650 karma · joined March 25, 2008
Everything was singular in its overwhelming confusion of intricate detail.
Perhaps a microscope could give you a few more blips of success?
To broaden the applicability, consider that a clever person goes through life with their bullshit checker set to medium. Normies try to persuade him of stuff, and he quickly spots the flaws in their arguments. But a clever person doesn't lock the setting at medium. They adjust it to suit their adversary. If it is a negotiation for a large business deal, the skepticism gets turned up to a high level.
One can imagine a situation in which an executive, Mr E, for company A discovers that company B has hired away one of company A's due diligence team. Whoops! Mr E thinks "They know what we look for, and will have made very sure that it looks good." One adjusts the setting of ones bullshit detector not just on the raw ability of ones adversary, but also to whether they have access to your thought processes, that they can use to outwit you.
Assume for the sake of argument that the previous paragraph is for real. Adjusting your bullshit detectors to allow for an adversary reading your secrets is an option. Then it leads to actionable advice.
How do you set your bullshit detector when you are trying to avoid deceiving yourself? You use the settings that you use when you fear that an adversary has got inside your head and knows how to craft an argument to exploit your weak spots.
In 1989, a program like Henley was used to simulate netnews postings by well-known flamers. The fake postings fooled a significant number of readers. Like all good hoaxes, this one had an underlying point. What did it say about the content of the original flames, or the attention with which they were read that randomly generated postings could be mistaken for the real thing?
One of the most valuable contributions of artificial intelligence research has been to teach us which tasks are really difficult. Some tasks turn out to be trivial, and some almost impossible. If artificial intelligence is concerned with the latter, then study of the former might be called artificial stupidity. A silly name, perhaps, but this field has real promise - it promises to yield programs that play a role like that of control experiments.
Speaking with the appearance of meaning is one of the tasks that turn out to be surprisingly easy. People's predisposition to find meaning is so strong that they tend to overshoot the mark. So if a speaker takes care to give his sentences a certain kind of superficial coherence, and his audience are sufficiently credulous, they will make sense of what he says.
This fact is probably as old as human history. But now we can give examples of genuinely random text for comparison. And if our randomly generated productions are difficult to distinguish from the real thing, might that not set people thinking?
The program shown in Chapter 8 is about as simple as such a program could be, and that is already enough to generate "poetry" that many people (try it on your friends) will believe was written by a human being. With programs that work on the same principle as this one, but which model text as more than a simple stream of words, it will be possible to generate random text that has even more of the trappings of meaning.
Graham published ANSI Common Lisp in 1996, so the kind of program he was talking about was far inferior to GPT-3. There is an extra level of self reference when we read his footnote in 2023 and think that he is talking about today's programs rather than the incomparably weaker programs of 1989. We find the meaning that we want, even in texts that complain about us finding the the meaning that we want.
There is a kind of bone spur that you get in your ear if you swim a lot in cold water. Archeologists spot that in skeletal remains.
Middlemarch started serial publication in 1871. So George Elliot gets to allude to telescopic philanthropy, confident that her audience will be familiar with "the wag's definition of a philanthropist" and get her dig.
So I know why I'm procrastinating, I don't want to tackle a 4 Ugh chore right now. But I can plan. Tomorrow morning I'll be up for tackling a 10 Ugh chore, I can plan to do it then.
Sometimes we don't know the upper bound. The article mentions religious doubts, which can escalate to nihilism and despair. One cannot plan for that. No-one says: I'm too tired for nihilism and despair this evening, I'll do it tomorrow morning.
Abyss? The examples in the article have a feel of a scarily large upper bound, which we are putting off estimating. But we could estimate it, realize that we are not at the abyss, we are just facing a tough choice, and get on with it.
http://www.cofault.com/2022/08/3-lisp-infinite-tower-of-meta...
Discussed three months ago
When Alpha Go adds one of its own self-vs-self games to its training database, it is adding a genuine game. The rules are followed. One side wins. The winning side did something right.
Perhaps the standard of play is low. One side makes some bad moves, the other side makes a fatal blunder, the first side pounces and wins. I was surprised that they got training through self play to work; in the earlier stages the player who wins is only playing a little better than the player who loses and it is hard to work out what to learn. But the truth of Go is present in the games and not diluted beyond recovery.
But a LLM is playing a post-modern game of intertextuality. It doesn't know that there is a world beyond language to which language sometimes refers. Is what a LLM writes true or false? It is unaware of either possibility. If its own output is added to the training data, that creates a fascinating dynamic. But where does it go? Without Alpha Go's crutch of the "truth" of which player won the game according to the hard coded rules, I think the dynamics have no anchorage in reality and would drift, first into surrealism and then psychosis.
One sees that AlphaGo is copying the moves that it was trained on and a LLM is also copying the moves that is was trained on and that these two things are not the same.
Searching to find out whether "original" is the current version, I've noticed that https://en.wikipedia.org/wiki/Stratego#Versions answers your question
"European versions of the game give the Marshal the highest number (10), while the initial American versions give the Marshal the lowest number (1) to show the highest value (i.e. it is the #1 or most powerful tile). More recent American versions of the game, which adopted the European system, caused considerable complaint among American players who grew up in the 1960s and 1970s."
The naive analysis just looks at the military effectiveness of the options. But there may be political implications. Perhaps if country B spends the money on tanks, that gets noticed in country A. That changes the politics in country A and lets the Fighters there persuade their own Peace-niks that more "defense" spending is needed. Whoops! The money that country B spends on tanks doesn't help defend country B as much as you would expect, if in unleashes counter-spending in country A.
What about country B spending the money on static defenses? That could help Peace-niks in country A push for cuts in country A's "defense" spending. That would multiply the effectiveness of country B's spending on static defenses.
I don't think it worked like that with the Maginot Line. Nevertheless, what a General says about military effectiveness misses part of the story.
The enumeration inherently includes functions of several variables, so I wasn't restricted to examples such as 1->1, 2->4, 3->9, 4->16 etc.
I could try it out on examples such as (1,2)->3 (2,1)->3 (0,2)->2, etc. Perhaps with enough it would "learn to add" = find a primitive recursive function that did addition.
I got as far as finding the first problem. The enumeration technique that I used was effectively doing a tree recursion, like that function for computing Fibonacci numbers that bogs down because Fib(10) is computing Fib(5) lots of times. I had a lot of numbers that coded for the identity function, lots of numbers that coded for the first few functions, making the whole thing bog down, trying the same few functions over and over under different numerical disguises.
I thought that I could see my way to fixing this first problem. Have some way of recognizing numbers that give forms that give the same function. I guessed that I could approximate this by saying that if two functions give the same value on a variety of arguments they are probably the same. Then I parameterise this criterion and tune. That opens the way to creating a consolidated enumeration, analogous to fixing the tree recursive fibonacci function by memoization, except trickier.
But my health is poor and I ran out of energy.
Also, I have a guess for the second problem. What happens if I fix the first problem and my enumeration reaches decently complicated primitive recursive functions. While they will all terminate, some might run for far too long, causing the process to bog down. Rejecting them on the basis of limiting the run time might work well. We are happy to only learn reasonably effect functions for doing maths.
It is a fun idea and I encourage others to have a go.
I've seen this approach implemented for teaching logic and proof, both
https://www.ma.imperial.ac.uk/~buzzard/xena/natural_number_g...
and
and it seems to work well. It is certainly easy to get started.
So you are fairly safe. You can only fall victim to heroic terrorist attacks.
But not as safe as that sounds. It is about being a hero according to the attackers head cannon, not your head cannon.
Terrorist attacks are down stream from culture. Keep an eye on what movies, novels, religions, and preachers call heroic. The terrorist attacks of 2052 will be distorted versions of what is heroic today.
A normal person probably doesn't see the double-indirection and ambiguity in this. The first indirection lies in using 'I'm a vi user' as a proxy for 'I don't like Emacs'. The second indirection lies in using 'I don't like Emacs' as a proxy for 'I reject you romantically'.
The ambiguity is even less visible. Some women use an indirect 'No' as a shit test. The man is supposed to ask again. This is phenomenon invisible to normal people because it occurs in a social context of body language and flirting. Normal people find this social context conspicuous. Consequently, normal people are oblivious to the phenomenon of men with poor social skills failing to navigate the treacherous waters around the "ask again no" and the very different "don't ask again no". These waters are a normie's paddling pool.
Remind yourself of the big picture here. We have a highly intelligent man with embarrassing blind-spots around certain social skills. We have intelligent women attempting to communicate with him. They choose a code with two properties. First, it aims at a social skills blind-spot; the message is likely to be misunderstood by the target. Second, it is a trivial code that normies will not even notice breaking. They get to have a sneer and a laugh at the expense of the loser who misunderstands the coded message. The joke lies in the weakness of the code. Everybody except the victim gets the message effortlessly.
So long as 'I'm a vi user' is a private joke, its heartless cruelty is mitigated by its pettiness. But RMS is the victim. To tell the tale as a strike against him lacks self-awareness.
https://github.com/kennytilton/cells
and he has slides from a talk
https://github.com/kennytilton/cells/blob/main/Lisp-NYC-2018...
to give context.
Medicine is much trickier. If an eminent doctor declares a medical treatment to be successful and others doubt, can the controversy be settled by a randomized controlled trial? That requires that the control group be deprived of a treatment that has the approval of an eminent doctor. Tradition views that as unethical and forbidden.
The next big date is 1972 and the publication of Effectiveness and Efficiency: Random Reflections on Health Services. Archie Cochrane moved things along with straight talk about how few medical treatments actually worked. Trusting eminent doctors had been tried and failed. It was time for randomized controlled trials to be used, not just for proposed new treatments, but to review old treatments that some doubted.
But returning to 1907, one notes that Archie Cochrane was born in 1909. The man who persuaded the world that biostatistics was for medicine, not just plants, wasn't born yet.
Dismantling the state "asylum" system in the 1970's looks poorly timed in retrospect. The mentally ill end up missing out on the transition from eminence based medicine to evidence based medicine. Too late to go back and fix it; we've moved on to money based medicine :-(
Temperature too high again? Have a long pause to make a little money from geothermal energy.
https://en.wikipedia.org/wiki/Lake_Peigneur
So I'm picturing water leaking into a salt mine full of pressurized air. Naively it makes a hole, and a hole is just a leak. But if the water starts dissolving the salt in a process that runs away, then it is more like popping a balloon, but on a grand scale.
Consider the first one hundred spins breaking 60 red, 40 black. That is a little odd, but not outrageously so. The ratio is 60/40 = 1.5
Suppose that the next one hundred spins break 50/50. There are no surplus blacks to push the ratio towards 50/50. So what is your intuition? It seems clear enough that the totals are 110 red, 90 black, for a ratio 110/90 = 1.22...
Having imagined the next hundred spins vividly, we are comfortable saying "Unless the surplus reds keep coming, the ratio is going to drift back towards 50/50 all by itself.". There is a striking clash between different "things we might say" seeming sensible, depending on the vividness and detail with which we imagine the situation.
But P(B) just sat there. Puzzling. Then I realized that it starts at P(B|A) = P(B| not A) which is a special case.
An excellent philosophical toy :-)
But there must be different facets to antibiotic resistance genes, because a bacterium that is producing an antibiotic needs to avoid being poisoned by it. Perhaps some genes produce antibiotic resistance by coding for proteins that metabolize the antibiotic. Such genes are not helpful when paired with antibiotic production because they are destroying the antibiotic as fast as the other genes produce it.
By contrast, some antibiotics work by destroying the cell walls of bacteria. Presumably some antibiotic resistance genes work by coding for resistant cell walls. And a bacterium that is producing such an antibiotic had better have such resistance genes or it will end up destroying its own cell walls.
The article is silent on the differences between antibiotic resistance genes that a producer needs, to survive its own antibiotics, and antibiotic resistance genes that a producer mustn't have, to avoid sabotaging its own production.
The article raises the issue of mutualism between different species of bacteria, with shared resistance and production genes letting the diverse community resist outsiders. Perhaps that is already a complicated story without adding in different kinds of resistance genes.
x = a * sqrt(t) + r * cos(t^2) y = b * sqrt(t) + r * sin(t^2)
If you are monitoring long term progress towards (100a, 100b) you are disappointed to see that progress is slowing. And yet the short term whirl is whirling ever faster.
I suspect that a weak version of Sapir-Whorf is in play. It is difficult to describe the equation in prose. The limitations of natural language handicap us in our efforts to defend against the pathology it exemplifies.
We could end up in the situation that computers are put in charge of important things because we have mastered the art of writing computer programs that can trick us into thinking that they are intelligent. Then it turns out that the programs are stupid. Disaster ensues.
The key question is "how clever does a computer program have to be to convince humans that it is intelligent?". That depends on two sub-issues.
First, are our assessments hackable? We might in time stumble over a clever trick that makes computers seem impressive to humans. The glib talkativeness of GPT-3, which impresses every-body, hints at this
Second, how keen are we to fool ourselves and just believe, without all that much prompting from the computer? The article worries me because of all the love given to the narrative in which the computer really is intelligent, impressively so, dangerously so. Erring on the other side, thinking a stupid computer is intelligent, is also a big risk. We seem blind to it.
This seems to be about halving the strength requirement because you can use both legs on every step, while also eliminating the balance and coordination requirements of climbing and descending stairs. Maybe that is all that is needed if some-one has had a stroke and falls just short of being able to use stairs. But I don't see it buying much time with a progressive condition.
Consider ME/CFS. The human body has remarkable powers of recuperation. Feel ill with a cold? Just take two weeks off to rest and you will recover. Broken bone? That will need splinting, but again it will heal up in six weeks. Ah! Now we come to the difficulty. There are standard times, two weeks, six weeks; what happens if something goes wrong and healing takes a lot longer?
One psycho-social explanation of ME/CFS is that patients aren't patient. Something nasty has happened, bio-medically, and they need to take two year out of their hectic life to recover. After six months they cannot take resting and pacing any more and do too much, too soon, provoking a major relapse. That restarts the clock on two years to recover. Rinse and repeat indefinitely. Call that "trying too hard".
The rival psycho-social explanation is that patients are deconditioned due to too much rest and need to exercise and build up their stamina on a "normal" schedule. Call this "not trying hard enough". This doesn't really work as an explanation because some of the patients are ex-athletes and familiar with training and building back up systematically after time off for an injury or illness. They are going to the doctor because they have done what always worked in past, because it hasn't worked this time. The rival psycho-social explanation isn't relevant to the actual problem.
It is worse than that. NASA attempts to study the effects of weightless on the ground, in gravity, by having people lie in bed and get deconditioned. Deconditioning is a rather delicate effect. It depends on a rigid discipline of lying down all the time. The effects wear off rather quickly. Crucially the effect wears off monotonically. It is not like ME/CFS with its key symptom of Post-Exertional Malaise: sufferers think they are getting better, but a day or two later, wallop, they have a flare up of symptoms.
The natural expectation is for a fight between psychologists. "Trying to hard" versus "not trying hard enough". One expects it to be a one-sided fight. "not trying hard enough" is a much worse fit to what the patients experience.
Back in the real world "not trying hard enough" has won a total victory. When patients are angry about the "not trying hard enough" explanation, that is understood as anger about the psycho-social model, as if there were only one. The important question is "Why prefer the implausible psycho-social model over the plausible one?". And that question is difficult to ask.
The article manages to skip over the central issue with popular "all in your head" explanations. It is not that they compare unfavourably with "physical illness" explanations. The problem is that they compare unfavourably with less popular "all in your head" explanations. What is going on there?
See https://reframingthevictorians.blogspot.com/2013/11/shops-an...
The article doesn't say how many generals Indonesia had before the attempted coup. Obviously seven fewer after the attempted coup. I wonder what the surviving generals thought of the recent murders of their colleagues?
The article is written as though the surviving generals sat around afterwards waiting for Britain to incite them to mount a response. That seems unlikely. It would be more usual for the surviving generals to take the murders of their colleagues personally and to attempt a response on a kill-or-be-killed basis, without waiting for British incitement to rouse them to action.