Thursday, December 15, 2022
Monday, December 12, 2022
Saturday, December 10, 2022
ETEC 511 Group Project Review and Retrospective
In this design process we created a guide site to using a very complex tool, TinkerCAD, as a beginner to Electronics ADST teacher.
In this project we initially tried assessing how to use the Classes feature of TinkerCAD, but in the end created a tool that simply tried to improve the usability of that limited-LMS in the context of non-microcontroller circuits.
 |
| from xkcd.com by Randall Munroe https://xkcd.com/2141/ |
The first challenge was an inability to go into priscilla mode and build our content within the TinkerCAD - Classes platform. This was frustrating and forced us to use the next best thing which was a series of glued together pieces of work that we believed a beginner ADST teacher would find engaging.
In navigation terms we laid out a very simple linear method of navigating the site which was designed to be as minimalist as it could be to avoid attentional overload. The pages with lesson plan PDFs and pre-built TinkerCAD circuits were coded in such a way that clicking them immediately downloaded or created new tabs for easy access to those resources.
Our work was to configure the user to get them used to the experience a variety of mediums (website, video, Genialy, simulator). This is simpler to the experience of being an ADST teacher where you are frequently expected to be an expert at many systems in those roles. In this sense the project was a success. I believe that those multi-modal resources are well put together and beginner friendly. A late addition to the project was the site walkthrough video which helps introduce the user to the navigation, format, and content which reinforces the idea that this is a guide to curated resources specific to the target audience.
Using Issa and Isaias's (2015) Usability Criteria we achieved some of the following:
Learnability
The site has been designed with easy entry points in the form of video tutorials/introductions as well as a Genialy animation that provides the expected feedback to “hand hold” someone getting started with this complex topic.
Flexibility
This is a place our project struggles. Being a limited scope we were not able to add many branching connections on the site to provide flexibility.
Efficiency
This tool is meant to solve an efficiency issue as the amount of effort required to find resources like these in the dark corners of websites like Instructables is immense and the payoff is minimal. Using this tool to its ceiling (which isn’t hard to do) provides massive payoff in terms of an ADST class.
 |
| from xkcd.com by Randall Munroe https://xkcd.com/1425/ |
In terms of what I learned: Design for usability is a slow and tedious process, testing for it doubly so. If I had to do this project again I would have spent less time on fighting with having a functional product and more time on the proof of concept; it may have taken a different form but our process definitely illuminates the challenge of managing sunk costs in development and how it can negatively affect usability.
Overall, I’m happy with this project. I feel as though it presented well in the synchronous session and the site is effective despite some technical limitations to implementation. It is a resource that I as a beginner complimentary course teacher would have coveted had it been available ten years ago.
Issa, T., & Isaias, P. (2015) Usability and human computer interaction (HCI). In Sustainable
Design (pp. 19-35). Springer.
Tuesday, November 22, 2022
Case Study Assignment
 |
| From xkcd.com/1215 by Randall Munroe |
Sunday, October 16, 2022
Monday, September 26, 2022
IP 2 - Artificial Intellegence
(*All text in brackets is “flavor text” and is not included in word counts...including this header*)
Who were these people, and how did/does each contribute to the development of artificial intelligence? How did/does each think “intelligence” could be identified?
Alan Turing
The “Father” of modern computing. Turing was the innovative genius responsible for many foundational concepts in Computer Science including some of the earliest electromechanical devices.
His work on applying these concepts while working as a codebreaker during world war 2 helped the allies win the war in the Atlantic.
Much of his work was not public due to wartime restrictions and because of his tragic death in 1954.
John McCarthy
(Can we applaud the level of nerd it takes to get kicked out of Caltech for refusing to attend PE classes?)
If Turing was AI’s father, McCarthy was the parent who named it, along with three other names on this list (Simon, and Minsky) in the Datrtmuth workshop in 1956.
In addition to writing on the field of AI and its philosophy, McCarthy also solved a large number of technical problems including being the first to implement time-sharing, which is the backbone on which most computer infrastructure runs in the modern day (we know them as servers). (Wikipdia,
Herb Simon
Herb Simon is one of the early theorists of AI. He focused on decision making and leaned towards the social sciences. His push, the one that earned him a Nobel prize, was around the need to collect data before a decision can be made. With this he developed a theory for simulating the problem solving process This is a core tenant of how machine learning algorithms process information today. (UBS, n.d.)
Marvin Minsky
If Herb Simon represents learning by crunching massive piles of data, Minsky represents the Neural Network which relies heavily on multiple layers of machine decision making processes modeled on the human brain. Ironically, some of his work with Seymore Papert is understood to have changed the course of AI research erroneously away from neural networks due to pessimistic views on future problems in the discipline. (Minsky, n.d.)
(While you want to hear about Minsky’s contributions to AI, I also want to note him being credited with making famous the useless box)
Timnit Gebru
A modern example of AI progress to contrast with the white guys above. Timnit is credited with bringing discriminatory AI into the public consciousness.
As we will discuss below, machine learning is only as good as its dataset and Timnit was someone looking at how biased datasets could create biased AI. (as an aside, 99% Invisible did a great segment on this in their You’ve Got Enron Mail episode) (Hao, 2020)
How do “machine (programming) languages” differ from human (natural) ones?.
XKCD.com - Misinterpretation - by Randall Munroe (2018) (xkcd comics have scroll-over text. This one’s is: "But there are seven billion people in the world! I can't possibly stop to consider how ALL of them might interpret something!" "Ah, yes, there's no middle ground between 'taking personal responsibility for the thoughts and feelings of every single person on Earth' and 'covering your eyes and ears and yelling logically correct statements into the void.' That's a very insightful point and not at all inane.")
The communication flow between natural and computer language can be summed up by the fact that all computer language is meant to eventually be translated in machine code of zeros and ones. Each term is translated to a specific meaning when interpreted by the compiler. Most compilers will also refuse to compile language that it does not recognize. Natural language is far less logical. It includes multiple layers of meaning to interpret ranging from simple syntax, to variations in tone(when verbal communication is considered). It also complicates communication that natural language changes over time and has to be interpreted by the receiver no matter how error-strewn it is (though, as a teacher I do like to throw error messages at my students who communicate poorly much like a compiler does).
How does “machine (artificial) intelligence” differ from the human version?
“Intelligence has been defined in many ways: the capacity for abstraction, logic, understanding, self-awareness, learning, emotional knowledge, reasoning, planning, creativity, critical thinking, and problem-solving. More generally, it can be described as the ability to perceive or infer information, and to retain it as knowledge to be applied towards adaptive behaviors within an environment or context.” (Wikipedia)
Machine intelligence is currently good at abstraction, logic, understanding, emotional knowledge, and problem solving.
The major difference between machine and human intelligence right now is the human intelligence’s ability to draw novel connections between disparate pieces of knowledge. This is what makes humans much better at self-awareness and creativity, though recent events with Google’s LaMDA AI may beg to differ. (Collins, 2021)
How does “machine learning” differ from human learning?
XKCD.com - Machine Learning - by Randall Munroe (2017)
(Is it bad to keep using xkcd comics? You told us to me multimodal….)
Again, a bit big for 100 words. ETEC 512 is a whole course on the latter concept.
In short: machine learning is when you feed a lot of data into an algorithm so that it can recognize patterns. That algorithm can then be used to check a novel piece of data against the existing data to recognize patterns in the new data.
Human learning starts with sensory perception before that perception is committed to different synaptic branches of memory. Those branches are squishy and are not always reliable. When actions cause feedback, those things are re-wired into the synapses creating systems of learning that can explain parts of the more complex pieces of human thoughts and memories.
A machine learning system is only as good as its data. A Human learning system is only as good as its complex brain architecture.
And for your LAST challenge, a version of the Turing Test: how do YOUR answers to these questions differ from what a machine could generate?
Example AI Generated Essay from MyAssignmentHelp.com
The first question is: “Is the AI trying to answer these questions with the best possible answers, or the best answers to seem human?” (are we getting too meta with the Turing test?)
The above text is from MyAssignmentHelp.com when asked the differences between machine and human learning. Is it a mess, sure; is it assembling things within an imperfect set of rules like a seventh grader learning to assemble an essay? Also yes. This would be an example of an AI trying to be human. After all, this site exists so that students can convince their teacher that it is their work (the site emphasizes their premium, paid :Hire an Expert feature for that human touch).
There are layers to this question here. My original thought process is if the AI is tailor-made to respond in natural language with the parameters being “convince a human that I’m human while answering these questions”, I have reason to believe that it would answer in a similar way to me. What the AI produced was not nearly as convincing as I thought it would be and set my faith in machine learning back a bit.
Ironically, the AI answers the question about its own limitations: “Machine learning needs lots of sample data or data in general to learn and be able to find valuable information respectively results in patterns.”
And I love the Jerry Kaplan quote it found noting where human learning is far superior “...[when] there's no data, just some initial conditions, a bunch of constrains, and one shot to get it right” (Kaplan, 2016)
(footnote: - the AI pulled the name with the quote but not the full citation which could only be found via Google search on a textbook that someone at SUNY Oswego forgot to make private, I could only find it in German via the UBC library. A machine learning Ai is only as good as it’s dataset after all.)
XKCD.com - Turing Test - Randall Munroe (2007)
References
Collins, E. , Ghahramani, Z., (2021) LaMDA: our breakthrough conversation technology, Google - the Keyword [blog] retrieved from https://blog.google/technology/ai/lamda/
Hao, K., (2020) We read the paper that forced Timnit Gebru out of Google. Here’s what it says, MIT Technology Review .https://www.technologyreview.com/2020/12/04/1013294/google-ai-ethics-research-paper-forced-out-timnit-gebru
Intelligence, (2022 May 21). In Wikipedia, https://en.wikipedia.org/wiki/Intelligence
Kaplan, J., (2016) Artificial intelligence [textbook], Oxford University Press
Minsky, M., (n. d.) Brief Academic Biography of Marvin Minsky, Retrieved from https://web.media.mit.edu/~minsky/minskybiog.html
Munroe, R., (2007) Turing Test , xkcd, retrieved from https://xkcd.com/329/
Munroe, R., (2017) Machine Learning , xkcd, retrieved from https://xkcd.com/1838/
Munroe, R., (2018) Misinterpretation , xkcd, retrieved from https://xkcd.com/1984/
UBS (n. d.) Herbert A. Simon, Nobel Perspectives, Retrieved from https://www.ubs.com/microsites/nobel-perspectives/en/laureates/herbert-simon.html