ETEC 540 Task 3: Voice to Text

I used the voice typing tool in Google Docs for this voice-to-text assignment. Below is what this technology captured, unedited, and my analysis of it.

The Story

This past weekend we had my daughter’s birthday party and I think it was a pretty good success we had 10 kids total and the plan was they were going to arrive and come over and do some crafts the first activity that we had was my birthday daughter was greeting all of her guests at the front door and show them where to go in the living room where my other daughter was waiting to teach them how to do origami she had found this really cool this really fun design where if you went to fold it if you hold it one way it’s a dog and then when you flip it to the other side it becomes a cat so that add some variety for people who prefer dogs or cats or most of them both actually once everyone had arrived we moved into the room with the long table and the first activity was painting coasters the reason this was the first one is because they wanted to make sure that the paint had enough time to dry so some of them chose to tape off designs others just did a free paint and yeah with 10 kids it was surprisingly successful not super messy once they finish that we handed out canvas pouches and Sharpie markers and they got to do Sharpie tie-dye canvas pouches and this was really cool I’d meant to look up the kind of scientific properties behind it but not entirely sure if a bunch of 70 girls would be interested in that so instead what we did was just had them colour however they wanted on their canvas pouches with these Sharpies and then once they were done we sprayed it with rubbing alcohol had a little bit of trial and error had a tiny little spray bottle at first and it wasn’t going to work very well so then we move to a larger spray bottle the kind that normally you would have kind of your cleaner in and yeah it works so much better so the trick was definitely to saturate the fabric with the rubbing alcohol and then it was fun right before your eyes you could see the colours melting in matching in the Fine Lines Blurred new colours came about and where created as existing ones Blended together and I think all in all most of them we’re really pleased with it not everyone chose to spray not everyone seemed to understand there was one person who was pretty upset that her design had faded after we sprayed it but yeah I mean that’s quick fix she can go back over on Sharpie and I’ll write her name again whenever she wants so once that was done we figured it was time for a snack we had a rainbow fruit and veggie platter some little pepperoni stick for them to a snack on and variety of juice boxes and then once that was done we did washer necklace is so I picked up some metal washers from the hardware store and got a bunch of nail polish just whatever we had around probably had about 10 or 11 bottles and yeah they just painted the washers and some of them did some cool design some of them just painted it one colour and the whole idea was at once that was painted with nail polish and had dried send they would be able to string it on some kind of a a row port wine that we provided with them they had a choice of colours and then they could add some beans as well so all in all they would have ended up going or the end up going home with her tomorrow call me some coasters canvas bag and and washer necklace the final thing that we have them do was of course we needed to keep with the craft thing so we brought out cupcakes confetti rainbow cupcakes with white icing and then presented those on paint palettes with candy in the 10 different spots for it until they all got to decorate their own cupcake however they wanted with all of the candy or not eating some of the candy on the side or not and yeah it was really fun to see those kinds of designs and I think we were able to get everything done everyone seemed to have a good time and like I said all it all I think it was a success.

The Analysis

The most glaring deviation from written English is that this is one massive run-on paragraph. I don’t use voice-to-text very often in my day-to-day life, so I forgot that one must actually say “comma,” “period,” and “exclamation mark” to add punctuation. It wasn’t until the very end that I remembered this, hence the final period at the end.

Because I was speaking off the top of my head and not from a script, there are a few places where I think the text is wrong, but it could have also been my narration style, such as repeating words (“kind of a a”) and incorrect tenses of words (“so then we move to a larger spray bottle”. However, there are also several places where I am confident the text is incorrect. Examples include “70 girls” (should be “7-year-old girls”) and “row port wine” (should be “rope or twine”). These mistakes remind me of the ‘telephone’ game I played as a kid (and that Gnanadesikan (2008) mentions) where accents, emphases and subtle slurs can make huge differences in whether the receiver correctly interprets the message to pass along. These, and many other blunders in my oral recount of the events, would have been rectified if scripted in advance, as a pre-conceived written version would have been more deliberate and precise (Gnanadesikan, 2008).

These errors could also be a suggestion that I should improve my enunciation, which is something I have noticed deteriorating since 2016 when I left a job that involved lecturing/class facilitating for 5-6 hours each week. If someone had been listening to the story, they may have misheard “beans” for beads but could have either used the context to figure out the proper word or asked for clarification. Similarly, if I were to increase my weekly oral presentation time, I believe some of these words would have been presented more clearly.

The use of technology here has taken what would otherwise be a collection of fleeting sounds and turned it into a visible, permanent record of my daughter’s birthday party. One aspect that I find interesting is that a written recount of this event would not exist had it not been an assignment. Sure, we would have told family and friends about the event, slightly adjusting details and incorporating emotional nuances with each iteration, but it is not something I would have otherwise written down unless I was trying to convey the story to someone I could not speak to directly. This is not because it was an insignificant event (it was my daughter’s first ‘friend’ birthday party, thanks to the timing of COVID); when asked to speak an ‘unscripted’ story, I automatically thought of a personal event that had strong emotions tied to it (joy, pride, exhaustion, love). And yet, none of these emotions seem present when reading the words captured with voice-to-text. Had this been scripted, or intended to be conveyed in written language, I would have included words to express emotions that I know my voice conveyed in the original oral story.

Reference

Gnanadesikan, A. E. (2008).The first IT revolution. In The writing revolution: Cuneiform to the internet (pp. 1-10). John Wiley & Sons. https://doi.org/10.1002/9781444304671

ETEC 511 IP2: Artificial Intelligence

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? (50 words each)

Alan Turing, a British mathematician and a ‘founding father’ of artificial intelligence (AI), proposed that humans solved problems and made decisions by applying reason to the information available to them (Anyoha, 2017). He believed that machines could show ‘thinking’ by mimicking the human process and that machine learning could occur by following “the normal teaching of a child” (Turing, 1950, p.22).

Building on this, John McCarthy, a computer and cognitive scientist, added that understanding how humans think was key to unlocking how to build problem-solving machines, but that AI goes beyond simulating human intelligence (“John McCarthy,” 2022). He sparked debate by claiming that AI means machines have ‘beliefs,’ referring to their ability to solve problems using question-answering and ‘if-then’ logic programming.

Herb Simon, a political scientist, connected AI to how humans make decisions, showing that they start with information and then follow a series of rules and used information processing languages to create logic and problem-solving machines (“Herbert A. Simon,” 2022). Simon highlighted that AI differed from human intelligence, as the latter still functions, albeit inconsistently, despite knowledge gaps and preferences (UBS Nobel Perspectives, n.d.).

Marvin Minsky, a mathematician and computer scientist, expanded ‘intelligence’ from procedural thinking to the result of several non-intelligent parts working together, creating the first artificial neural network (“Marvin Minsky,” 2022). Although critical of previous theories on human brain function, Minsky believed brains were machines that computers could copy. He believed AI would lead to machines outperforming people and strongly advised thorough testing (BBC News, 2016).

Timnit Gebru, a computer scientist, continued to highlight similarities between brains and machines, flagging flaws with AI like bias, racism, and ethical issues (“Timnit Gebru,” 2022). Gebru’s work emphasizes the need for representation not just in AI data and the individuals researching it but also to break down the systemic barriers within corporations that prevent diversity at all levels (Levy, 2021).

How do “machine (programming) languages” differ from human (natural) ones? (100 words)

Despite their similarities, machine/programming languages and human ones have notable differences. Programming languages are artificial creations where context is bound by pre-set rules (Harris, 2018). AI may ‘learn’ but it lacks morphology, the idea that word meaning can change based on context. Furthermore, human languages incorporate additional cues beyond simply words (body language, intonation, punctuation) to convey emotion and can use context to understand the meaning even when the words are unclear (mispronunciations), whereas programming language must be logical and precise. As Harris (2018) eluded, imperfections mean the code will not run while many humans possess the powerful ability to interpret and adapt.

How does “machine (artificial) intelligence” differ from the human version? (100 words)

Intelligence, according to McCarthy (2007), is the ability to achieve goals by means of processing information. AI is created with hard-coded rules, from which it cannot deviate, and operates within the confines of the designer’s understanding of a process at one specific time (Chollet, 2019; McCarthy, 2007). This makes AI excellent for collecting and managing massive data sets, processing data according to pre-set parameters, and completing tasks (Chollet, 2019). Human intelligence, however, is better able to generalize, adjust to nuances, and incorporate subtle changes in information like skin tone or cultural norms (Chollet, 2019; Hao, 2020; McCarthy, 2007). AI, therefore, is more about manipulating information and human intelligence is about trying to understand information (Hao, 2020).

How does “machine learning” differ from human learning? (100 words)

Machine learning is designed to make predictions about new information using coded datasets and algorithms (Heilweil, 2020). Unfortunately, when datasets are incomplete, which most are, it produces biased algorithms: varying accuracy rates and/or different decisions for different demographics such as sex, gender, and race (Buolamwini, 2019; Cirillo et al., 2020). These biases can result in “neglect[ing] desired differentiations…or amplify[ing] undesired ones” (Cirillo et al., 2020, para.58) that continue to promote inequalities and racism in systems governing human lives like education, recruitment, policing, and healthcare (Buolamwini, 2019; Cirillo et al., 2020; Heilweil, 2020). While humans are also biased, they often possess the flexibility to adapt to changing contexts and objective scenarios, which is not a feature of machine learning (Heilweil, 2020).

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? (200 words)

One of the first indicators that this was written by a human versus a machine is the (attempt) to connect the biography snippets in question 1. Words and phrases like “Building on this,” “expanded” and “continued” refer to the progress of AI in the context of information already included in the text on these specific people, not necessarily on the information on them found online. An article in Futurism suggested that acronyms and punctuation, like hyphens or apostrophes, can indicate human-generated content (Robitzski, 2019). In that light, the use of single apostrophes to imply additional context to terms understood by humans suggests a human author (e.g., “‘beliefs,’” “‘intelligence,’” and “‘learn’”). For fun, a sample from the above text was entered into a Giant Language model Test Room [http://gltr.io/dist/index.html] that highlighted several instances where the likelihood of a machine using that word was 1/1000 or less. These included adjectives (e.g., “notable differences” and “powerful ability”) and the word “morphology.” Another example is the use of the word “unfortunately,” which is included to convey an opinion. These examples suggest that this text is trying to represent emotion and a deeper connection between concepts, which are more characteristic of human intelligence (Hao, 2020).

References

Anyoha, R. (2017, August 28). The history of artificial intelligence. Harvard University. https://sitn.hms.harvard.edu/flash/2017/history-artificial-intelligence/

BBC News. (2016, January 26). AI pioneer Marvin Minsky dies aged 88. https://www.bbc.com/news/technology-35409119

Buolamwini, J. (2019, February 7). Artificial intelligence has a problem with gender and racial bias. Here’s how to solve it. Time. https://time.com/5520558/artificial-intelligence-racial-gender-bias/

Cirillo, D., Catuara-Solarz, S., Morey, C., Guney, E., Subirats, L., Mellino, S., Gigante, A., Valencia, A., Jose Rementeria, M., Santuccione Chadha, A., & Mavridis, N. (2020). Sex and gender differences and biases in artificial intelligence for biomedicine and healthcare. npj Digital Medicine 3, 81. https://doi.org/10.1038/s41746-020-0288-5

Hao, K. (2020, December 4). 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/

Harris, A. (2018, November 1). Human languages vs. programming languages. Medium. https://medium.com/@anaharris/human-languages-vs-programming-languages-c89410f13252

Herbert A. Simon. (2022, June 2). In Wikipedia. Retrieved June 5, 2022, from https://en.wikipedia.org/wiki/Herbert_A._Simon

John McCarthy (computer scientist). (2022, May 23). In Wikipedia. Retrieved June 5, 2022, from https://en.wikipedia.org/wiki/John_McCarthy_(computer_scientist)

Levy, M. G. (2021, November 9). Timnit Gebru says artificial intelligence needs to slow down. WIRED. https://www.wired.com/story/rewired-2021-timnit-gebru/

Marvin Minsky. (2022, May 25). In Wikipedia. Retrieved June 5, 2022, from https://en.wikipedia.org/wiki/Marvin_Minsky

McCarthy, J. (2007, November 12). What is artificial intelligence? Basic questions. Standford University. http://www-formal.stanford.edu/jmc/whatisai/node1.html

Robitzski, D. (2019, March 11). This site detects whether text was written by a bot. Futurism. https://futurism.com/detects-text-written-bot

Timnit Gebru. (2022, June 5). In Wikipedia. Retrieved June 5, 2022, from https://en.wikipedia.org/wiki/Timnit_Gebru

Turing, A. M. (1950). Computing machinery and intelligence. Mind, 49, 433-460. https://www.csee.umbc.edu/courses/471/papers/turing.pdf

UBS Nobel Perspectives. (n.d.). Herbert A. Simon: Nobel 1978 – Do we understand human behavior? https://www.ubs.com/microsites/nobel-perspectives/en/laureates/herbert-simon.html

Spam prevention powered by Akismet