reading your post as a stat major gave me a bit of a twitch when I had to read about them again in your response! im only half joking. your post also talks about how you seek to make statistical terms more accessible, where the language used in a lot of statistics-based predictions can often times be exclusionary. this is true! one cool (but also very bad) thing that may be relevant in your future discussions is highlighting the fact that even amounts researchers and statisticians, i feel like these terms are used, but also represent inaccessibility in a way. it’s sort of hard for me to explain, but nowadays published research only shows p values below 0.05alpha (stats standard), which is a typical sign that results are statistically significant. this is interesting as the p value is commonly adjusted in order to make research look significant, which represents a larger issue on how many researchers see certain p values and confidence intervals as confirmation of results rather than having the added nuance that all research needs. when looking at making stats language more accessible, I think making sure this nuance is baked into such definitions is super important!

While both the New York Times and The New Yorker articles that we read contained information that was not too surprising to me, I found myself taken aback by the difference in tone between the two articles. At its core, the NYT article had a friendly tone in an attempt to explain AI to people who were vaguely familiar with it. While this structure is very informative, the article was truly an introduction to AI, at a point encouraging readers to sign up for chatbots like ChatGPT. This tone was in stark difference to the more cynical tone seen in the the 2015 New Yorker article titled "Doomsday Invention" on Nick Bostrom, which centered the opinions of many AI researchers with near-apocalyptic views on AI and its implications, at one point stating: “The A.I. that will happen is going to be a highly adaptive, emergent capability, and highly distributed. We will be able to work with it—for it—not necessarily contain it.” Given that Boström's quote stuck with me the most, I sought out to look for articles that talked about the relationship between AI and work, with a focus on one of the concepts seen in the NYT article for added context. The concept I chose from the NYT article was self-generation, where the first thing that came to mind was to focus on Midjourney, an image generation algorithm. The article I chose was "Is AI art stealing from artists?" by Kyle Chayka, also in The New Yorker, which begs the question as to whether or not artists can work with it or against it. But briefly, the article partially agrees with Boström's point in two ways: 1. "The A.I. that will happen is going to be a highly adaptive, emergent capability, and highly distributed": in Chayka's article, we see A.I., in this case Midjourney, become highly adaptive by being able to self generate image based on a text prompt. We've seen it skyrocket in recent history with over a million members with accounts to date, distributing images on demand. Take, for example, this (fake) image of the Pope that went viral last week: https://www.cbsnews.com/news/pope-francis-puffer-jacket-fake-photos-deepfake-power-peril-of-ai/ 2. "We will be able to work with it—for it—not necessarily contain it.”: in Chayka's article, we see people "work" with Midjourney by making images within it. I think his correction of for it was interesting in the context of this example as those who use the generation tool truly work under it while in tandem with it, given that he bulk of what generates values in their labour (creating the artwork) is not done by an individual at all. I thought these points are interesting as it showed that Boström, for the most part, got it right in his opinions.

Your point about needing real life models to not be pristine due to nature's imperfections made me think about how important it is to delineate the difference between seeing imperfection models and prediction though. In other words, while I think that it is true that models that model life themselves may not be perfect, it's important to say that models themselves should try to approach as perfect representations of imperfect things. lots of words there!

Your point wondering about how biases may affect how people make self assessments to maximize their possibility of getting a better outcome made me think about how the same concept could be applied to people's perception of events of which they asses instead of the questions they ask! In other words, I think it's likewise plausible to say that biases (or external factors like hope, goodwill, etc.) may affect one's perception of an event to align with preconceived opinion, too. The first thing that I thought about that would support both of our points is the availability heuristic, which refers to people's tendency to rely on readily available, typically recent information when making decisions (something talked about in Behavioural Economics, which was my subject area!). In the context of predicting the future, this bias may leads people to focus on recent events or vivid experiences when making predictions, creating that bias you talked about and further altering what reasoning looks like, rather than considering the full scope of information available at the time. For example, someone who has recently experienced a setback in their career may be more likely to predict further setbacks in the future, even if this is not necessarily likely. Tying this back to the context of subconsciously chasing after accurate predictions, I think the existence of cognitive biases are pretty interesting as it means that, when reasoning is done in a largely qualitative manner, it's hard to have a standardised process that can truly be the same two times over as we typically think of in terms of processes like the Padua Rainbow!