I asked Wolfram Alpha to paint for me the face of 50 Cents, by computing it, and this is what I got:
Math indeed can translate into art, in fact, this portrait has been plotted by Wolfram Alpha, using a very long parametric equation, which looks like this:
This is only a slice of the equation, as it goes on much much longer.
And I know, it looks like gibberish.
Parametric equations, though, are very useful, because they can help us graph the complete position of an object over time.
Well, it’s to show you that there are various forms of engines on the web.
Why am I telling you this?
Wolfram Alpha is a computational engine. In short, instead of giving you data, based on a ranking system, Wolfram Alpha uses a computational system.
What does it mean?
Take the case of Google, the father of all search engines.
Google follows three steps, before providing you a
- Crawling: Google uses robots that from time to time, go on web pages, to discover content on the web. Those crawlers are looking for many, many parameters. Those crawlers also called “spiders” look for hundreds of rankings factors and thousands of parameters. At this stage, the role of the crawler is to start understanding what might be relevant, to include in Google’s index.
- Indexing: at the indexing stage, Google starts to build a corpus of content, which might or might not be served in its ranking. While at the onset Google indexed pretty much the whole visible content on the web (the content that could be crawled), over time it stopped. As you can imagine this makes sense. With almost two billion websites on the Internet, prioritizing those that add more value, to its index, is critical, to keep it viable, in the first place.
- Ranking: once crawled, and indexed, Google builds up its ranking system in various ways. For most of its life, this ranking system worked so that when a search query would be input (things like “car insurance) in the search box, the users got 10 blue links on the first page of Google.
Things (as we’ll see in another newsletter issue) have changed substantially today. But this is the basic mechanism behind Google.
This is what a search engine does!
Wolfram Alpha instead doesn’t crawl, index, and rank.
Instead, it gathers data, curates it, then computes and answers.
The process for a computational engine looks like this:
- Data gathering and curation.
- Computation,
- Answer.
In other words, if you ask Wolfram Alpha the “top companies by market cap” (as of today), the computational engine will spit this out:
In short, where with Google you get a fixed (but continuously changing) ranking system.
On Wolfram Alpha, you get a computation on the fly, as you perform the search, in a natural language (spoken English).
But in the age where machines are starting to answer direct questions, and understand more and more (see previous newsletter issues), about what we say, is Wolfram Alpha still relevant?
It is!
Why? As it turns out, machine learning models, like GPT-3 or DALL-E, while extremely good at drawing abstract stuff, at the point to have them compared to digital artists.
In reality, they are not good (yet) with data.
I asked a model, similar to DALL-E to generate an image with three cats and this is what I got!
Apparently, the model can do basic math.
However, if I go more into advanced counting, asking you to paint me five dogs and three cats, things get a bit trickier:
As you can see in some images the count is not correct, and it’s not clear whether the machine is distinguishing between cats and dogs (in the bottom left images they mostly look like cats).
Now, I know, asking a machine about painting cats and dogs isn’t an exciting endeavor.
Yet, it shows one thing, the machine is learning to make sense of data, but right now it’s matching better at abstract stuff, where the margin of error is much much wider.
In short, computational engines like Wolfram Alpha are now as relevant as ever, because they enable us to do things with data, that the machine, for now, can’t do.
From a business standpoint, when and if machines will become extremely good with data, this can change the whole business intelligence landscape. So watch out!
