Episode 15: When Philosophy and Python COLLIDE! - Part 1

Episode 15: When Philosophy and Python COLLIDE! Wins of the Week Sean Tibor: Hello and welcome to Teaching Python with Kelly and Sean. This is Episode 15: When Philosophy and Python COLLIDE! Welcome. My name is Sean Tibor. I'm a coder who recently began teaching. Kelly Paredes: and I'm Kelly Schuster-Paredes and I'm a teacher who recently began coding. Sean Tibor: How's your week going Kelly? Kelly Paredes: It is going remarkably well. We had a busy week last week and start off busy week this week. And so it's starting to level out on Wednesday, you know hump day so good for that. How about you? Sean Tibor: It's been good. It's been a lot of training over the last week learning a lot of new things both inside the professional development courses and outside. Wins of the Week Did you have any big wins this week anything that you want to share that went particularly well inside or outside of the classroom? Kelly Paredes: Well, I'll let you talk about the win probably for training because that was a good win, but \ I have to say that the pi code was pretty cool that you helped me. Sean Tibor: Wait, wait tell me what tell me more. So you made a Jupyter notebook, right? Kelly's Jupyter Notebook Kelly Paredes: I made a Jupyter notebook again, because we've been using that a lot and it's kind of like a fun project. And one of my colleagues wanted to do something for Pi Day, which is tomorrow, Thursday 3/14. Well, one of the days that we celebrate Pi Day and unfortunately wasn't there to teach a class. I wanted to develop a program that the kids could use in order to guess the numbers without searching on the internet. See how many numbers they can recall of the pi. So I wrote a code uhm that allows the kids to guess every five digits and yeah, that's pretty much it was a good win and then you help me clean up the code with a little bit of a more pythonic and we're actually tweeted that out. You tweeted that out for me. That was really nice of you and hopefully you guys can follow me on that and get that. Sean Tibor: Yeah. It was pretty cool. uhm It's not as easy as it sounds to try to come up with the number of digits of pi without having rounding issues. So uh we ended up going through making it a cute little like looping program uhm and I think it worked out really well because it's simple and it's easy to understand how the code works. So in addition to running the code and playing the game, students can also see how you wrote it, which is a good way for them to learn. Kelly Paredes: Yeah, and I we used f-strings right with we use 1f string Sean Tibor: We used a little bit of f-strings and some uh the MP math library, right? Kelly Paredes: That was pretty cool. I was that was a good find. Yeah, I actually uhm wrote the code for the found the code for 217 digits, which I can't imagine anyone memorizing but you never know. Sean's Alexa Skill Sean Tibor: Yeah. Well, we have some pretty impressive students around here. Maybe they'll try to tackle it. So for me on my side, there's been a couple of wins this week. I'm actually going to save the big part of it for our conversation our deeper discussion but something that was a win for me was learning how to write an Alexa skill using uh AWS Lambda just to just to try to have a little fun and teach myself uh things. So our school is having a dad joke competition among some of the male teachers and administrators uh in our in our school. And so they asked me to come up with a dad joke. So in addition to making a dad joke, I wrote an Alexa skill that will tell you a dad joke on command. So it was kind of fun to figure out and just uses a simple API that's publicly accessible to pull down these dad jokes and then serve them back up, but it was really kind of fun to learn how it all works and connect all the intents together. I have some plans for what can be done next with it to make a private skill for our Middle School that can answer more questions than just "What's your favorite dad joke?" Kelly Paredes: You want to test that out on air? Sean Tibor: No. Kelly Paredes: We're learning about the inconsistencies of talking to Alexa and Google. Sean Tibor: Even though it's been a really frustrating experience at times to learn the ins and outs of it. It definitely is that feeling of victory that you have when it lights up for the first time and actually responds to your question instead of saying, "I don't know how to help you with that." Kelly Paredes: Exactly. Someone told me to say, please so I tried that too. Philosophy Meets Python Sean Tibor: Well, so that's an interesting thing that leads us into our next conversation, which is: the things that people do with inanimate objects. How we start to see our students saying please and thank you to our voice assistants in the classroom. Why do they do that? What's the the reason for that? Why do we as humans start to ascribe these human behaviors to the machines that help us out on a daily basis. So that leads us into our next conversation and this really started we were started talking about this last week. Last week, Kelly and I had the opportunity to attend a two-day workshop led by professors from the Machine Perception and Cognitive Robotics Lab at Florida Atlantic University, which is just maybe 15 or 20 miles away from us. So we attended this two-day workshop and it was all about machine learning and deep learning particularly. And one of the things that came up in the conversation that we're having was around the philosophy of artificial intelligence and the philosophy of computer science. And with all of these great things we could do with machine perception, computer vision, robotics, what were the philosophical implications of that? And it was really fascinating. I mean we were deep into Python and math and neural networks and convolution layers and things like that. So it was a deep heavy session in a lot of ways but one of the things that stuck out to us was these philosophical questions Kelly Paredes: I think for us I think that was the easy for me at least the easy connection of really, uhm, "what are we doing?" We have become and this is something that one of the professor said is: we have become the translators from what was in the past to what is now. AI is brand new to all of us. So we're the people that were not around when AI was really in our face as much. It was around, obviously. It started in what 1945? In fact, some of the earliest machine intelligence work was done in the the 19th century, right? So there's been some of these neural networks and everything that a lot of the modern AI is based on, and modern machine learning is based on, for over a hundred years. But the pace of development has really accelerated over the last 10 really. Kelly Paredes: And it wasn't something that was like on the news or flashing in front of your face. "And here's the AI!" So we're like the ones that are the translators of: here's what it was before Google. Here's what it was before we had all this data and we're the ones that are going to be teaching these kids and helping them to realize what it is now and what that means for them. So I thought that was really interesting and just kind of thinking that it's going to be a bigger impact on the kids than just the TV, just a smartphone, just you know, should I say a "home device" or another device as such. It's going to be something that is just out there for the kids and it's a lot to think about. Deep Fakes Sean Tibor: And as Educators and especially as Technology Educators we have a unique position to be able to help with this. One of the examples given was "deep fakes." The ability to create photographs or video or audio that sounds so much like someone else to be able to fake or train a model AI to imitate someone else that our students are going to face that. How are they going to be able to tell what's authentic and what's real versus what's faked? The ability to fake things, even over the last six months, has grown exponentially. Kelly Paredes: Yeah, and it's even the faking of the writing. I just saw an article today about how they're looking at ways to use a bot that can write text and write stories in order to identify fake stories. So that the the whole concept of we are now at a point where we need to be able to identify fake photos. We need to be able to write programs that can not only I guess make the fake photos but to cite them and to source them out. Sean Tibor: To detect them, yeah. Kelly Paredes: And it's something that philosophically and ethically and morally we need to think about. So not just be nice in the classroom, but now not just be nice on online. But now be nice when you're creating something new., The Academic Relationship Between Computer Science and Philosophy Sean Tibor: Right. So it was interesting to us. There's the ethical questions, a subset of philosophy, but one of the things that I wanted to just reflect on for a minute was that many of the underpinnings of computer science actually come from the study of philosophy. So one of the unique things that I had the opportunity to do when I was a freshman in college was to have a freshman seminar course, so I went to Carnegie Mellon which is well known for Computer Science and Robotics and AI and Vision. There's so many things going on there. This freshman seminar course was actually taught out of the Philosophy department by a professor who was extremely well-versed in an expert in logic, induction, reasoning: all of the underpinnings of computer science. The theorems of: if A equals B, then B equals C. Transitive logic. Things that we take for granted as fundamentals in math and computer science, but understanding how you would actually have proofs of that. How would you prove that those things were actually true instead of just assuming that they are based on the underlying theorems. So it was interesting to me. That was the first place where I actually learned about Turing machines, and we actually made our own Turing machines and programmed them to be able to do simple calculations. It's really fascinating when you start to go back and model some of those out and simulate them or emulate them on your computer. The way that those work is essentially the fundamentals for the way modern computers work. We just do a lot more of that. And so I was reminded of that fact in our machine learning workshop because the underlying math is fairly simple, its matrix math. It's dot products and some basic calculus. But from there, we just do a lot of it. We do more of it. We approach it in a very smart way. So I feel like we're in the same sort of transformative step going from simple Turing machines to massive computers with lots of resources as we are in AI going from simple AI to really sophisticated, advanced AI that can do things we never dreamed of. Kelly Paredes: It's such a complex uh topic for me. You always think of: you took a Philosophy degree or you get a computer science degree. I did a quick search on how a lot more colleges are offering a joint or a double major in philosophy and CS. I was just looking and Stanford has a joint major: CS plus Philosophy and the whole idea is to not only blend the intellectual traditions of two departments, but they do it in a way so that you only have one requirement for that major because it's so important. And just looking at how philosophy teaches you to be creative it teaches you to look at what people believe in. And it's that aspect where you question the truth. I think by having that double major with those with computer science and philosophy is going to develop something maybe more humanly, you know, so it's been interesting. Sean Tibor: Yeah. It was really a fascinating starting question for us. Many of the things that we talked about or spoke about was how do we take these Concepts in Python, AI, computer science, machine learning, deep learning, and philosophy and teach them at a middle school level or a lower school level or upper school level, right? So how do we make them age appropriate and developmentally appropriate at each of those stages? And that's something that we haven't quite figured out yet. It's something that's still relatively new and changing so quickly. It reminds me of that whole Donald Duck cartoon where he's building the track in front of the train as it he's riding on top of it, right? Kelly Paredes: I think the major thing that I was taking away from that is that we have to constantly remind the kids that behind all these technologies a human-made that. And every time that they ask Siri or they asked a Google home or they drive in a Tesla or they do some sort of you know, I guess web scraping kind of machine learning aspect that human created that and that it has that human feature behind it and it's not the program or the computer that is being human. So it's something to really keep reminding the kids and to ask them, "how do you think that was created?" Machines as Creators Sean Tibor: Well, and then what's interesting is to go the step beyond that and how to start thinking about the machines are now creating the things that we're using. So one of the fascinating things to go through was that right now the humans are designing the layers of the neural networks. We've got some models that work really well. There's also a lot of work going on in machines that develop the models using genetic algorithms, so they are using a genetic style evolution of the models to train better and better models. Now we're getting back to another layer removed where it's not just that you can say the human is training the AI that is doing the thing that you want. It's: "the human is training the AI that creates the other AIs and models them to do the things that you want" Kelly Paredes: It's incredible. When they were explaining it and they were going into the Neuroscience behind it. It just turns another spin on everything for me. I went through the process with the kids earlier in the year with AI. We trained the computer to identify a heart and a diamond and a club and to me it wasn't really anything. I just thought you know here we're putting the algorithm. We were. We're just putting the algorithm. Here's a heart. Here's a diamond. Here's twenty hearts. Here's twenty diamonds and when we showed that heart , it comes up and says it's a heart. And you don't really think too much into that because that's at the basic part of AI. But when you start thinking about the things that it can do. Does it make a left turn? Does it make a right turn. Does it stop for a pedestrian? What type of pedestrian is it going to stop for? Is it going to choose? You know if a female's walking across the street or male's walking across the street or an elderly person's walking across the street, who's writing that program? It's not just a heart or a diamond or a club anymore. It's a life and do we hit a dog or do we swerve out of the way for a dog? And it's all these questions that you don't have the answers for and that we want the kids to start talking about. So that whole talk that they went through just had me asking all these questions. Sean Tibor: The biggest question for me coming out of those: now, how do we teach it? So where do we start? There's a starting point to this where you start with the fundamentals. You start with the basics of how humans learn and that's where we started with this workshop. We started with how we believe the brain actually works and our professors for this workshop were both computer vision experts. That was their field of expertise and interest and they looked at it with: "How do we see and recognize images?" Not necessarily, "how do we learn how to move?" "How do we learn how to talk or speak or any of those things?" But, "how do we process visual images coming through our eyes to the brain?" How does the brain interpret that make sense of it? And the study of that Neuroscience has been going on for a very long time. The way that modern neural networks and particularly convolution networks work at this point are to emulate that process in the brain, that Neuroscience in the brain as much as possible. And so I think that's a good place to start, particularly for middle school students so we can talk about simple things: about how people see things, how we learn to recognize new things, how our brain then processes that. And then how we make decisions based on that information. Kelly Paredes: It's a huge jump for middle school and that's the thing that was on my mind a lot. So trying to explain to a 13 year old or 12 year old that we see images that are inverted. They're picked up by our rods and cones and they're sent down these nervous system, these nerve cords and they're going to our brain and then they put together this other image and we say, oh that's Sean that's sitting there right there. How to Teach Neural Networks? And I think that's a huge jump and then to explain to them that that's kind of what happens when neuro networks is even a bigger jump. I was a little bit of apprehension going on when we were listening to these guys and explaining how they're similar to an AI into the brain. I think once they do kind of don't have to know how it works really so much but just kind of get an idea. I think our kids could pick it up. Sean Tibor: The good news is there's a lot of metaphors that you can use. So a lot of the discussion that we had was metaphorical to make sure that we understood it really well. Now there were definitely parts of it where the meth was way over my head and I hadn't looked at stuff like this for 20 years and I'm trying to keep up. As long as I can come up with a metaphor to do that it made sense. Kelly Paredes: I liked when he said, "I want to take a machine that sees X to do Y" kind of thing. I think going back to the kids and saying, "I want a machine that can identify dogs crossing the street so they don't get hit." I think that's something that a middle school student, even a high school student could obtain. Or, "I want to help a blind person find their way around the school." I think that's things that they can grasp. So if I want X to do Y then maybe you can start pushing them into that AI. Sean Tibor: Even before we get to that point though, it's the "how do we learn what things are?" That was the primary metaphor we used was the way that children learn things, learn to recognize things, how to associate words with objects, or how to classify and categorize things. The big one is the Dog vs Cat problem. So it's actually really hard when you think about it. It was really interesting to think through that problem. It's really hard to distinguish between a cat and a dog if you've never seen them before but the way we used to do machine learning was it we used to try to define all the rules. These are all the things, all the heuristics, all the elements that make up a cat and these are all the things that make up a dog. It's like the angles of the ears are like this and this one has whiskers and this one doesn't and and how would you even think of all of those rules ? The rules are exhaustive. You could spend your entire life writing rules to differentiate between dogs and cats and the big jump was when we realized collectively or there was a breakthrough when we realized that you don't have to tell all the rules because that's not how we teach children. That's not how we learn it. What we do is we say that's a doggy. There's the kitty. Pet the kitty. Pet the kitty. Pet the kitty. We reinforce over and over again that that's a dog and that's a cat. That's a dog. That's a cat and over time. Even when the kids get it wrong. It's no no, that's not a dog. That's a cat. They update that mental framework. We are training our brains , our children's brains, our brains. Every time we see something new to be able to recognize it. That act of recognition is the metaphor for the machine learning, the deep learning that we were doing, where we don't tell the computer, "these are all the elements that make up a cat." We just tell it, "here's the input. Here's the visual image we're giving you." Then we're telling you this is a cat and you figure that out in between. What are all the rules that say: this is a cat or this is a dog and so it's generating all of those rules and inferences that make it work so that in the future if you ask it, is that a cat or a dog? It just runs it through the model and it can tell you. Its 94% probably a dog. Kelly Paredes: And I like that because we always refer back or you refer back to comparing Python to music and I like to refer it back to learning another language. The same thing applies. We know how a piece of music is supposed to sound. How do we know it sounds that way? Well a music teacher might say well there's a pitch there's a tone but a musician will know that because that just sounds right or that's the way it's supposed to be. The same thing when you're learning a language. You can walk up to a hotel and ask for a loaf of bread and they're going to look at you weird when you really meant to say, you know, I want to get a room. One bed. If you constantly refer to that learning language, how do you learn as a child if you're not a parent? Maybe you have a younger sibling where you used to teach them how to identify items? I think that's a good way to keep it going. It just hits home for everyone. When you have to constantly remind them that we're not trying to build the program or not trying to build the program or not trying to teach specifically, in a certain way. We're not the holder of all knowledge of what makes a cat what makes a dog but we are just going to let you know that that's what a cat or dog looks like. Sean Tibor: It was fascinating as now having taught, to be able to apply all of the things about the way we learn and the way our students learn to this. It's interesting: in deep learning, failures are just as important as successes. That when you have something wrong, that teaches you about what to do next time and that's something that may be a metaphor for our students that they can apply in the opposite direction. So if they see that in machine learning or deep learning, that failure's important. Success is important. That each of them teaches you something about what you do going forward, then maybe that's a metaphor that they can apply to their own lives. Kelly Paredes: Just got me with a great idea. What if we can write a program. Here's that "what if?" What if we could write a program that highlights how many failures it takes in order to be successful? Can you imagine the insightfulness for anyone? Because we assume in in school, "oh, I got an A. I'm successful." But what is that success? However, how many failures did you have to have an order to get that A? How many failures does a machine have to have in order to identify a cat or a dog? Can you assess it and I know there's a lot because we were watching the epochs and we are watching the time to learn against the... What was the other name of the graph? The time to learn against?? Sean Tibor: It's the loss Kelly Paredes: The loss. Thank you. It was pretty interesting. There's a lot a lot of times that we test that. Can you imagine the breakthrough that you can have for a child? Say: "Listen, in order to be good, in order to be perfect, in order to have a 97 percent rate, not a 97 percentage but ninety seven percent rate of rightfulness. Or rightness, uhm you have to lose a lot. That's huge. Sean Tibor: Yeah, it was really fascinating to see the math behind it, the process behind it, the way the things fit together, and be able to think about: "Now how do we teach this? How would we get students to understand this at a deeper level and understand what's possible with it?" There were some resources that we used along the way that were really quite good. So there were some really good primitive resources that were more historical in nature. So, at AT&T Labs there was a video of a researcher who had trained a mouse to navigate a maze and built a primitive neural network from that. But once he had it learning the maze it would know how to navigate it or traverse it without running into any of the walls. And then he could change the walls and then it would hit the wall return itself. Learn the new path and go back. So it was really kind of fascinating the way that this research has been going on for a very long time. A lot of the elements have been there all the way through . We found some of those resources to be particularly valuable that would help demonstrate this in a simple way. There's some great tensor flow visualization. So showing how neural networks flow data from point to point. How they learn over time. That can be run right in the browser so you don't even have to download a program or write Python code, just it visually shows how these these flows happen. The other one that I really liked was the Visually Explained site. That had a lot of information about how regression algorithms work, linear regression in particular. Visually explained and so if you need some of these tools, we're going to post those on our show notes for you that help talk about these basic math concepts that go into the neural networks and then how you can use them in your teaching to teach the neuroscience and the philosophy behind deep learning and machine learning that may be more appropriate for your students at the level that they're at right now. Kelly Paredes: Yeah, I liked some of the other ones as well with just how they were doing the colorizing of the old photos. How they're taking the different renderings and changing photos to do different styles. Also, I found something called that Juke deck. It was creating unique music that was created artificially. So there's all kinds of little bit of a ways to do the jump in. I think the one of the things that the first part of the training that really helped kind of made me feel okay about introducing this into the English classes. There's just so much we can get into with the kids right away, with just looking at what is already done. What is already out there. So yeah, there was also an image clean up. I thought was really cool. Can you imagine if we could clean up all those old photos from the past in order to get a better outlook of history? I think there was a little bit of talking about that, getting rid of some of the scratches, putting colors. What did it look like back there in black and white ? Sean Tibor: That was really interesting to see how many different things can be done with the same basic network. It's really just a matter of training the model to do the thing that you want. And then once you have it trained, to reapply that over and over and over again. Why Computer Science? Kelly Paredes: I had a student this week was questioning why we had to take computer science. One of the professors who was doing the training for us told us a story that when students come in, he asked what their major is and they say, "English" or "Math" and he says, "perfect." So when the student asked me, "why do we have to take computer science?" I asked her, "Well, what do you think you want to do in life?" And uh she said to me, "I want to be a chiropractor. Do something like that in medicine." I had the three cases with the Alzheimer's detection using machine learning and brain scans, the bone fracture detection, and then the breast cancer detection. I had those graphs and I told her you know, listen, this is the rate of success when a doctor does a scan for breast cancer. This is a rate of success when a doctor uses AI for breast breast cancer, and this is a rate of success for someone surviving when the doctor uses AI plus his or her own knowledge in order to save the patient. What you see is that the doctor who can combine AI plus his knowledge was able to have what was like one percent failure rate or Sean Tibor: Yeah, like it was it was like a 10X. It was an order of magnitude better. Kelly Paredes: It was incredible. So when I said to her I don't expect you to do programming, if you want to become a chiropractor, that's great. I don't expect you to program for the rest of your life, but I do expect you to understand there's some ethics there are some things that you need to know that you need to be able to do and to investigate in order to serve your clients or your patient better in the future. If you don't know that then you're not the doctor that I want to have. She's like, "okay, I'll keep learning computer science." So it was such a it was such a great way to put in again. We keep saying this is why we have this Philosophy and Python when they Collide. Python's where it's at right now with data science and machine learning. We have to remind the kids about having that ethical understanding of what we can do for people. How can we do social good for the world using machine learning versus negative side of machine learning? Self-driving Cars and Non-driving Kindergarteners Sean Tibor: That was that was the big take away because we're talking about our kindergarten students. So my daughter is in kindergarten right now and the claim was made and I think it's not that far-fetched. In fact, it might happen sooner that my daughter will probably not learn to drive. Because there will be no need for her to learn how to drive. It will be safer for her to get in a self-driving car than it will be for her to drive herself. Having been a 16 year old driver when I was a kid and seeing all my friends, that's very believable. Kelly Paredes: It's not such a bad idea to know that my boys won't be driving cars either. Although they might? What were they saying that they're predicting an $8,000 a month insurance if you have an old car or classic car versus one of the self-driving cars? I was in Germany as a rental car and I did not have a Tesla, but I had a car that literally almost drove me to where I needed to go. It had sensors. It had a motion detector. It told me how far to be away from from the car in front of me. I started to swerve to the left because I was looking at some beautiful castle in the far off distance and the car re-corrected me. Can you imagine? We have those safety measures already in place. Sean Tibor: It's unbelievable. I mean if you buy a car that's even two or three years old. You might have bought something from the last century. Right like you might have bought something that is a little bit better than 1999 in terms of safety and comfort and features and affordability . But even a car that you buy today is a step change ahead of where we were two years ago in terms of the automation, the self-driving features, the convenience all of those things. It's incredible how fast this is moving and so to your point about the ethics of this: by the time our students are entering the workforce, which at this point is what probably ten years away? Eight to ten years away. It may not be Python. It might be Python 4, or 6, or whatever it is. It may be a completely different language. But the ethical questions will still be there. New ethical questions will arise and our role as educators is to help them think through those questions, have the discussions, to be able to have the foundations in technology and computer science to be an informed debater about those questions. Kelly Paredes: Yeah, so it's not necessarily like the code, right? What if someone came up with just the block program? Here, I'm going to use this neural network in order to train the difference between helicopters and jets. And just plug in that block and push it through and we're going to we're going to direct it here and let it go. So yeah, need to have some nice people out there. Sean Tibor: In the last four years, the power available to do machine learning has increased by 50 times. In four years. Kelly Paredes: Well, they say Andrew Ng from Coursera says artificial intelligence is like the new electricity. Sean Tibor: I don't know that I believe that. At least, not yet, but it may be a simple something that right now. Four years ago, it may have taken us a week or two to train a model that could tell the difference between airplanes and helicopters. But today may only take 15 minutes. If that curve continues, in another four years that may take 15 seconds. So if we start to get to a place where we can do this training that quickly it could be as simple as: "now use the the model that can to tell if this is an airplane or a helicopter and use that as part of my program." Train it up and it'll be done as part of a block-based code. So Josh Lowe, if you're listening, we have some work for you to do with EduBlocks. Kelly Paredes: I gave his name out to the instructors. Hopefully they contact because they were really interested and saying, "oh we have a new block. We want to do a block program of artificial intelligence. That's great." Sean Tibor: So what we're going to do because we wanted to divide this topic up into two sections because there's so much to cover here. So what we wanted to cover today was really just around the philosophy, the questions, the ethics of artificial intelligence and specifically, machine learning and deep learning and what that means for our students. Hands-on Applications Next week, what we're going to cover are the hands-on applications , really the reason we needed to do this is we have some more research to do. We've got homework to do before we can present back to you what we think might work and some things that we're going to be trying in our classroom with machine learning and deep learning. One of the things that was really reassuring for that though is that all of the work we were doing was using Google Colab and a Jupyter notebook and Python. So we're on the right track. If you're teaching Python now and you're interested in this area, you're using the right tools and there's so much that's out there and just freely available to start doing this work. So as a little preview of that one of the things that the FAU MPCR team put together for us is a Jupyter notebook that contains the method for classifying images from Google Image search based on a ConvNet model. So it's a really great tool. I'm want to be able to share that with you. So I'm going to be talking with that team to see if we can share that out with. But that's something where you can give the program two different search queries for Google Image search and let the machine train the difference between the two. I think you got to a 94% accuracy or 97% accuracy? Kelly Paredes: Is pretty high granted it was about wine. So... Sean Tibor: Yeah red wine versus white wine. Kelly Paredes: Like I didn't go. It was okay with red and white. So yeah, so there was a really neat quick program. We only chose the the categories. I was assigning variables. So it was something that was quick and easy. Sean Tibor: Next week we're going to be talking more about that. Some of those other resources that we found. Places where you can go to get more information. W e'll also share some information about the MCPR lab at FAU. They're doing some really amazing work there and we are so incredibly grateful for the time that they spent with us to learn about deep learning. Kelly Paredes: It's going to be good. We're going to also follow up a little bit about images and just looking at the comparisons of how computers and phones manipulate the images and make them blurry or sharper and how that's passed into machine learning as well. Listener Questions Sean Tibor: So now we're going to start adding a new thing to our show every week. So this is brand-new for the first time. We experimented with it a couple weeks ago. I have some listener questions this week that we are going to answer. So this is hot off the Twitter feed. If you have questions for us, please feel free to share them with us. Reply to us on Twitter. Send them our way. We love to see it we're at @teachingpython on Twitter, but our first question comes from our good friend Peter Kazarinoff. Kelly Paredes: Aww, thanks Peter! Sean Tibor: We interviewed Peter a few weeks ago. Peter teaches mechanical engineering at Portland Community College and has really done a lot with Jupyter Notebooks with his students as well and really transformed the way they've done things. So Peter asks, "does each student in your classes buy their own micro:bit? Or do you have a class set? How do you keep it organized?" Kelly Paredes: So that's a great question Peter. You always have these amazing questions. Currently they do not buy their own micro:bit, but we are looking and begging to put into our budget the opportunity for the kids to have it as sort of like their paper and pens for a class. For us, we think that's just going to be a powerful tool instead of purchasing a book that can cost us a hundred something dollars. We'll have them purchase a micro bit so that they can take it home. You'll still need to have a class set because just like you always need a pen and paper in your classroom because, for that child that forgets, I think it will always need a class set or at least something that someone can borrow. We had about 60 micro bits at the beginning but we go through them and we beat them up pretty bad. And so we printed it out a 3D model. Where did we get that from? Sean Tibor: I think it was from thingiverse. So it's a parametric model micro bit holder. So you can 3D print it. It has little slots that keep all the micro:bit's organized. So we printed out two or three of them so we can move them around as we need to. We basically have baskets that keep all the little bits and cables and everything that go with it. I know we talked a lot on our micro bit episode about different accessories and everything. The basics that we use are these short little micro USB cables that came with the micro bit go kits as well as some short six inch long USB-C to micro USB cables because we do have a lot of our younger students that are buying computers for the first time. They're getting computers at only have a USB-C port on them because we tend to be a fairly Mac heavy environment Kelly Paredes: The only thing I was thinking is you remember those old CD cases where they had like the hard plastic top? I'm hoping to find something similar like that to put in our 3D printed holder into our micro bit so that we can almost have our old-fashioned cd case to keep our micro bit in. So that's currently what we're doing. Can't get enough of those boxes or baskets to store things in. Moving them around because I just grab it have a handle on it. Grab it put it into my cart and we go to other classrooms with it. So great question Peter. Sean Tibor: For all of our other accessories that go along with it, like battery packs and other wires and alligator clips and things like that. We have a rack of bins that we can organize everything into so I think you can find them on like Uline or other sites where it's just the clip-on bins that clip onto the rack in the back and we can reorganize them and move them around as we need to. So it stays reasonably well organized. Especially since we have so many middle school students in here. It stays pretty well together, but we probably spend what 10 or 15 minutes a day at the end of the day just kind of sorting and organizing things again. Kelly Paredes: Yeah, but we have to keep it visual, you know out of sight out of mind. We don't want that to happen. So one of the biggest thing is to keep them out couple out on the table. And so that the kids can always come up and say what is this sir? What can do it this? Sean Tibor: Yeah. Yeah, so great question Peter. Thank you very much for reaching out to us. If you have a question of your own you can always reach us through our website, which is teachingpython.fm or find us on Twitter at @teachingpython. We have a form for listener submissions. So if you have a question, if you have a request, if you have a topic idea, if you'd like to be a guest on our show, please reach out to us. We're happy to hear from you every time our Twitter dings. It makes us a little bit happier inside. So please feel free to reach out to us. It makes it a lot of fun to to run this show when we know that we're engaging and connecting with our listeners. Kelly Paredes: Yeah, so stay tuned for part two is we're going to talk again, like we said a hands- on AI Python exercises what to do when students cross the line and some more resources will have those posted out next week. Innovation Institute Sean Tibor: So looking forward to it and also one last reminder Kelly and I are gearing up for the Innovation Institute at our school. So we have lots of topics. We'll be talking a little bit about Python, a little bit about microbits. We're going to be talking about how we try to keep things fresh interesting and innovative inside the classroom for our students and how to encourage innovation in their lives. So if you're interested in that Institute, I believe it runs in early April. So April 6th 7th and 8th of this next month. We will both be there and available for questions, meetings face to face. We're also looking forward to learning quite a bit from our colleagues and our attendees. It's a small conference. So there's a lot of face-to-face time. So I highly encourage you if you're interested as a Python educator to attend and and join us. Kelly Paredes: We have these what we are calling poster sections not necessarily poster, but if you can kind of think of it like an old science fair exhibit. We're going to do a poster about our microbit genetics project, our database with the Dragsters and just how we use Jupyter notebooks in the classroom. Sean Tibor: For those of you who can't attend we will post it after the Innovation Institute so that you can get a look at it also, and if you have any questions, we'd be happy to answer them as well. So I think that's everything we wanted to cover in part 1. As always, I love the conversation Kelly. Looking forward to next week. Kelly Paredes: Thank you. Sean Tibor: All right, this is Sean Kelly Paredes: And this is Kelly. Sean Tibor: Signing off.