jorleif

My own experience from academia tells me that a big part of the problem is the fact that everyone is competing for a small pool of faculty positions and those positions are filled based on individual performance. This means you can never win as a team, only as an individual, and everyone needs to get their share. Only a professor can build a team, but even those teams are temporary. Add to this the fact that only new things count as science, which means that good work can easily turn out worthless if the desired results are not achieved. For this reason, science then is naturally elitist, and has big differences in the status of participants (tenured vs not), and heavy competition. Might it be that this environment breeds nasty behavior such as misogyny and bullying?

If a professor treats a student badly (I don't mean harassment, I mean overwork and bad guidance) then that is rarely punished in any way, it is probably beneficial for his career.

It's not either "just luck" or no luck involved. Sure, most people who do get rich try and work very hard. The question is how big a proportion of people who work hard do get rich. If it is only 1% and it cannot be predicted in advance who it is, then there certainly is an element of luck.

It is certainly true that workforce reduction is just one possibility, but what I still wonder is when coops work and when they don't. Somehow it seems to me that they only work when the business is not too scalable, but profit is directly related to work input.

While I think co-ops can work very well in some cases, I wonder how well they cope with significant change for the worse in their environment. If you need to lose a third of the workforce suddenly, can a co-op decide on that kind of thing or will it lead to some unproductive compromise, or even worse, denial?

We are all semi-technical outside of the narrow situations and technologies we have a lot of experience with. In other words, you can learn to be technical, but only in narrow subfields like a particular programming language, programming environment, application and so on. University education is good for getting at the shared core of these things (computer science and especially mathematics).

But what should you do? Well, I suppose you need to decide if you want to become more technical in some subfield, and what field that would possibly be, or if you want to learn the metaskills of managing projects where you are semi-technical in the sense that you know a lot about the project you are managing, but could not put it together yourself.

Note that these are complementary skills, so you can have both (and often should strive to have both).

As a technical person I would tend to think that the easy path towards a less-technical managerial position would go through an entry level technical position, but that is certainly not the only way.

I think the process of peer review often causes papers to be optimized for being hard to critique rather than easy to understand. I have myself participated in writing papers where we decided to leave out some non-crucial but very useful detail just because it opens up too many questions and opportunities for critique.

I wonder if there would be any way to implement some alternative identification system that would enable the benefits the article mentions, but still allow for privacy that does not require a trustworthy government. I think, if the government can be trusted (always questionable, but sometimes true), a national ID is clearly an efficient solution. But if government can't be trusted, then maybe some kind of decentralized pseudonymity-based system could work. The state could tax the pseudonyms, they might hold the access rights to medical records etc. Of course this would be cumbersome and allow a certain level of fraud, but might still be an interesting alternative in hostile environments (the state either opposes what one does, or at least does not support it). Maybe one should look into how guerrillas and organized crime identify people that belong to their networks.

That is very well put. I am often amazed what a big difference certain ideas make it possible to build complex things quickly and with low amounts of bugs. For example building a parsing subsystem "properly" as taught on compiler courses vs haphazard combination of regexps. It is mentally much more demanding, and requires more effort up front, but then to your surprise, every change you make just works almost immediately, and there are no hairy corner cases to resolve (this is kind of declarative vs imperative, but not quite). Another example is mathematical models, such as Markov Models and the like, where you can have a really complex behavior, but your parameters are simply matrices and all operations are some sort of matrix algebra. Difficult to understand at first, but very powerful and "clean" once you do.

In contrast, as you say, there are huge amounts of armchair ideas that are not worth very much even if they are correct. For example "software will eat everything" is something I believe is mostly right. It is also a business idea with direct implications for action. Despite this, it is only valuable to set the general direction (let's do software), but not any specific cases. It does not mean that every case of transforming a traditional business into a software one is going to be a grand success, or even profitable. In fact, these types of ideas have a tendency to cause one to forget the exceptions from the rule, and thereby being actively harmful.

Why is the mental model building not seen as a special case of induction? One needs to generalize (i.e. interpolate/extrapolate) from the known instances. The mechanism may be just "replaying" known sequences, but there is still a measure of induction because one needs to decide that which simulation applies to the ball I'm seeing right now (and that what I am seeing is a ball, and that intuitive physics models of balls apply).

Had he said that Bayesian decision theory with optimization of utility functions etc. was behaviorism, then I would agree, but Bayesian formulations can be used for many other things as well. At least one can propose very complex internal models, which behaviorists were criticized for not considering.

I think there is a difference, but my view is that it is impossible to reason without sufficiently accurate prediction, and that the concepts we use for reasoning are largely formed based on how predictive they are. Consider playing a game where two players throw dice, and the one with the larger number wins. There is no predictability between turns, so there is no point in reasoning far into the future, unlike in chess, where the positions change slowly, one piece per turn and in predictable patterns. In chess you still don't know what the other player will do, but you can reason about it, because there is a level of predictability.

I would define reasoning as answering a question or optimization problem, e.g. What is my best move? What is a good move? This can be formalized as calculating some kind of score for a possible action, whereas prediction is just about assigning a probability value to an event.

What do you mean is unfounded? My suggestion for what creativity might be as be? I guess I was unclear, I don't mean to say I know what creativity is. What I do mean to ask, is whether creativity is something more than just solving optimization problems? To me that way of thinking about it explains many of my own experiences (a-ha moments, unknown time to completion of thinking process, thinking going on in the background etc.) using a formulation that I understand. It is free for anyone to prove it wrong.

"only a tiny component of thinking is about prediction at all"

Here I really struggle to follow his argument. It seems, he is saying that thinking is not prediction, but how does he know that? As I see it, we don't know what thinking is, so is he then saying that thinking does not feel like prediction, and therefore it cannot be prediction? It is well known that we are only aware of our very highest level of thought. What goes on below the surface may very well be very much about prediction without us knowing about. When we think about "the world" we must have some internal representation of the configuration of objects that we think about. In my opinion this is implicitly a kind of prediction, because we create (at least mostly) configurations that are plausible, if not true (counterfactual, but possible).

It seems to me that creativity is mostly about finding good approximate solutions to hard (e.g. NP-complete) problems. How we do it is so far unknown, but it seems that it is not in principle so very special. The explanation for the calendar he speaks about is actually not very complicated inference at all, and is probably in reach for artificial systems. All he has to do is to generalize 19 as belonging to a number category, and then apply the predictive properties about how numbers work. Not all prediction is naively predicting that the world is as it was before. If change is predictable in the past, then change can also be predicted to the future. Prediction is difficult, but that does not make it less important.

It seems you are assuming there are direct observations. Musk is talking about generating hypothetical observations from a model (the cost of the battery is bounded from below by the cost of the battery materials). This sort of bounding does not always work outside physics, because the uncertainties are so ill behaving.

I agree, but would like to add the aspect of uncertainty. When reasoning from analogy, you usually have good statistical knowledge about the properties of the problem. For example, the targetted product category may exist, and customer behavior is known, and therefore predicting what would happen in some nearby configuration is usually somewhat accurate. On the other hand, in the first principles case you need to have a very accurate theory, because you are "far away" from what exists currently. If the theory in questions concerns physics, as for Musk, then this can work well. On the other hand if it is about social sciences and you need to predict the behaviors of customer from some kind of first-principle model (e.g. rational agent models), then you are very likely to make mispredictions, since human behavior is complicated, and the theories are inaccurate and overly general.

Exactly what I thought. There is an awful lot of damage a semi-truck can do if things go wrong. But, then I realized, that there is not necessarily a need for the "trucks" to be big. Maybe you could go the Roomba route, and make them much smaller than traditional trucks, maybe even smaller than conventional cars. Since no driver is needed, you can scale out with many of them, rather than having one very large truck. Maybe that approach does not work for everything, but somehow I think having a large bunch of cargo ants is much less risky than a single very large cargo elephant.

That is a very modern take on it. I think the cultures that matter are the ones that will protect you if you protect it. Traditionally they are very much based on ancestry or political power. Still today, when people run into trouble, the ones who pick them up are their oldest friends and relatives, not people from their professional or hobby subculture, even if they are big shots within those circles.

Could the teachers statement be rephrased in terms that the kid is developing analytical skills so quickly that they are dominating over the more "experiential" part of socializing with others? It seems that this kind of thing persists into adulthood, because I notice that it is hard to experience and analyze at the same time. For example, it is hard to read a book to experience the story and analyze the narrative structure at the same time. A similar point is described in "Gödel, Escher, Bach" that one can either listen to a musical piece as a whole (experientally) or as individual pieces, such as notes (analytically), but not both at the same time.

While I tend to think analytical skills are more important, it is probably not good to entirely neglect the skills to experience and "live in the moment" either, since that seems to be the foundation of many social skills.

If you want to buy, say furniture, a big barrier of web shopping is that you cannot automatically reason about their dimensions, because they are not exact enough (and even more seriously, they are usually available only as free text).

The 3D scanning technology would enable products that are add-ons to objects where the original manufacturer did not see value in adding an interface. If I can scan my car interior, I can attach my GPS to nearly any surface "perfectly". This gives aftermarket products very nice finish with a much lower cost.

The article has an interesting argument structure. First it undermines the science behind psychiatric drugs by referring to their discovery as serendipitous. Then it proceeds to point out that there is no good theoretical understanding for how the drugs work, even if there is understanding in what neurotransmitter levels are effective. Then the author proceeds to drop the "not better than placebo" results.

Without knowing any psychiatric science, this narrative nevertheless seems suspicious. First of all, it is entirely irrelevant whether a discovery was made serendipitously or as a result of incremental science, as long as the effect is clear enough to be repeatedly found. This is not true in well understood fields, such as physics, where the theories are mostly not changed every now and then, but in complex fields, such as medicine and social sciences, the fact that something matches pre-existing theory is no evidence per se. The problem is "bumpy" enough that any theory is likely to flatten every bump that is not directly observed, even if that bump would be very useful in practice. This leads to most discoveries of any value being serendipitous. It also leads to incremental research failing to advance rapidly most of the time.

So the lack of theory and serendipitous discovery are not problems in themselves. The finding that psychiatric drugs would not be better than placebos is indeed a problem. It does raise the question, of how the serendipitous discovery was ever made. Did it arise out of spurious correlations, so that when the drug was first discovered there happened to be a big incidental recovery in most treated patients? Then in later retrials people managed to trigger a sufficient placebo effect to get the whole thing going. It does seem a little unlikely, but one would have to examine how the first trials were performed.

> we've watered our teaching of history (and most things) down to soundbites

Exactly, or at least to neat narratives that are very much retrospectively constructed and do not really give a realistic picture of the messiness of how the events felt to those who were living through them. I think reading actual speeches, or documents written at the time gives a very complementary view to the neat narratives and insight into the value systems that people had at the time. Of course this is precisely the reason why reading them demands very much of the reader.

In general, I think, one should read more original works, and less later histories describing said works. It is more demanding, but also more satisfying.

Exactly, I think it's a good heuristic that people who ask for loyalty don't deserve that loyalty. People who do deserve it rarely have to ask, because they are busy turning the situation into a win-win, or at least symmetric in risk-profit, so that you both take a risk for profit, and if the risk is realized you are in the same boat, rather than you in the sea, and them sitting comfortably somewhere else.

For a big corporation it is not important to succeed in most acquisition. The real risk, is that some startup eats your current business model and you didn't buy it when you could still afford it. If most acquisitions "fail" that is ok, you still sit on a ton of money.

In other words, probability is more or less irrelevant, because if you fail it means a small loss which does not risk the company's future, on the other hand, if you don't acquire, and the small business succeeds big time, then that is a big loss, and may eventually kill you.

Since we there is no exact definition of what a mentally healthy human being is, in the same way that those specifications would be available for a machine for example, the definitions of mental problems have obviously been developed mostly in other ways.

Conditions where a person causea problems for himself or his surroundings are then labeled as mental health problems. Naturally, the opposite conditions may be benign. The person may be overly obedient, or overly trusting (opposite of paranoia), irrationally peaceful (opposite of depression) and so on, and this may indeed be caused by the same mechanism that causes the opposite condition, but as it does not cause problems, it is not considered a disease.

Yes, the solution would be more along the lines that the owner or one of his direct descendants has to live at the site. Then hurting the environment hurts yourself.

The much bigger problem is that nobody tries to repeat findings, because there is no incentive to do so. There is an asymmetry between positive and negative findings. Say I make a study that shows that X causes Y. Then someone else tries to replicate that. They find no statistically significant evidence that X causes Y. Now, in medicine this may be a result in itself, if X causes Y is a well established result. But, in many other fields, such as computer science, a negative finding is usually just because the researcher made a mistake (bug in code etc.), so you have to work really hard to prove that there is no relation between X and Y, and it is hard to publish these kinds of results unless you can somehow rigorously prove the opposite result.

Usually these kinds of data are not part of the publication or up for review. I don't know why exactly, but I suppose it is a combination of: 1) paper publishing tradition, where distributing this material is too much work. 2) too much work for reviewers to comb through raw data, the real solution would be to require sharing of the raw data, so readers rather than the reviewers would find mistakes, but this is often difficult because of privacy issues. E.g. videos of the Utrecht train station study would make fraud more difficult, but publishing the videos would probably require permission from all participants 3) Trust in the authors' integrity. Most scientists would not make up data, but they might make logical errors in their reasoning, or use bad procedures. Peer review can find these errors, but if the authors lie about their procedures, it is very hard to check. Peer review is about internal consistency.

It's not entirely clear to me how givedirectly.org operates, but it seems they give in an unpredictable fashion, rather than continuously. That might make a big difference, since fixed amounts of money every week or so can lead to the receiver becoming dependent on the aid. But if receiving is not ongoing or predictable, this cannot really happen, while it might of course lead to stupid buying decisions when the money is received, that's not nearly as damaging, and might even stimulate the local economy.

It seems that no one has mentioned that the effect of relative wealth might be severe also from a rationality point of view. In a more primitive environment, if your neighbor becomes too powerful, he might enslave or even kill you, so better keep him in his place.

Similarly, if you take the $100 and are silent, and protect the killer, you risk severe punishment if you are caught. Even if I agree that the decision is made based on moral conscience for most people rather than economic calculation (which might be the case for sociopaths), there is a rational basis for this. Fundamentally, to make sure that I don't get murdered myself by the murderer later, I better not comply with him, as he is proven to be ruthless. In addition, because other people are moral beings as well, $100 is a very small compensation for risking jail and being shunned for the rest of your life, a very bad economic decision.

You can surely call that playing the cards, the difference with actual cards, is that you can't reliably calculate probabilities, and avoiding too large harm is a good strategy in that case.