<img height="1" width="1" style="display:none" src="https://www.facebook.com/tr?id=338168267432629&amp;ev=PageView&amp;noscript=1">

Let’s Talk about Algorithms

Algorithms are simply statements of logic chained together, but where they get really powerful is when machine learning is implemented in them as well. These are the types of algorithms we mean when we discuss how social media platforms can attempt to influence our engagement. Full Transcript Below: Cody: …


Algorithms are simply statements of logic chained together, but where they get really powerful is when machine learning is implemented in them as well. These are the types of algorithms we mean when we discuss how social media platforms can attempt to influence our engagement.

 

Full Transcript Below:

Cody:

How does Facebook, or YouTube, know what to put on your feed based off of what you like?

Cris:

Okay.

Cody:

Machine learning algorithms are entirely different.

Cris:

Facebook algorithm, the Instagram algorithm, LinkedIn has an algorithm.

Cody:

Algorithms sometimes, not even the companies that write them know what they are.

Cris:

Let’s jump in today, and let’s talk about algorithms. This is the buzz term that you always hear floating around. The Facebook algorithm, the Instagram algorithm, LinkedIn has an algorithm. What is an algorithm?

Cody:

That is a fun question, and one that marketers don’t want you to know about.

Cris:

No.

Cody:

Because here’s a little tidbit. Algorithms, sometimes not even the companies that write them know what they are. And we’ll get into that in a minute here. But there’s a couple of different kinds, to start out. Typically, in computer science and even in mathematics, an algorithm is a procedural logic tree.

Cris:

Okay.

Cody:

Basically anyone who’s slightly familiar with code knows about if else statements, but for those who are not initiated, that basically is, “Hey, if this thing is true, do this. Otherwise, do this.” And then an algorithm is basically a chain of those decisions typically. At least in traditional computer science they are.

Cris:

Gotcha.

Cody:

And so that could be like, “If name is Steve, do the thing that Steve should see. If they’re not Steve, do for everyone else.” That kind of a thing. So, that’s super simple.

Cris:

Right.

Cody:

Going from that, you can see if you chained that together, “If my name is Steve, and if I was born on this day, and if I had this car,” you could chain that logic together to make very complex decisions based on known parameters. And that’s typically what a traditional algorithm is. It’s really not that complex. It’s just a logic tree of, these steps equals this result over time. And the part where it gets complex and where people start talking about the Facebook algorithm or the YouTube algorithm is when we bring up the topic of machine learning.

Cris:

Okay.

Cody:

Machine learning algorithms are entirely different. And the main part of why they’re different is because they’re not defined. Whereas logic in a traditional algorithm is a known analysis of parameters, a machine learning algorithm, while they typically know the parameters, they don’t necessarily know what the output will be at the end, and that’s where things get a little complicated.

Cris:

So we’re, again, used to algorithms because we hear thrown around with all these social media apps and that stuff. So you’re saying that the social media apps, they’re more on the machine learning side than a standard linear algorithm? Okay.

Cody:

That is correct. And now that said, a lot of the times they have very specific use cases. So when people say “the YouTube algorithm” or “the Facebook algorithm”, they’re not talking about what happens when you log in, they’re talking about how does Facebook or YouTube know what to put on your feed based off of what you like? And that kind of information, that kind of qualitative analytics of people’s insight into, or I should say the insight into people’s desires and what will actually keep them entertained, that’s where a lot of the social media industry has taken that word algorithm and turned it on its head.

Cody:

Because it’s not so much an algorithm, even though in a, I guess you could say, electrons on the metal sense, it definitely is, it’s more of a machine learning than it is an actual algorithm, which are almost not quite the same thing within the computer science domain.

Cris:

Interesting.

Cody:

Yeah.

Cris:

So it’s almost like we liked Kleenex and we just started calling it all nasal tissue Kleenex.

Cody:

Yeah. And I don’t want to get away from the fact that for instance-

Cris:

I know, semantics.

Cody:

Yes, semantics.

Cris:

It’s always my favorite analogy.

Cody:

Of course, you got to bring the Kleenex in.

Cris:

You got it.

Cody:

I don’t want to get away from the fact that machine learning is some sort of special thing that isn’t an algorithm.

Cris:

Sure.

Cody:

It definitely is. But the part of it that’s a little different is that machine learning is an abstraction away from what is traditionally known as an algorithm within computer science. It’s a whole nother level above.

Cody:

A good example of this is in traditional algorithms you would program step-by-step on what the logic decision tree actually would be.

Cris:

Okay.

Cody:

Whereas in a machine learning algorithm, you would train it against data, much like you would train anything that has to learn. And a lot of the time, for instance, one of the reasons that large companies are typically the ones that benefit from machine learning is that it takes a long time to get machine learning algorithms to be trained correctly, to give accurate and predictable and useful results.

Cody:

It takes often petabytes of data, the one above tera, and just constant reiterations. And the fun part about machine learning, and the reason I mentioned early on that a lot of the time these companies don’t know their own algorithm, is that you really only have control on what is fed into an algorithm for machine learning. You have no control on what goes on underneath the hood.

Cody:

So for instance, if, again, our character, Steve, all of a sudden starts getting something he’s not interested in, he’s a car guy and all of a sudden he starts getting ads for an air control board, and you can’t have catalytic converters of special kinds in your car. He’s like, “I hate that. Why are they showing this to me?” No one can go in and tell the algorithm to stop showing. You can only give feedback as an input. You can’t change it under the hood.

Cris:

Interesting.

Cody:

So that’s why a lot of the times you’ll see on Facebook and YouTube and other social media platforms, they’ll see a little way to report an ad is, “I don’t like this.” And then it will just stop showing that. But it’s only managing that on the output and it tries to send it back through the machine on the front so that just deletes that and tries to say, “This was a bad result. You tested and failed.” That’s what it’s telling the machine when you do that. And it tries to like, “Okay, this was wrong. Let’s do it again. Let’s get something new for Steve.” That kind of a thing.

Cody:

But there’s no way to just say, “Hey, this guy like this particular thing versus something else,” it’s more of a feedback loop than it is an actual process of changing code.

Cris:

Gotcha. So we’re using algorithms to obviously solve these problems. What kind of problems are working well, and what kind of a problem would a client potentially bring to us to be like, “Let’s just throw an algorithm at it,” that maybe is not the solution?

Cody:

Well, for one, if we’re talking about machine learning algorithms, they’re incredibly complex to do correctly. Now, there of course are experts in the field that of course can do things instantaneously like anyone else out there. But doing anything meaningful with machine learning really does require a large amount of effort if you want to make, well, money with it.

Cody:

So that said, when someone says, “I want to just throw an algorithm at it.” Well, that’s like saying, “I want to make a code project.” It’s one in the same, really.

Cris:

Yeah.

Cody:

So in a lot of ways, saying, “I want to throw an algorithm at it,” is like saying, “Well, I want to write some logic to do something.” And that makes sense, but it’s also an incredibly broad term.

Cris:

Sure.

Cody:

So from a standpoint of how do we actually benefit clients with using algorithms or using that, it’s a little bit of, I guess the term would be a non-sequitur term, I think?

Cris:

Okay. Yeah.

Cody:

Filler words.

Cris:

Filler words.

Cody:

Yes. But regardless of that, it doesn’t really make a lot of sense because we’re using algorithms for everything. And ultimately, if we want to, let’s say, make a complex section, let’s say we were making another social media app that wanted to have a feed to show specific elements from other users on the platform that it learns dynamically. It’s like, yeah, that would be machine learning.

Cody:

But let’s say we just want to decide what to show this person based off of settings they chose. And it’s not based off of dynamic content that has to be curated by a machine. So to speak. In that case, we would just write a standard algorithm, which is just normal programming, in other words. Really an algorithm is just normal programming.

Cris:

Got it.

Cody:

That’s what we do day in and day out. So, using it as a buzzword, which marketing loves to use-

Cris:

Loves the buzzwords.

Cody:

Typically, at least in today’s modern world, usually means actually machine learning. And the irony is that it’s almost not quite the same thing as what computer science experts actually think about as an algorithm, because they’re typically not writing it. They’re inputting data and then hoping for a result.

Cris:

Yeah. It’s always interesting, because that’s how things often are in the world. You created it as one thing, it’s meant to be something, it takes on its own life. And not that it’s incorrect, but if you really start to dig deeper, you can get into semantics and get pedantic with stuff, and you realize that algorithms are not just this very… Well, I guess they are a very specific thing, but we’ve used it in a much broader level.

Cody:

Yeah.

Cris:

Interesting. Any final thoughts on algorithms, how clients can utilize them, integrate them into their projects? I mean, I think we did a really good job of helping understand what algorithm means, but yeah, any other closing thoughts?

Cody:

I would say algorithms will definitely help you with your project, because algorithms are your project. It’s the truth. And with that said, obviously there are going to be complex areas of your project where you want to decide how certain things should be done in a logical flow, and that’s an algorithm too, just like it is to actually have it serve anything in the first place. So, don’t get the terms mixed. Algorithms are there, and they’re effective.

Alexandra:

Thank you joining us for this episode of Bixly Tech Tuesday, where Cris and Cody discussed what even are algorithms, and how are they used in web technology? I hope you learned a lot from this episode, because I actually learned a lot from this episode as well.

Alexandra:

And don’t forget to check out the links that we have in our description box down below. You can find a link to our free custom software guide, which will walk you through the process of planning out your own app idea. And you can check out our website, Bixly.com. And you’ll see at the top, we have a “validate my idea” button, which means that you get a free 60 minute meeting, with Cris, to talk about your next step idea. I hope you enjoyed this episode of Bixly Tech Tuesday.

Similar posts

Get notified about the latest in Tech

Be the first to know about new tech from the experts at Bixly!