CharonY

If we look at voting behavior of Trump as an indicator of making a mess, the clearest differential is between men and women. Trump was 10 points up among men and Harris 10 points up among women. Looking at Brexit data, men were 10 percent up for leave, whereas women were 2% for remain (https://www.statista.com/statistics/567922/brexit-votes-by-gender/ I do believe the age differential was higher here). For the Reform party, 12.9% of young men voted for Reform, compared to 5.9% of young women (https://www.kcl.ac.uk/news/gender-gap-strongest-at-opposite-ends-of-political-spectrum-in-the-2024-general-election.) In Germany, AFD was at 22% with men, compared to 16% with women. So clearly, we should disenfranchise penises until they settle down and make rational decisions.

It also the question whether you want to. Few papers are just merely a series of statements of facts, but rather they try to push the boundary of knowledge by providing a particular perspective, which are then underpinned by evidence. Whether that evidence is convincing or not (or require further discussion) is ultimately a value judgement. Fundamentally, you can think of scientific papers as a discussion among peers and if AI takes over, it is going to be a discussion among algorithms. Regarding the use of AI as support, I could imagine that they could help in finding additional literature, given the exploding (and also deteriorating) literature landscape. However, I can see that backfire spectacularly, as traditionally experts in the field will know the relevant lit they want to see cited. AI would move the discussion more likely to an amalgamated consensus, with no discerning voices.

There has been somewhat extensive testing in Universities, for obvious reasons. However, for pretty much all tools the conclusion was eventually that they are not good. They have high rates of false positives, as you mentioned. But in addition, even minor is sufficient to confuse some of the software. Our school has abandoned that attempt to prevent plagiarism and I think that is or starting to be the overall consensus.

I am very much looking forward on how it is implemented and what effects it might have. Anecdotally, pretty much every of the top students I talked to had some limit on screen time or were older before they got their first cell phone.

You mean the guy who likes to snort cocaine from toilet seats and indulging in sewage baths?

Oh come on. Just a tiny portion to munch on. Barely a bite, really.

There is actually quite a bit more data regarding that, as well as a proposed mechanism. One of the key factors regarding dementia is neuroinflammation. And shingles is, of course an infection of neural tissue. A range of previous studies have shown that shingles is associated with dementia. See e.g. Chen et al., 2018 Clin Psych. There was another paper early last year (I think the first author was Tang, but cannot recall the journal) also showing reduction in dementia risk in vaccinated groups, but also looked more closely on whether folks were diagnosed with shingles. Together, the lit makes a compelling case that neuroinflammation, in this case caused by a virus, is associated with dementia, and preventing such occurrences, e.g., using a vaccine, reduces dementia risk. Outside of shingles, there is also literature showing that vaccinations in general might be associated with reduced dementia risk (including for example influenza, tetanus, diphtheria etc.). Suggesting that beyond protection of specific neural infection, there might be another effect at play. The main hypothesis is modulation of the immune response and IIRC, the idea is that vaccinated folks have a lower inflammatory response to an infection and in the time period after an infection. Thereby, they lower overall inflammation, thus reducing dementia risk. That link is AFAIK not fully validated, but in my reading the evidence is somewhat compelling and does explain some observations made in dementia research.

Gotcha If I am allowed up to 10 minutes, my go to is usually some sort of pasta. When I cook pasta I always cook too much and most of the time I will have some in the fridge. If I want a snack, I'll do a quick carbonara (especially if I got pancetta, otherwise bacon, but do not tell the Italians) or cacio e pepe or, especially in the winter, put them in a miso soup.

Wait, how do you make flapjacks in 2 minutes?

I think the difference here is that traditionally new developments or tech have made certain tasks obsolete, which often is not a bad thing. However, one of the stated goals of AGI and similar movements is to make human thinking obsolete.

The great vowel shift of Ni!

Sounds right. Knecht in the military sense was often used to describe levels of foot soldiers, but got elevated to the equivalent of Ritter (mounted but also usually associated with nobility) in English.

On the phonetic side I always found it funny that some words in old and middle English are easier to understand for folks from other language backgrounds rather than native English speakers. As mentioned in the blog this is due to the use of loanwords that were closer to their origin first and then got changed (often in a French way) over time. One of the words I was always wondering about the origin was "knight" until I saw that in Old English it was pronounced "cniht" (knɪçt), i.e. similar to the the German "Knecht". Pretty obvious in hindsight, if you pronounce knight the Germanic way.

Well, this how the tech industry responds: CompanionEinstein - AI Education CompanionEinstein logs into Canvas and does your homework automatically. He has his own computer — he can watch lectures, read essays, write papers, and participate in discussions.I will also note that what I am talking about are not media reports. I am talking about the experience of myself and colleagues over the last decades (social media), acceleration of trends over the pandemic and the rapid adoption of AI to get out of work by students. While it is possible that things will level out at some point, we are still in the midst of figuring out what is going on in the first place. Even more worrisome, I don't think that we are even starting to address the issues in education in younger folks starting about a decade ago. I.e. we are already a decade behind and now things are accelerating in the wrong direction. For many years I have been wondering whether that is just a normal generational change in attitudes. There are pathways where ways of learning might change, but the final output (e.g. academic work) still largely maintains its quality. I will also say that the simple fact that education has been expanded will result in some decline in quality but should be level out at some point. However, all indicators point to a sustained decline ranging from basic skills to mental health. It has come to the point where the benefits and purpose of academia becomes questionable. And that trend is not entirely AI driven (but will likely accelerate because of it). Underlying everything it seems that GenZ seems to be the first generation to be less capable adopting to new tech then the previous generations and at this point I think the data suggests that we are looking at something different happening to our brains than what the previous technologies did. I also think that this development is happening because we keep imagining how great these tools are going to be for productivity and are not sufficiently clear about the human side of the equation. Most importantly, I don't think that leadership, neither in academia, business/tech or government is seeing the value of proper deep development of human intellect and are only to happy to outsource that task to boost the economy. Or I might just be grumpy because I will have grade exams that are barely intelligible and will have to curve so hard that the space-time continuum becomes infinite.

I am not sure what the current status is, but interestingly there were some cafes with robots, which were remotely controlled, in part by elderly folks. Only somewhat related, but I have read somewhere that the Chinese strategy for AI is very different from the US. The latter aims to generate AGI with the assumption that it would create an enormous strategic advantage in almost all realms (how that would work is a bit unclear to me, most of the articles I read are a bit like sci-fi mixed with handwaving and not terribly concrete). The Chinese strategy otoh seems to be more about implementation of AI in somewhat more specialized tasks, such as improving fabrication (they are already the most automated workforce, from what I understand). If that happens, it will require yet another generation. The two youngest two generations right now are struggling. As I mentioned before, the issue many of them are seeing is that they only have a vague concept of live before the internet and social media has become the single most dominant element in their lives. I have likened that to the fish in the water issue, where younger folks struggle to define their position to the internet as its absence has become inconceivable. The fear that they have now there is that AI is going to take over the internet pushing them even out of those spaces.

I think one of the issues your touching on is that getting to the point of understanding requires often boring repetition. However, we have done everything to eliminate even a second of boredom and thus these the boring tasks are becoming and unendurable task for many. I have students which I would consider decently intelligent and they really do work hard, in terms of investing time. But by the time they come to university, many basic abilities have not been developed and things that would have been routine even 15 years ago, require massive time investment on their part, often resulting in immense stress, which many are also not equipped to deal with. The number of folks having burnout symptoms just from regular coursework is just too high. There is little wonder why relief from AI is so attractive, though.

I am not too worried about specific exercises, or skills. What worries we most is the decline in fundamentals, which to me ultimately boils down to reading, or perhaps in a more abstract sense, information comprehension. In my mind, so far other media (video, audio) are not as well suited to transmit complex information. The only exception I can think of are oral traditions. However, cultures who used those have specialized Knowledge Keepers with highly focused training to retaining, contextualizing and transmitting information. It would be the opposite of how it is used in modern times.

When I learned that (not today) I also learned about the abuse Judy Garland suffered starting as a child actor (including groping and repeated solicitation for sex from executives). Between asbestos and the abuse stories, one might start to think that Hollywood might not have been a healthy environment, especially for kids.

An issue is that AI is not a human mind. It does not think like one and it mostly simulates. Avi Loeb has dubbed it "Alien Intelligence" to make the distinction.

There are part of curricula already. However, I do feel that there is a disconnect how some educators and especially administrators frame it and how it affects learning in practice. Essentially there is an emphasis on academic misconduct and plagiarism to counter the use for entirely AI-generated works. And then promoting beneficial uses. Yet quite clearly, we can see that it has little (positive) impact. That is already the case, though oral is problematic for large classes and usually has a high level of complaints regarding subjectivity. Furthermore, there is strong administrative push to let students pass, which is a long-standing problem. Essentially if too many students fail, the assumption is that the prof is at fault, rather a decline in learning abilities. This has led to at least a decade of grade inflation. These are already being done, but frequently if you poke them for more details (see a) we can see massive gaps. Together with a system that disincentivizes failing students (largely because of tuition) and the fact that students fail at building basic skills, there is not a lot material to work with. The easiest ways for educators is to go with the flow and we are already seeing that in high schools and increasingly at universities. Thus, the overall issue here is how do we disincentivize the easy ways of using AI and promote the better ones. Students frequently do not see the benefit in wasting their valuable time on social media by doing homework or exercises.

I had a discussion with younger students and one of the things that they noted is that they feel unequipped to navigate the whole social media and now AI landscape as they lack the knowledge gained compared to folks who grew up in the "before-times". What is happening now is the opposite. Quality of students goes down, number of folks going through the system increases. Few, if any imagine being able to outperform AI.

it is mostly an issue for online exams (there is a big push towards online learning, due to a) ability to reach more folks but also b) to save money. Administration has discouraged in-class work and want profs to provide a more interactive experience. As such lengthier writing assignments (for credit) have become take-home assignments. Short-answer tests are still in class most of the time. But not only on the usability side, whole purposes change. Facebook marketed itself original as a privacy-driven sharing platform when myspace was relevant. Once they got a monopoly their whole purpose was to extract and sell client information. I have no doubt that this will also influence AI implementation. What is also somewhat troubling to me is that folks cannot even predict the strategic risk or benefit of AI or even AGI. Maybe I am reading the wrong articles but there is enough wandwaving there to power the whole endeavour via wind energy. I am also curious about how efficient the regulations are going to be. Parts of EU do have stronger privacy laws, but often legislation fails to keep up with tech.

Coercion by definition is caused by some sort of power differential, and factors pertaining to can include age, but also threats or use of force. So as you mentioned, there is no age that immunizes against that. But the point regarding age is that clearly folks are more susceptible and less powerful the younger they are. Again, the differential in power is what is relevant. What in this spectrum is considered appropriate will be up to society to decide. The concept of coercion and that it is inappropriate in Western societies is a rather recent development, for example. Legal age of consent have varied a fair bit as have societal attitudes. Up until very recently a power differential between men and women was assumed to be the norm and there have been long struggles to address that, or to move the scale back towards it. AFAIK there is no universal standard that one could simply apply, even within a given society. That would be a very pragmatic criterion, for sure.

I agree on that, the issue though is that it is an universal tool and while it can help you to get deeper into things, it can also be used to bypass the elements that take effort. Often, this is a good thing, but the the issue arises if the bypassed skill is actually a fundamental one. Most students write reports from articles without even reading summaries. Others, are more active, and have for example created an audio summary, that they listen to while they are doing other things. While the latter is at least somewhat creative, it is also ineffective and in-person neither group is really able to hold a normal discussion on the topic that they were supposed to write and think about. Personally, I think it boils down to lack of basic reading and comprehension skills and while I think there is a way where AI can assist in that, I think on balance, folks will choose the easy way. It like having a tool that at the same time invites you to work out or to relax and enjoy yourself. Given the chance most folks will do the latter and I wonder how we can convince folks to choose the former. That is for sure, but similarly to social media I still think that it shouldn't be up to the companies to decide how it is implemented. Because another thing is for sure, they do not have the betterment of mankind in mind. That is a very interesting viewpoint and makes sense from a corporate viewpoint. I am reminded of studies among children and young adults, where the majority have indicated that social media has been a detriment to their mental health and that they see it generally as a bad thing. At the same time, the overwhelming majority won't or is unable to stop. When I read those studies, I was reminded of typical addict behaviour. I am not sure why you think we won't lose scientists or philosophers. There is a pathway to that actually. In academia (where most of the research is happening), focus on vocational aspects and certificates rather than quality of education. That opens the way to reduce faculty and replace with sessionals with AI support. Then, rather than broad funding of research, focus narrowly on priority areas and train AI to address them. Over time, areas where AIs are still weak or yield poor results (there are molecular biological areas with huge gaps where, from all I have read, AI is massively underperforming), will be considered non-priority. That shift will save millions if not billions, which is what most folks will focus on. The role of scientists will be limited to what probably is already starting to be the case with senior devs (i.e. agential supervision rather than leading research groups). There will other AIs on top of those. Students will lack the boost to climb to the first set of shoulders. Considering how the US government is wielding AI, do you really think that there will be short-term benefits beyond boosting the AI-economy?

Inspired by a comment from StringJunky and to avoid derailing a thread further, I started this topic to explore the idea how much of the challenges of AI is real and how much of that is just an outdated viewpoint. SJ mentioned that calculators and their introduction has likely led to some reduction in arithmetic abilities. The rise of wikipedia and search engines have reduced the ability of folks to look for obscure sources, especially in libraries to some degree. But while they have reduced certain skills and abilities, they enabled other, and often more efficient approaches. I.e. the much feared loss of overall competency did not manifest, rather it resulted in a shifts of abilities. But from an academic/educational viewpoint, AI feels different and I have struggled for a bit in figuring out whether my view is just old-fashioned, or whether I am looking at a valid problem. I will preface that the COVID-19 pandemic has accelerated an already ongoing decline in academic abilities in young folks almost world-wide. My personal opinion is that decline in literacy is a key element, as this skill affects almost all higher academic abilities. Now, in other threads we have explored uses of AI, and how it could for example be a personalized tutor. At this point most leading LLMs tend to be decent in undergrad-level topics. However, it does not seem that it is used as such. Rather, most students use it to bypass the learning process entirely (similar to the example of AI writing emails to be read by AI). Once they arrive at Uni, many struggle with basic comprehension, and as a consequence, it is very difficult to teach advanced concepts that build on simpler ones. They struggle to see connection between those elements and if they memorize advanced concepts, they cannot use it to extrapolate ideas from them. Perhaps unsurprisingly, they often also struggle to explain what their (likely AI-generated) reports actually mean. We are thus again at a point, where tools seem to make certain skills obsolete. The issue I am having is that the skills in this case are not specialized, but very fundamental how we think. Folks have trouble reading, and there is little evidence that modes of media consumption (e.g., videos) can fill that gap (most studies point to the opposite). Folks struggle with connecting ideas and synthesize information and increasingly offload it to software. But without that ability, I do not know how higher learning is possible. In short, if we offload all that mental tasks that, IMO ultimately makes us human, what else do we have left? Where is the space that the human mind still can prosper? Again, I do think that there are scenarios where AI is being used to better one's mind. But the fact that the incentive is to offload, most folks will. And if the system is geared towards that what would be a possible best case scenario? I could imagine regulations and other means to reign in AI for certain uses, though our track record in regulating tech has been abysmal at best. In short, I would like to discuss whether my perspective is just too skewed and missing elements and if so, which are those?