All episodes

Episode 80

Unsupervised Sentiment Neuron

0:000:00

Show notes

In this episode, Mike and Steve explore the intricacies of unsupervised sentiment neurons and their implications on AI. They also discuss the recent rise of Scala in programming language rankings and Google's controversial move to integrate ad-blocking into Chrome.

Topics

  • Unsupervised sentiment neurons and OpenAI's advancements
  • Tool-assisted speed runs and AI in gaming
  • The rise of Scala in programming language popularity
  • Google's plan to integrate ad-blocking into Chrome
  • Australia's data retention laws and VPN usage
  • Amazon Lex and chatbot development
  • Media: Ghost in the Shell, Mystery Science Theater 3000, and Arne's Brush With Fame
Show transcript

Hey guys, it's time for another episode of Space Welders. Episode 80, recorded Friday 21st of April 2017, Unsupervised Sentiment Neuron. With your hosts, Mike Wise and Steve Rogers. Welcome back from Easter Steve.

It's now downhill to Christmas. Surely they've got Christmas out in force. I think so. I mean I've got my decorations up.

This is now, because like when did they have Easter was only like 250 odd shopping days till Christmas. I think it was in November they had. Oh, much before then. Even Easter eggs out at that point.

I mean, what's the excitement around it? I just, I think we've lost it. The cynical person might offer that it's a capitalistic, commercialized thing where the earlier they get something holiday related out, the more money they can make off of it. Of course, I wouldn't say that being the non-cynical person that I am.

That's true. I would never suggest that. Previously religious holidays are now just completely commercialized. By the way, Mother's Day is coming up soon.

Give generously. Get your cards and flowers and chocolates in now. Make sure you pre-order. And furthermore, if you're listening to us from an underground bunker, good on you.

You escaped. Well done. Maybe you want to put your Pip-Boy on sometime soon. Who knows?

Yeah, so the unsupervised sentiment neuron. What's that about? So just recently, and we do, one of the fun things to watch on YouTube is like playthroughs, but they're kind of like speed runs. And quite often you can see speed runs carried out by...

I love Mike trying to stay relevant. These quick runs, I believe they're called speed runs? Am I saying that correctly? Those speed runs by those kids.

Is that what the kids are calling it? No, but what they... Speed runs. There's a couple.

We'll put some videos on that we saw before where you can, basically they train software to play the game or keep, you know, they detect all of the baddies on screen and places to move on to. The baddies. The baddies on screen. 2D platformer.

Gotta get them baddies. And shoot them up type games. And they let the computer play it. And they let the computer play it against the actual hardware.

So there's some excellent examples where they get in sync with the screens and they're detecting all of the sprites on screens and moving the spaceship around so it doesn't blow up sort of. It's also known as a tool-assisted speed run. Tool-assisted speed run. So just recently...

And not just when an arsehole plays the game, but as in a hardware tool. Complete arsehole plays the game. That would be the AI. So Elon Musk is doing a number of things apart from SpaceX and digging tunnels and whatnot.

He also has... Taking over the world. Taking over the world. Taking over OpenAI.

Helping the lizard uprising. So he and a bunch of other dudes got together and formed kind of like a not-for-profit type foundation. And primarily at the moment, what it does is it trains itself playing computer games a lot. A lot of the time.

But just recently, OpenAI, beating their chests, have beaten Google DeepMind at an Atari game. So they basically have these two types of AI platforms. Even though it's kind of done in a small form where you can watch these speedruns, so-called speedruns, happen. But OpenAI have just recently beaten Google's DeepMind at it.

But... Is this when DeepMind kind of goes back to his room, sits quietly, and then like does the whole... Emo smile. Hey, some of you are cool.

Don't go to school tomorrow kind of thing? Yeah. Is that a bit too edgy? That's a bit too edgy.

But if you want, you can go check out blog.openai.com. It's a beautiful site. I'm going to probably copy it. Let's be honest.

I'm just... I was just playing... Just before we hit record, I was playing with the parallax-y, rotation-y icon things with like... They've got like lighting on them and everything.

Yeah. It's amazing. There's a great layout for a site. It's like it's all CSS.

Anyway, so... It's terrific. One of the main articles on there is unsupervised sentiment learning. So there's different types of AI that you can have and the obvious leading to is supervised and unsupervised.

And so supervised is where you're informing the AI, that's a good choice, and that's not a good choice, and that's a good choice, and over time it builds up a bank of weighted or learned responses which has been driven by a human. And that's not really... That's sort of not the ultimate goal here, but you can store that up and send it around in files. And natural language parsers work kind of like that where they've been trained and Apache actually have online where you can download files for their NLP, pre-trained word detection where they can pick out words and sentences, they can pick out locations, they can pick out money.

Anytime you talk about a country, Apache NLP can pick it out for you and it gives you a nice little struct of, you know, this is what was made up in the sentence and, you know, these are the key things I thought about the sentence. But what's going on, what Elon's group is doing? They've got like an ability to work out the sentiment of something that's being spoken about and that what they've been doing is actually not to, you know, meanwhile they've been playing 80s console games, but training their AI and tools and they've got like a gym. You can go to the AI gym on the site and play with their tools and you can submit some of your tools up onto there and they give you hints and tips on how to use their software and so on.

But the unsupervised sentiment neuron, what they've been doing is working with neural networks. Neural networks are generally the ones where anytime anyone says like deep learning, generally they're working with a neural network of some kind. And then, you know, over the period they're basically feeding it mounds of text in order to be able to train it endlessly. And so they use things like sigmoid squashing functions and other sorts of ways of being able to weight the network.

And there's quite a few ways that you build up the weights and the various vectors in which the network fires when it sees particular sentences and so on. But the interesting thing about this, when you read the paper for it, is that, you know, they're really targeting unsupervised and they're pointing it at things like movie reviews and reviews of products online. Now you can probably, the conspiracist in you thinks, oh, why would they be trying to work out, you know, product reviews or some kind of understand, you know, what's the best products online? Is that some kind of, you know, wouldn't humans normally rate products quite well?

That worked out quite well on Steam, Steve, didn't it? Not so much, no. But unsupervised is quite difficult. So what they're doing is looking at the sentence structure and they're also, in this particular case, they're looking at the next character that appears in the word sentence order.

And then what they're trying to do is build up basically a graph where they can position and say, oh well, this series of texts were kind of happy about what the product was doing for them and they seem to, you know, it's quite strange, like how would you get a product review which was saying, gee, this, you know, popcorn maker was doing great things for me. There must be stuff out there. Maybe people who write movie reviews are more articulate. But anyway, they're using it to do that and they've developed this neuron which is a part of the network which is basically a sentiment neuron that works out whether or not the text it's reading is either happy or sad or, you know, and how to do that.

Which is kind of interesting because Zipf algorithms and Bayesian algorithms, you kind of would think, like, if you're just looking for the word the, how many times does the word the appear? So they'd have to be removing that. And they must have, like, some kind of Bayesian or naive Bayesian algorithm to remove all of the shit. Like, most spam filters are Bayesian filters anyway.

So they're looking for, you know, Russian prostitutes, you know, Nigerian money handler type people. Those are the sort of, like, pre-loaded keywords that you feed into these things. But anyway, I urge you to go and check out the blog openai.com. It's a great read.

They produce, they put lots of, all of their papers are up online, PDF format, so you can sit there and learn them and read them and kind of think you're smart at the end of the day. But yeah, there's plenty of good stuff going on over at Elon's other company, which is OpenAI, and surely they'll save us from pending nuclear disaster. The important question is, have they done a blog post on how to disable the AI remotely? I don't know.

Maybe they're teaching it to tell jokes. That should be the first blog post that they put out. It's like, when this inevitably goes wrong, here's how to turn it off before we unleash the AI on the world. It's got some of the sexiest paper titles ever.

Like it's got attacking machine learning with adversarial examples. Faulty reward functions. Like a phased plasma rifle? Yeah.

Faulty reward functions in the wild. If it seems like Google wants to put better ad protection in Chrome natively, don't you agree? Is ad plus or adblock plus? Or ad plus?

Of course I do. Because the internet without ad blocking is a horrific, nightmarish hellscape. Don't you agree? I just press the ads.

I love seeing the ads all the time. The same ads. Do you? You click on every single one?

Yep. Every single one. I love this ad. Especially if you're like watching YouTube and you see the same ad 50 times, because they do the ads based on your demographic.

And my demographic hasn't changed in the last half an hour. So I see the same ads again and again and again and again and again. I love that. I wonder if your ads are different to my ads.

Probably. Am I missing out? I've got FOMO over here, like I'd love to be seeing your ads. They're probably a bit more, you know, ooh la la compared to my ads.

My ads have got like missing tax reform. You know, what was it? All the exciting ones. I honestly can't remember because even when they're on, I just ignore them.

I refuse to even give the ad companies even 5% of my mental capacity. So I literally prefer to stare off into space for the 30 seconds that they're on. Yeah, it's like I regularly get emails like, thinking about changing to this particular online tax accounting software. You must live an exciting life.

Presumptive. You know how people are afraid of their browsing history falling into the wrong hands? I don't think you would have that problem. You would single-handedly take down the NSA because it would be the most boring browsing history they've ever found.

They'd be like, ooh, we've got a new one. What's this one going to be about? And they'd open it up and it's like, tax accountants? Yay!

Are they terrorist tax accountants? No, just income tax accountants. My wife gave a presentation this week at TechOne and there was literally a thousand plus people who registered for the seminar just to learn about the new laws in contract tax accounting. And boy, wouldn't you love a dinner party with one of those people.

You know the usual thing, like 230 people turned up and she had the marketing lady in with her whilst they were doing the presentation itself. And like, you know, it's one of the hardest things because you can't hear them but you've got someone who's reading out the lines of questions and you're just talking to nobody, nothing. Like this. It's like there's no one talking back at you.

Yeah, so was it 250 people turned up but due to some clever practices they actually only had like 100 people in the room? It was the same. Is that how it works? Yeah, it was the same person.

The tax accountants conference. They can fit 300 people in a room designed for 50 just because some clever writing down. It's like, it's like going, it's like a... This is actually a fire safety deductible so we can shove more people in.

You wanted a FaceTime? All that matters is the post-fire safety count. You want a Facebook Live with all the hearts and likes flying across the screen. That'll at least give you some motivation as you're getting into the hardcore.

You should have GIMS for that. Yeah, well, yeah. Or even, I don't know if you've ever been onto Google Live and you see the chat that's flying off the screen. It's completely useless.

Oh, any live stream chat is just terrible. Live stream chat is just stupid. But anyway, Google's trying to ban, no, no, they're not banning, they're blocking. So Google, an ABC company, Google.

Alphabet. Alphabet company. So they're planning to put an ad block feature into Chrome as a default. What's that kind of mean?

Like, are they, if you're suspicious about this, then you can kind of think, hmm. Well, at first glance, it seems kind of conflict of interest-y and you'd be right because Google doesn't like their ads. Google or... Don't they earn money from ads, Steve?

Yeah. Let's be more generous than that. Alphabet is an advertising company, really. It's not a search engine.

It's not a video hosting site. It is, at its core, an ad-serving company. And all of their other products, Google and YouTube and whatever else, just exist to allow them to serve ads to as many people as possible, to make the billions and billions of dollars that they make that will be worthless once the apocalypse hits and we all use bottle caps for currency. But what they're doing is they're seeing, and quite cleverly, they're seeing the way that the market is going, which is towards ad blocking, and they realize that, well, if we control the ad blocking, then we can make sure, first of all, and again, the cynical person might think that we can block other people's ads and only let our ads through or serve a Google ad instead of some other ad company's ad.

But I wouldn't say that, being the non-cynical person that I am, of course. It's true. It's like building a superweapon to bring peace to the galaxy. Yes.

But to be fair, most of Google's ads are quite unobtrusive. There's certainly worse ads out there. Could you tell a Google ad from any other ad? Well, generally, Google ads, so Google subscribes to the Coalition for Better Ads.

In itself, it's a bit dumb. Well, so these people realize that ads are a necessary evil, because otherwise, how do you make money? You've got to make money somehow, unfortunately. That's just the way the world works, and a good way of making money is advertising.

But what they've realized, the people that are in this Coalition for Better Ads, which is an industry that releases ad standards that are for inobtrusive or unobtrusive, unobtrusive, inobtrusive. If you know Latin, let me know. Ad standards that prohibit things like pop-ups, auto-playing video ads with sound, and prestigial ads with countdown timers, which are ads that come up on top of the screen. If you're scrolling through an article, I think the Wall Street Journal is particularly, I mean almost every blog out there now, you'll be scrolling down and halfway through, you'll get a pop-up.

It's like, click here to continue reading, or you could subscribe, those kind of things. Just to be clear, if you've been living under a rock, you have to click the ad to get the cash, right? Well, it depends. You can get cash just for impressions, which is the number of times the ad is loaded, but it's lower than the number of clicks, so that YouTube ads in particular, you do get some amount per thousand views, but you'll also get some different amount, probably more per thousand clicks of the ad itself, but I make it a point never to click on an ad because in this day and age of tracking cookies and everything, as soon as you click on one ad, it's goddamn everywhere.

I don't know if you've ever been to a website that advertises on Facebook, but you'll go to a website and suddenly their ads are everywhere on Facebook, and my opinion of that is I don't need ads. I already know that you exist because I've been to your website. You're showing me ads for things which I already have. I remember a funny story for a previous company and a previous product we used to develop, and we sent some people to look at the website, and that then turned the advertising on, and then wherever they were browsing next, they seemed to see ads from us in their stream, and I remember getting a phone call about a week later going, oh my god, Mike, you're so popular.

Your product's amazing. And it says, no, no, you're just being stalked. So yeah, you can be tracked all you like, but what does it mean for, like there are free ad servers out there. You can download your own ad server.

In fact, we can have the Spaceworld as ad server, Steve, if we truly desire from a company called Revive. It used to be called Xblock or Xad. It's kind of like the WordPress of ad serving. So Revive, they haven't got like a statement.

And I was searching and searching and searching to see if anyone had mentioned it, but then there's plenty of ways that you can overcome ad servers anyway by changing the way that the content is presented in the HTML and received in the browser. But Revive's been out for ages, and you can create your own ad serving, and you can do ad serving to mobiles and- Well, what I like with Revive is they have a forum, and there's a post on their forum about beating Adblock. And nicely, one of their members has a number of steps you can take to beat Adblock that you could then implement in Adblock to beat the beating Adblock. Things like, add something in front of the zone ID.

It's okay. Well, I'll just search for zone ID anywhere in the URL and we'll block it or call it something different. Well, you only need to find it once and then you've got it. They're like, yeah, well done, guys.

You've told me how to beat it. You've done it again. Public forum. So there's ways and means around this, but this depends.

It's now coming default. So if it's default switched on, are people going to complain? And then if it's default and controlled by rules external, which are Google's, people are going to complain. Well, this is what Google's- the reason they're doing it is they're seeing how many people are using Adblock because they know how many people are downloading it in the Chrome plugin extension store thing.

And they want to make sure that they can control how people are blocking ads. And I'm sure they would rather block their own ads with their own product than have someone else blocking their own ads. At least then they're controlling the system and they can put these what are so-called acceptable ads. I tend to agree.

If you look at the Coalition for Better Ads, they have a list of examples of acceptable ads and they tend to be text only or non-moving picture ads on the sidebars. Nothing in the main content of the page, no pop-ups, no flashy animations or sound, things like that, which are all very reasonable alternatives for the necessary evil that is advertising. So, speaking of- matter is programming languages, Steve, and it seems that Scala, yes Scala, has made a move up the charts. Just recently RedMonk, who follow these trends online and tracking, and if you've ever been to GitHub and get annoyed that your favorite language isn't in the list when you go and check out the trending, well there's a reason for that.

And they do a lot of research and, you know, across the tags and show the, you know, the highest ranked ones. But there is actually a ranking survey out there from RedMonk. So JavaScript is obviously numero uno. Java coming in second, probably because it's a shit ton of legacy and a shit ton of Android development going on.

Then there's good old PHP and Python fighting it off. But C Sharp is coming in there, Steve, it's up there. And then good old C++. And there's movement at the station of C++.

It's alive and kicking and very, very active. I think C++14 is coming or somewhere. They've got concepts and updates coming out soon. Objective C is running up from, following up from behind.

And then C and Ruby. And interestingly, interestingly, R and Swift. R and Swift are the newcomers. And then Scala, right down, right down the list there.

But at the bottom, the least used, least used, what have we got? OOC. Source Porn Eagle. Well, I want to use that.

What about, there's Vala. Well, that's someone who's, this prologue is way off the map. XML is a programming language. They've got that in there.

So, so this is a, an investigation, an article they've written ranking programming languages in January this year. I have a little bit of a problem with the way that they've, or the conclusions they've drawn from it. It depends on your definition of popular. Popular doesn't necessarily mean people like it.

So, what they've done is have a scatter chart rank, comparing the popularity rank on GitHub, which they define as the number of pull requests for a project using that language, versus the popularity rank on Stack Overflow, which they're determining by the number of questions for the language tag. So, what that means is if you have, just because something is a lot of questions on Stack Overflow, doesn't necessarily mean it's popular in that it's, in that it's, it's nice to use. That's why XML has a, has a shit ton. PLSQL has a load of, of Stack Overflow questions.

So does Prologue, Mathematica, Cucumber, whatever the hell that is. NSIS, which is if you've ever used NSIS, it is arcane magic to make that work properly. You literally have to reference window handle, Windows handle IDs in the window to actually like change the color of the installer. It's, it's terrible.

It's black magic. So, all that means is maybe those languages, or, or languages, I use that in a loose term, are maybe harder to use or have more gotchas, versus things like, like HCL, which is, I believe that's Salesforce's kind of query language, HCL, something like that. Objective-C, Logos, Pascal, a few other things that are kind of below the line, as it were. And then you've got things with lots of, lots of pull requests on GitHub.

Again, doesn't necessarily mean people are liking it. It's just what people are using. And it's all, all the standard things. It's C and it's various permutations in Java and JavaScript and Python and Ruby and Objective-C and PowerShell and all of them.

So it's not necessarily, it's, yeah, ranking languages by popularity, depending on your definition of popularity. Yeah, I think it's, it's, it's kind of interesting of how they've pulled this together. It does explain, like, why CSS are up there, even though they've kind of got, like, CSS and Ruby ranked around the same usage rate. But it sort of gets back to, you know, also public, sort of more broad understanding.

So this, this article did do the rounds, you know, through your LinkedIn's and so on. So is that being liked and pushed up by, by management folks who kind of wanting to understand a little bit more about, you know, the development or how sausages come to be in the Sausage Factory? And if they were choosing a Sausage Factory, which Sausage Factory would they choose? Whether it be a JavaScript based one, maybe recruiters use this as their standard filtering technique.

25 years in Scala. That's like four, four years I'm hot, you know, I'm lit as fuck in NSIS. That'd be, that, that's what, that's what, you just go with that straight on the letterhead of your resume, Steve. You have the, like, the little flame symbol.

You have the flame, A-F-N-S-I-S. And then no questions asked, you'll be just hired straight away. 140k, you're done. Also coinciding with Easter, Steve, is something that's going on in Australia.

If you're not familiar in the US, you're probably more worried about your data being collected by the ISPs. But in Australia, we're now required by law, not we, me personally, but our telecoms. No, I am though. Are you?

Really? Yeah, I have it all. Do you? Yep.

Great. It's on a USB stick. Is that the truck parked outside your home? No, it's on a USB stick.

Smaller than you think. Is it? Yep. Did you zip it up?

You've got some amazing compression algorithm. Yeah, I had to buy a copy of WinZip. And then you formed a startup around this compression algorithm and doing video from your garage. Yep.

That could be the premise of some kind of show. No, it would never work. No. So, anyway, so they've got to, telcos in Australia have to start collecting metadata about you.

Parts of this is really around a data retention plan. And it's, you know, it's now enforceable from this point forward. We've spoken about it before on the podcast. But what, you know, in the media at the moment that we're talking about, now's the time to get your VPN sorted out.

And even so, what reputable VPN supplier would you go to anyway that would still be stealing and taking from you, as explained last week, when, you know, you will, they will still see your traffic. There's not not too much about it. You might as well do it yourself. And, you know, sort something out.

Get a free plan VM off Amazon and go for it. But anyway, so there's no surprises here. They've got to be compliant. So I wonder if there's a lot of developers out there who are looking at, who are trawling logs for a living.

Maybe Splunk is being deployed everywhere inside of ISPs to carry out all of the metadata tracking. If so, then what sort of metadata are they looking at? Non-content? Who knows?

So also, speaking of Amazon, Steve, Amazon have got something new. It's like they've got something new each week. They've got a couple of shows going on at the moment. So you're probably seeing a lot of that stuff on YouTube or through your LinkedIn feeds.

Amazon have now provided an online service for doing bots. Now this week, and you're probably seeing me tweet, and I'll continue to tweet and annoy, MusicMatch. So the MusicMatch, we've developed a tool inside of MusicMatch and it does AI lyrics for songs. Long story short.

And so here's an upstream provider who already does this. And so I've been bothering them to try and get back to me. But they proudly make their, and so on their website say that they proudly make their software in Italy. And if you're in Italy and you hear this, buongiorno.

Anyway, so on their Facebook site, when you go to it and you use the messenger on their Facebook, and this is the disappointing thing. So you think you go to the bother on going to someone's Facebook page, talking to the messenger and typing something in. And what happens on MusicMatch's site is it comes back with a bot response. And I'm going, huh, how is this possible?

Well, it turns out Amazon Lex is there for you. So if you've got a standard Amazon account, you can log on in, you can go to Amazon Lex. It's available in this region. I'm using one in North Virginia at the moment, but what it does is allows you to build bots.

There's an API and you can basically load in intent and various other things that allow you to have a bot like conversation with somebody. Their ultimate purpose is to be able to execute Lambda or Webhooks or whatever to be able to drag you across to your site or answer something. Specifically, people are using mostly with Lambda. And what it does is pretty cool.

You can create a bot. You can create a bot, you know, based on any name and then you kind of go through the process of building it, giving it utterances, Lambda validation, all various channel settings and so on. But the killer of this is that you can actually distribute this bot to a channel. And one of the channels is Facebook.

Another one, a favorite of ours is Slack. Everyone's using Slack. If you're not using Slack, who are you? And Twilio of all places.

I think there's a drinking game around Twilio. It still exists? Yeah, it probably does. So anyway, so you can have this go to your Facebook site and then become a bot for your messenger chat.

I think I might just replace everything I do with this. Can you use them to host podcasts? We can record a bunch of noises, you know, like phonetic sounds and then have the bots do the podcast instead so we don't have to do anything. Yeah, well you've got an editing facility and you've got the little test bot in the bottom right hand of the browser.

And then you go through a process of creating intents and utterances. An utterance is kind of like pattern matching for, I want to buy a video. And then you build up a set of responses to, I want to buy a video. Or I would like a slice of pizza.

And you hear that as an utterance and the intent is buy pizza. And then you provide the steps to buy pizza. And then you would say, what sort of pizza? Would you like something like pepperoni and cheese?

Or would you like a vegan selection? And you would type go fuck off. And you will go through that sort of motion. So yeah, you can sit here and use Amazon Lex to build your next bot.

So Steve, it's come to a bit of media. So I went and saw Ghost in the Shell eventually. I think most people have seen this way before me. But have you seen it?

No. You've not seen it? Nope. I'm not planning to.

Never was into it. I don't care. Right. It's very pretty to look at.

I think it's badly edited is my answer. You see a lot of the set production and design around the place tested. Did they actually use sets or is it all green screened and CGI'd in? Seems like it would be a movie that would be CG'd.

No, that was the interesting thing about it. The set design for this thing is absolutely amazing. It's gorgeous. They shot it all somewhere in New Zealand, Queenstown, I didn't know.

And it was all Weatherworks design and that sort of thing. So yeah, they put a lot of time, like the opening sequence of the major being created, which is also essential to the plot as to why she looks the way she does, is all practical, the whole thing. And they really only CG'd in just a few things, which I think is quite amazing. And so you can, yeah, there's plenty of behind the scenes stuff.

I'll actually recommend, there's a great and amazing YouTube channel you need to watch. It's called Cinematography Database. And after I saw Ghost in the Shell, I wanted to see a bit more, you know, you see Adam Savage's, he spent time at the Weatherworks talking to the guys, and I wanted to find out a bit more. And I've come across Cinematography Database.

Now, Cinematography Database, he's a DP, or a light designer, and he spends most of his time talking about lights, as if lights could be really important in movies. I don't know. These movie video reviews are getting really specific. Yeah, this is, I love this.

This is, hello and welcome to 200 Watts. This is the video review channel where we talk exclusively about the lighting in the scenes. This is exactly what he does. How, unbelievable.

Anyway, there's Cinematography of Ghost in the Shell. I think I'm going to start one talking specifically about background characters, but only background characters that are out of focus. I think there's a real niche for it. Like what?

How would you, like? Well, I don't want to give anything away. All right, so it's coming soon to your YouTube channel, Steve. It's going to be called Background Action.

But there are plenty of folks, like you've got Nerdwriter, who's looking at the story and how it's made and so on. This is hardcore. This is, there's this kind of like, this is the camera that we use. These are the interesting things.

One of the amazing things about this particular channel that I like is, you know how sometimes behind the scene shots, you actually see the camera on the dolly and all the people standing around. He sits there and explains every one of them. But if there's a screen that has like, you generally see the numbers around the edge of the screens. There's two screens, maybe, where the directors are sitting behind.

He can explain every number on every screen, and he explains them in such detail. I hope most people that work on a movie would know what most of those numbers mean. No, some of them are not what you think. And then you kind of go, oh, I never knew what that number was, and why was that number repeated?

So it's not necessarily like, you know, how many frames per second and why the frames per second were such the same way. Like, for example, I know that they had a very strange, like, and the different types of camera and different types of lenses. Like, they used, like, for example, Rogue One and Ghost in the Shell used the same lenses. They used the same camera equipment.

They used digital Sputnik lighting. So, you know, I just learned so much from just watching this guy, not pick it apart, but actually tell you how it was done, how each of the scenes work, and all of the practical effects that are actually inside of it. So, surprisingly, Ghost in the Shell, even though it's so luscious with lots of CG background, but that's the background that sort of builds up the city, which is what you need. But in the foreground where all the, kind of, acting happens, that's all practical, and there's a great degree of it.

But I will add my two cents. I don't think of it as a whitewashing at all. That's, like, it's a part of the story. There's a reason why she's a white girl.

And you find out at the end, it's like, oh, okay, yeah, sure, that's why she's a ghost in the shell, you know. You know, you could be, if you came back, like, what sort of robot shell would you want to be in? Would you want to be in a good-looking girl, or you want to be in an Asian body? What sort of body would you want to be?

Would you just want to be... Trump's. Trump's buddy. Yeah, that's good.

Yeah, that's pretty good. There's a good chance he is already. I mean, come on. Yeah, he could be.

He's Trump bot. Maybe he's just an AI bot. Who knows? So there's that...

Controlled by the Illuminati. Hey, Arne's Brush With Art, or Brush With Arne, is back on ABC, and it's fantastic. It's two episodes in at the moment. The second episode's very, very heavy.

It's an interview, or he's painting the strain of the year. So I won't spoil it for you, but it is a very powerful... I'm going to guess he finishes the painting. He finishes the painting.

Oh, no, I've spoiled it. There is no jeopardy in Arne's Brush With Fame. Sorry, everyone. Ruined it for you.

No, very, very powerful. Also, I think it's on ABC at the moment that you can't ask that, which is also very good as well to watch. So a little bit of free-to-air that I've been watching recently. Steve, do you have anything?

I don't think so. I've been... I don't think so. I've been watching the reboot of Mystery Science Theater 3000, which has just been released on Netflix.

What episode are we up to? I'm up to episode three. I've watched the first three. Yeah, so I'm actually up to episode four.

It's really good. It's funny. It's up-to-date. The problem I had with watching the original one is it's a little bit dated because it's all kind of pop-cultury references, but the new guy playing Jonah is really funny.

They have the same... Wait a minute. Pop-culture references? That's right up your alley.

Not from 25 years ago. Oh, right. In your wheelhouse, not 25 years ago. Oh, so 50s.

No, not even that. What? It was like the mid-90s. Mid-90s pop-culture references.

Yeah, which I don't necessarily get. But they have the same puppeteers doing the puppets, who I don't know the names of, even though they do the robot roll call at the beginning. Jeff. No.

And Felicia Day and Patton Oswalt are really good as the villains, and it's funny they have a house band now playing light jazz and ska in the interstitials. It's terrific. Check it out. And House of Cards is coming back soon.

Thank god. The funniest thing is House of Cards would be like watching relief. Yeah, like last year watching House of Cards was like, oh my god. Yeah.

This year it's going to be like, eh. You don't know what. Okay, sure. Yeah, you know, whatever.

But if you're listening to us from your shelter, tell us. Write to us. We'd certainly like to know. So Steve, if you wanted to have some kind of sentiment, supervised or otherwise, how would you express that sentiment to us?

If you want to express the sentiment to the Space Worlders, just put it in an ad and then I'll block it and I won't see it anyway. So, oh well. But I'll change up my ad. I'll change up my sentiment.

I'll just block that one too. And I'll change it again. Otherwise, just head over to www.spaceworlders.com where you'll find the show notes for this episode and all the other episodes that we've talked about all the way back to episode two. And oh, then maybe there's episode one as well.

You can subscribe to the show on iTunes. That would be the pilot. Yeah, that's why it's crap. iTunes, Stitcher and SoundCloud.

Just do a search for the Space Worlders podcast. Make sure you like and subscribe and comment and rate and share and print it and put it on a banner behind a plane and sky paint. No, sky paint? That's not the right one.

Sky write. You could get Pepsi to do it. They did a good job with ads recently. No.

You can follow us on Twitter. Twitter.com.spaceworlders Facebook.com.spaceworlders Mike's on Twitter. Hey, that's some, some, some thing. I'm on Twitter.

That's some other thing. I'm sure there's a link somewhere. Make sure to check out our store. You can buy some t-shirts and some merch.

You can also buy a badge from the website. There's a big buy badge button. Put it on there. Buy us a cup of coffee and then we'll send you a badge.

It's pretty sweet. You can iron it onto your bag or your pants or your face. Iron it onto your face. Finally, if you have any questions, comments or feedback, email us info at spaceworlders.com.

So guys, 80 episodes. It's huge. We continue on, hopefully barring any kind of major catastrophe. We'll be producing more of these comical episodes of what we call Spaceworlders each week.

So it's Mike out. Steve out.

Unsupervised Sentiment Neuron · Space Welders