otakucode

This is a very interesting read, but it causes me some concern. When I went to college, I majored in Computer Science, but I minored in Philosophy. A substantial part of philosophy deals with how arguments are explained and proven, and mathematics is a fairly direct outgrowth of that. Reading mathematics and computer science papers or proofs is quite different from reading more typical prose. They are often very rigid and extremely precise. A single symbol or word can change the integrity of the argument from unshakably true to entirely false. It requires much more care and effort to gain what feels like a true understanding of what they communicate - and that is an important and necessary part of their character.

There are many ways to communicate ideas, and most people in most situations rely not upon proofs derived through deductive logic. They rely instead upon persuasion, intuition, and building associations between ideas. Those tactics are in many ways more natural (as in, fairly direct consequences of brain structure rooted in biology) but with known flaws. LLMs are pure persuasion, and operate entirely in that realm. They can be cajoled into repeating logically unsound ideas in much the same ways humans can. And, they can do the same back to us.

My concern is that having one model generate a paper, and then using another model to explain and answer questions about the paper, could run the risk, especially in arenas where the number of individuals with deep deductive knowledge of the subject is small, of optimizing for persuading those individuals rather than proving real results to them. Since it's reasonable to presume the model has been trained on other explanations which have convinced the expert, and on convincing people generally (which, on balance, uses more flawed intuitive explanations to persuade rather than prove), it should be very good at least at creating an environment where the expert would need to practice extreme skepticism and fully utilize critical thinking to attack every element of every claim. That can be very challenging.

I mean, I would argue that they are right, but there is still a hard problem of consciousness. What properties must a system hold in order to exhibit consciousness? Personally, I think I could fairly capably argue that consciousness is an emergent property of any feedback loop composed of a large enough collection of highly interconnected adaptive components embedded within a sufficiently pattern-rich environment. Every feedback loop which meets these requirements is conscious, and removing any single one of those elements causes consciousness to disappear. I believe the property emerges in a similar manner to how 'heat' emerges from electromagnetic interactions of atoms. This definition has implications, and all that I have thought of are real. Total sensory deprivation results in loss of consciousness, for example.

That was my immediate first thought. "Oh, is Bitlocker Not Safe Anymore?"

Are we going to simply give up when it comes to understanding the fundamental nature of complexity? Why do multi-component systems undergo phase changes, why does it occur at the points at which they do, how could we design simple systems which, in sufficient quantity, exhibit useful properties upon phase shift? What techniques can be used to predict or derive macro-scale emergent phenomenon from micro-scale complex interactions? Why, exactly and with the same precision as quantum mechanics achieves, is more different?

When I had to come up with a company name back when I was doing freelancing, I chose Entropy Fueled Productions on a rationale quite similar to what was expressed in this article. Then I got to find out very few people know the word 'entropy' much less know how to spell it, and 'fueled' always gets mangled over the phone...

Years ago, Chrome included an experimental vertical side tabs feature. It was great. Worked perfectly. But, it disappeared one day. People created a bug on the tracker to get it restored. It got thousands of stars. It kept getting closed and re-opened and kept accumulating more and more stars. Eventually, someone at Google stepped in and said "we don't like how it looks. We have no other solution. But, we don't like how vertical side tabs look, so we will never permit them in Chrome." and ever since, if you want to open more than half a dozen tabs in Chrome you either need a monitor that's 9 feet wide or you use Firefox with the magnificent TabTree extension which is the correct solution to the problem.

There also might be an issue that recipes themselves can not be copyrighted. Article content and writing alongside the recipe can be copyrighted, but the recipe itself is not eligible for copyright protection at all. So if you're trying to make money off of a site, if it only had recipes, there is no protection whatsoever from someone setting up an identical copy and monetizing it themselves.

You would be atypical if you were less productive by normal standards. Most studies I have seen on the topic usually show a marked increase in worker productivity in working from home. There is a bit of a confounding variable, however... Homes are not open floor plan for the most part. And we know from over 1,000 studies that open floor plan offices massacre productivity to a huge degree. So it might just be the benefit of not having all of the kneecapping and hamstringing taken away that is usually inflicted by business on their workers through open floor plans.

What I didn't see mentioned that concerns me (although I admit it might be a concern born purely out of ignorance): No mention of whether employees would be provided with a minimum of 2 weeks paid sick leave should they show symptoms of COVID-19. Any employee who has a reasonable chance of becoming ill with COVID-19 at their work, and who does not also have a minimum of 2 weeks paid leave, should, by all logic I can think of, refuse to return to work. In a situation with no paid sick leave or less than 2 weeks of paid sick leave, the options are mostly reduced to 2: First, return to work, get sick, then get fired for absence while off sick. Or, second, remain home and get fired for absence while remaining healthy. I suppose there is the third options, the danger-face economic kamikaze model that most Americans will likely be expected to endure - go to work, hope you don't get sick, get sick anyway, get fired for absence. Then you can't even file for unemployment. There is no winning play.

>also offers a glimpse into a potential future where companies of all sorts, not just auto insurers

No, not companies of "all" sorts. Just insurance companies. Other companies are regulated by Congress as part of antitrust regulations which would prevent them from doing this kind of thing. Insurance companies, however, are not regulated by antitrust laws. Insurance is declared to be "not commerce" and therefore not answerable to antitrust laws. They are permitted to break any antitrust law they please. This is why medical insurance companies, for instance, hold national meetings every few years to decide what price they are willing to pay for medical goods and services. This is price-fixing, and it is illegal in absolutely any industry - except insurance. In insurance, they can do it publicly because there literally is no law against it. There's a law PROTECTING it. It was called the McCarran-Ferguson Act. The plan was to repeal it as part of the Affordable Care Act. It was the clause that enabled the ACA to survive the entire revision process... and was then promptly removed from the bill immediately before its passage. This is why the insurers have never really fought very hard against the ACA. If it ever gets repealed, the McCarran-Ferguson act, that is, insurance companies will fall under antitrust regulation and have to change almost every business practice they follow. They will have to compete against one another on price, compete against each other on what prices they pay, etc. It would destroy their profit margins and probably result in the majority of them going out of business quite quickly. Don't hold your breath. It's why the insurance industry spends so much on lobbying and every lobbyist for every single industry will fight its repeal. In other words, it will NEVER happen. The law has been in place since 1945.

That idea of a critical period we're unaware even exists scares me. We know of several critical periods, like an infant with their eyes covered for the first two months of their life will never master binocular vision, a child born deaf can be given normal hearing with a cochlear implant if placed before or around age 2 with effectiveness falling off the later it is implanted, etc. What if there IS a critical period for learning how to develop deep and meaningful personal relationships? Everything about modern child life is designed to pervert and destroy any ability for a child to form such relationships. I have similar concerns about autonomy. Autonomy is a very strong need in all animals, so it would only make sense for this to be true of humans as well and we see plenty of indication of it throughout history and into modern times... but we continue to subject young people to almost total suspension of their autonomy during all of their formative years. What impact should we expect that to have on the kids later in life? The disturbing thing about critical periods is that even knowing you've missed one doesn't help, you just missed it and the most you can do is compensate.

It often surprises me when I hear about parents who are considering job-hopping and moving when they have children that are in the critical 11 - 15 age range. I had read ages ago, back in the late 90s I believe, that the chance of suicide skyrockets if an early adolescent is separated from their friend group. In my later reading and reflecting on my own experiences and whatnot I've come to believe that prior to the advent of agriculture, when most of humanity were nomadic hunter-gatherers (mostly gathering), when the single universally shared concept was of shared fatherhood, and when most children were raised in-common by the tribe, that friend group formed in early adolescence most likely functioned very much like your second family. First you get the tribe of adults, the family you're given, then you get to choose one and build it yourself. We spent the vast majority of human development at that stage, so it's a bit presumptive to think we can suppress it without significant consequence.

This sort of stealthy manipulation reminds me of an idea I had while reading 'Linked: The New Science of Networks.' It's a dangerous idea, I think. The book discusses what would be necessary to take the network of film participants and effectively 'break' the '6 degrees of Kevin Bacon' game. Most would presume the way you'd go about it is by finding the nodes connected to the most people and remove them. That's wrong. Because of clustering, removing the most-connected nodes results in almost no change to the general connectivity of other nodes.

Nodes which are actually important are 'bridge' nodes that provide a means of moving between mostly-disconnected groups. I started wondering what these ideas looked like in an actual social graph, like society. What would 'bridge' nodes look like, and what would eliminating the connection to them look like and what effect would it have? I think a social bridge node would be something like a biker whose main social group is his motorcycle gang, but who also participates in his elderly aunts knitting circle once a month. He provides a means through which ideas and concepts and information can flow from biker gangs, and those connected to them, to a group of elderly ladies and those they are connected to. They are, almost by definition, tenuous links. Ones which, if someone had influence over the communication networks they were using, it might be very easy to disrupt. What consequences would there be to breaking those links on a large scale? In the '6 degrees of Kevin Bacon' situation, you can get the average number of links needed to get to Kevin Bacon up over a dozen by only removing a couple handfuls of bridge nodes.

I think doing such a thing on a real social graph could be very quiet, possibly undetectable (drop messages from rarely-connecting pairs of users... they rarely connect, so how many of them will go through the trouble to re-establish contact? Have bridge nodes have something go haywire and they have to be issued a new phone number, 'their facebook got hacked', etc). And the consequence would be to freeze most things in place, or at least radically slow down any kind of large-scale social change. Disruption of the status quo on the scale of regime change in a government, say, requires buy-in from large and very mostly-disconnected segments of the population. If only pockets of people are interested in change, it doesn't matter how intensely they want the change to happen, it only matters if they can join forces with very disparate compatriots. If you had high-level control of communication networks and a vested interest in guarding the status quo against large-scale social upheaval, you could probably do it very quietly and without really needing anything more than the metadata of connectivity. No need to find out what ideas are being spread, you could just make sure ALL ideas remain trapped in their own little bubbles or that their spread is greatly contained.

What about making it so that the amount of latency added is dynamic? It could be set up so that frequency of visitation leads to greater and greater latency. Something like exponential growth of 1.5x every time the site is visited with an exponential decay function applied since the last time visited. This would encourage slowing down, and would most heavily penalize the most heavily used sites which seems coherent with the goal.

Look into the case where a Christian film company fraudulently claimed a tons of videos on The Bible Reloaded YouTube channel. They went to court. They 'won'. Kinda. They ended up settling, but their case was as iron-clad and perfect as anyone could dream. They didn't include a single frame of the movie they were talking about in their video, didn't include a single nanosecond of audio from the film, but the company still filed a claim. But when they went into court, the company blamed the actual employee who filed the claims and for some reason the court just decided that the company can't be held responsible for the actions their employee took while doing the job they were hired to do. I would expect the same thing to happen in any similar case. They just throw the one person under the bus and move on.