Do you like RSS? Add Smash Notes to your daily!
Smash Notes on Smash Notes

June 10, 2019 - Weekly summary of the best podcasts on the internet

Smash Notes podcast.

June 11

download episode
What is Microsoft doing with Machine Teaching?

Microsoft is tired of paying big bucks for data processing and is looking into new ways to do machine learning. They are calling it "Machine Teaching." This segment of the Smash Notes Podcast covers an episode from the Microsoft Research Podcast.

this'd smash note. A podcast that brings your weekly summary of the best podcast on the first up on this week's edition is Microsoft Research podcast. Have you heard about machine learning? It's a way of making computers do work for you without telling the computer exactly how to do the work for you but instead letting the computer learned from a bunch of data. Microsoft is apparently working on a new project that's called machine teaching, and that's all about making the machines do the work for you without having the data to teach them to do it for you. It sounds really interesting because the premises, basically how do we teach the machine toe? Learn like a human would Hold on, Let me get the expect to explain it the test. So when human T shirt a human, they teach in a very different way, then they teach a machine. Now, if you have millions and millions of label, then the task is about extracting the knowledge from that data. But if you start with no data,

then you find out that labels are not efficient at all, and this is not the way human teacher, the human so there must be another language. And the other language is what I call machine teaching. This is like a programming language, and just to give you an idea of how natural it is, what I see happen over and over in industry is that when people want to build a new machine learning model, they start by collecting a whole bunch of data. They write leveling directives, and then they also said, and then they get back there 50,000 labels and then you have a machine learning algorithm tried to extract that knowledge from those labels. But this is ironic because the leveling directive contain all the information to do the labeling. So imagine now that the laboring directives could be inputted directly into the system. Now, when you look at the leveling directives they'll features, they're saying,

Oh, this is a cooking recipe because it has a list of ingredients. So if we can make that the teaching language, then we can skip the middleman and get the machine, do the right thing. Okay, now that we talked about the next phase of teaching and computer to learn new tasks, how about we talk about teaching kids to learn. AD Astra is the school private school in Los Angeles founded by a lone mosque to basically educate his kids. They have a very cool and very nontraditional way of approaching learning. The school is based on campus of Space X, and their curriculum resembles more of a business than a typical school curriculum. Personally, I think it's really cool. It is the future of education, But let's hear from Josh down the head of school there. He explains what they do, how it's different from public schools and what public schools can do to incorporate this kind of curriculum into day to day life.


If you take the example of like, let's say that give you a catalogue of, like, 60 different works of art and artifacts overtime, you know from all over the world. And each of those different works has a unique demand from 10 different cities around the world. So say, like l. A so Paulo Mexico City. Lunden. No law against Nigeria. Extra have 10 different cities. You got 60 different works that are you know what the market is for each of those different cities. The idea would be four different teams that are competing to bid on these different works of art and live and silent auctions. And you're trying to put together a cohesive collection that you could take on the road and make the most profit. Bring the most attendance, and it obstinately put together some of the best marketing plan on exhibition humanly possible.

So, like this is the product design called Llama Los Angeles Museum of Art. And really, the idea is that kids, not only are you like learning about the different works of art dealing with the ethical issues that are like that are involved as well. I mean, Caravaggio murdered someone like, Are you comfortable having his work is part of your exhibition? What about Queen Nefertiti statute, like the Germans stole it from the Egyptians and it's been to the Noise museum ever since. Like, are you comfortable writing that piece of art in Station on exhibition? Knowing that there, you know, there's a lot of there's a huge dispute over its ownership. So,

um, having kids you know at at the age of eight and older go through simulation in teams working collaboratively, having to figure out what their plan is like Maybe you really want starry night? You've done the math. You figure that, you know, over these, let's say, you know, we do like 10 years off of the art exhibition. You spend a year in each of the 10. You know, each of the city's what cities you're going to go to work. You make the most money. If this team does that, how does that affect us?

What if you lose starry night in the auction like, what's your sort of backup plan? So having to manage all those different pieces of the project? Uh, it's just quite a magical what you get from the kids and you haven't the Constitution's going on all hours of the night. You got kids lunch conversations are about who did want it, how much it went for You make it all publicly available, so kids are always looking spreadsheets and checking out the financials. An airplane not only sort of their hand every else's hand as well. Um, it's just like, you know, that's an example of a project that isn't only memorable, but but one that brings out through the best of what we want kids to be doing in school.


This next episode is from a brand new podcast called Factually is Done by a comedian called Adam Conover, where he talks to exceptional experts revealing shocking truth and thought provoking new perspective. It's an investigative comedy podcast for curious people who never stop asking questions. It's actually pretty interesting. Here's what caught my attention on this episode. Adam was talking to a U. C L. A professor in the School of Law about Constitution and the Second Amendment, and they were talking specifically about the Second Amendment, as interpreted by us today versus how it was intended. Back in the day, basically, in 17 91 when the Second Amendment was adopted, guns sucked. So no one actually expected you to have a gun in your home. Never mind actually used a gun for your own protection. Here's an expert to explain this.

Yeah, no, I mean, the framers were really focused on militias. It's written right into the Second Amendment, and if you look at the founding era discussions about the right to bear arms, they talked about the right to bear arms, primarily within the framework of the militia. In part, that's a function of the firearm technology. At the time, farms were not very useful for self defense. At the time you couldn't store a loaded firearm in your home, for instance, the gunpowder was too explosive, so it would take you about a minute to CO to load your firearm.

So if you had a criminal climbing through your window, your firearm wasn't what you would use to protect yourself or your family. So that's not how we thought about firearms. But they were effective for malicious service. And as firearm technology changed and firearms became more useful for individual self defense, people started to re conceptualize the right to bear arms as about people's ability to have guns, to protect themselves and to defend themselves against criminals, which was very far from what the founders were thinking about. And how did that happen? How did we move from a culture where nobody really believed that the Second Amendment was about that individual right toe? One where well, now a majority of Supreme Court justices due, and now for all intensive purposes, that is what the Constitution now says, because the Supreme Court says that said that's what it says. Well,

I think it's the real product of a social movement to advance the cause of gun rights on that's been a slow, steady growth. I mean, like I say, part of it's the technology of guns changed on So, um, but for most of American history, the court said that the Second Amendment did not protect the right to bear arms or protected a right that was somehow associated with the militia service only and didn't have much to say about ordinary gun control. That really changed in 2008 with AH 54 decision of the Supreme Court. That said, the Second Amendment did protect right to keep and bear arms for individuals and have guns in their home for personal protection. And indeed, the Supreme Court said the core of the Second Amendment was personal protection against criminals or, in the case of confrontation, again, not what the founders were thinking about when they wrote The Second Amendment below the line has become one of my favorite podcasts.

It's straight to the point, it's honest, and it does agree job conveying the information that's otherwise taken people years to acquire. Here's James Bashar talking to Ryan Hoover from product Hunt about what it's like to start a company, and more specifically, what you should test first when you're doing it started some people also the kind of almost natural, too test out the things that they think you're gonna work first, like the things that they're most confident in working. But actually, in many cases you probably test the most difficult, most risky parts of the business first. And so if it's if that's what you're building a new D D C brand. If Thea you know, one of the rest is can you build a product? In many cases, that's not the biggest risk.

So maybe you shouldn't focus on validating that Maybe should focus on the hardest thing. Which is, Can you get users? And can you get them cheap enough for Can you return a positive? Are y in this purchase or the sale? Um, I think that's something that people maybe don't think about nearly. Oh, everything's going great. We've just validated the four obvious things that are gonna work. And the 5th 10 yeah, it's gonna be hard, but we'll figure it out later, and then you realize a two years passes and you're like we haven't figured out the 5th 1 And if the 51 doesn't get validated or doesn't work, then the whole business is broken, right?

Right. It's, um now 1000%. Last but not least, here's a bit of public service. There's a park. It's called Cat on Cloud, which is done by a couple guys from California who are on a coffee shop, and usually they talk about very different subjects. But today they're talking about patent lawsuit and trademark because Caterpillar incorporated $54 billion in revenue company as off last year is suing there for the trademark around the word cat, which is pretty ridiculous here, I'll let them explain. What Caterpillar is trying to do to us is to cancel our our trademark that we previously had applied for and been approved by the Patent Trademark Office caterpillars. Caterpillar Inc. Yeah, company that makes really large tractor correctly.

They I think last year they sold $54 billion where of heavy machinery and there, and we got a letter from them. Ah, August 4th 2018 letting us know that they were suing us $54 billion that's just in one year, one year. So they're doing okay. It seems


this is ridiculous.


That so when we got the letter, the first thing that struck me was Wow, a company is doing $54 billion in a single year. Actually has the bandwidth to go and find these frivolous things, dig them out and spend money. They're paying someone to talk to us. Yeah. Someone on their legal team is interfacing with us to try to get us to release the trademark. Too specifically. It's the word cat. Yeah, Caterpillar. Ah, you know the inside? Yeah, I feel as the insect like they have the trademark for the abbreviation cat, right,

which in my mind, is the abbreviation for the word caterpillar, not the feline animal corrects, but yet they're trying Thio. And they're doing this to a number of companies which we talked to in a minute where there essentially systematically, going after a number of smaller companies and trying to cancel those companies trademarks. And there we have It concludes this edition of smashed. You have told me that you really like this podcast that helps you find you an interesting segments. If you do, please go to iTunes. Apple podcast Google wherever you listen to your podcasts and leave it a review because that helps other people to discover this podcast. And the more people know about it, the more segments I can't produce. All right. Till next Monday. Thank you for listening.

share this