been collection the wrong data, biased data or highly random data. Without an understanding of what the data contains, how it was obtained and some digging for a useful signal it’s easy to use the wrong algorithms and draw faulty conclusions.
The second error is the belief that there’s just this simple algorithm or AI system you can use. I think that especially with the rise of “bubble AI startups” this argument is going to be heard more and more. Yes there are great toolkits and algorithms out there but they need to be combined with performance benchmarks and AB split testing (cross validation in ML terms) and a good understanding of the pro’s en con’s and application areas to actually deliver value. And they’re not just algorithms you run once, rather you keep optimising and improving as new data comes in. That’s why it’s critical that alongside your “data” you also build a validation loop. I’ll elaborate on this point more in the next post but it’s very similar to how you validate your Core Metric as described in this article by Hampus Jakobsson
Last but not least, misconception number 3…
Again I think this is something we’re going to hear more an more. If investors (or worse founders) lack understanding of how they use Machine Learning to actually deliver value they will start to simplify and generalise to a point where every AI company is thrown on a big heap.
Now the problem of-course is that most AI companies aren’t actually all that alike. Some might focus on Natural Language Processing, others focus on data clustering. There’s also a big difference in how these companies deliver value and how they can be monetised and grow. Shivon Zilis wrote a great article on different types of Machine Learning companies and as the space develops we should be starting to see more stereotypes, not less!
If you are interested in AI companies or are thinking about investing in them please make sure to do your homework. Even if you’re not technical there’s great books (like this one) to help you at least understand enough to have a feel for when things are off (or right ofcourse!). And I really do believe that we’re going to see some amazing new companies delivering value in unique and exciting ways. And naturally we’re working hard ourselves to make Lead Wizards one of them!
In the next post I’ll talk a bit more about how a lot of AI companies choose to get started and how this, seemingly deceptive, process is actually essential to understanding the problem and delivering real value in the long run.
— Rik Nauta, CEO Lead Wizards
Create your free account to unlock your custom reading experience.