Posts
Comments
The transmission utility is not purely a transmission company. It spends money on both generation and transmission. Some generation charges leave to other companies. This is not a competitive market, but even if it were, it would only give you a bound on the cost of generation and tell you nothing about the cost of transmission.
You say solar is getting cheaper, but it is only the panels that are getting cheaper. They will continue to get even cheaper, but this is not relevant to retrofitting individual houses, where the cost is already dominated by labor. As the cost of labor dominates, economies of scale in labor will be more relevant.
To a first approximation, solar is legal for individual residences and illegal on a larger scale.
Maybe you could learn something by looking at the public filings, but you didn't look at them. By regulation, not by being public, it has to spend proportionate to its income, but whether it is spending on transmission or generation is a fiction dictated by the regulator. It may well be that its transmission operating costs are much lower than its price and that a change of prices would be viable without any improvement in efficiency. This is exactly what I would how I would expect the company to set prices if it controlled the regulator: to extract as much money as possible on transmission to minimize competition. I don't know how corrupt the regulator is, but that ignorance is exactly my point.
Even in this last comment you keep making that very distinction. The regulator dictates the price but you assert that you know what the monopoly spends.
If you just want to assert that the current set of regulations are unsustainable, then I agree. But not a single one of the comments reflects a belief that this is the topic, not even any of your comments.
Yes, if we assume that there is a competitive market for generation, price of transmission may prevent grid solar generation from being built. But you asserted that you could learn the cost of transmission from the bill.
These numbers are dictated by the regulator. What mechanism is there to make them have any relation to the real world?
That breakdown is fiction dictated by the regulator.
Why do we even believe the claims about congressional trades? It is widely believed that Hillary Clinton's commodity trading was falsified by the broker. Why not the same for stocks? These records were created once a year. It would be easy to look back at the year to choose good trades after the fact. Today we supposedly have 1 day notice of Pelosi's trades, which would be hard to fake.
If the Ziobrowski data is dominated by a few big trades, why not look at them? Are they companies that were affected by congressional action? That is the worst scenario. If not, then I see three possibilities (1) noise; (2) falsified data; or (3) as Christian says, it was a bribe of information from inside the company, not information from inside the government.
What is Chaos Theory? It sounds to me like an arbitrary grouping of results of people playing around with computers, not a coherent theory. If it were about a social group, that provides more coherence. Indeed, the people who pushed the term "Chaos" do form a social group, but I do not think this group really includes all the people included in, say, Gleick's book.
A lot of the results were things that they could have predicted from theory before computers, but they don't seem to have been predicted. In particular Lyapunov died in 1918. If the theory is his theory, then it's hard to articulate what the people with computers contributed, but it may still have been important to actually use the computers. Similarly, I think it wrong to dismiss something as just information theory, not chaos theory. The only concrete result I know is the reconstruction from symbolic dynamics, but this makes it clear how to apply information theory.
Your history is definitely wrong. Patents don't enforce themselves. Hollywood is on the west coast to make physical distance from Edison's lawyers and muscle. The Wright brothers went down in history as the inventors of the airplane, but they wasted the rest of their lives fighting over the patents.
Linchpin patents are rare. Maybe you patent one invention to make it just barely work, but that's not the end of the story. Someone else patents something else needed to make it scalable. Now there are two patents and a bilateral monopoly.
None of this is to say that patents were unimportant, so it's not an answer at all.
What you say about OpenAI makes it the apotheosis of EA and thus I think it would be better for 80k to endorse it to make that clear, rather than to perform the kayfabe of fake opposition.
Instead of imagining if all trials were bench trials, instead perform the experiment. Or just look at the countries where this is true!
What does it mean to claim that these people are contrarians?
Is there a consensus position at all? For any existing policy, you could claim that there is some kind of centrist compromise that it's a good policy, so people who propose changing policy, like Hanson and Caplan, are defying that compromise. But there is not really any explicit consensus goal of most policies, so claiming existing institutions are a bad compromise because they pursue multiple goals and separating those goals is not in defiance of any consensus. Caplan, Hanson, and Sailer are offensive because they feel we should try to understand the world and try steer it. They may be wrong, but the people opposed to them rarely offer an opposing position, but are rather opposed to any position. It seems to me that the difference between true and false is much smaller than the gap between argument and pseudoscience. Maybe Sailer is wrong, but the consensus position that he is peddling pseudoscience is much more wrong and much more dangerous.
Sailer rarely argues for genetic causes, but leaves that to the psychologists. He believes it and sometimes he uses the hypothesis, but usually he uses the hypotheses 1-4 that Turkheimer, Harden, and Nisbett concede. Spelling out the consequences of those claims is enough to unperson him. Maybe he's wrong about these, but he's certainly not claiming to be a contrarian. And people who act like these are false rarely acknowledge an academic consensus. Or compare Jay: it's very hard to distinguish genetic effects from systemic effects, so when Jay argues that racial IQ gaps aren't genetic, he is (explicitly!) arguing that they are caused by racial differences in parenting. Sailer often claims this (he thinks it's half the effect), but people hate this just as much as anything else he says. Calling him a contrarian and focusing attention on one claim seem like an attempt to mislead.
That is a very clear example, but I think something similar is going on in the rest. Guzey seems to have gone overboard in reaction to Matthew Walker's book Why We Sleep. Did that book represent a consensus? I don't know, but it was concrete enough to be wrong, which seems to me much better than an illusion of a consensus.
That they have a "real names" policy is a blatant lie.
They withhold "real names" every day, even ones so "obvious" as to be in wikipedia. If they hate the subject, such as Virgil Texas, they assert that it is a pen name. Their treatment of Scott is off the charts hostile.
If we cannot achieve common knowledge of this, what is the point of any other detail?
How about cancer deaths? From the point of view of 2012, was Beau Biden's death in 2015 after diagnosis in 2013 due to quantum randomness? That sure had a big effect on the Democratic primary, if not the general election.
Sure, but this is not new. You start by saying "AI in 2024" but this is true of everything that has been called AI and a lot of things that maybe should have been called AI, such as the PageRank algorithm. Credit scores have made decisions based on based on statistical models since the 50s.
That sounds pretty similar to sublingual therapy. I think it is likely that sublingual therapy is better because of the denser dosing (weekly vs monthly), but the difference is small enough that it can only be assessed with a head-to-head trial. (If the difference is compliance, it would be difficult to measure, though potentially very large.)
The headline that environmental allergies are curable is a decades old. If this news has not spread, it is good that you promote it, but we should ponder why it is not common knowledge.
The medical consensus is that sublingual immunotherapy is inferior to the injected immunotherapy that has been used for a century. Did you try that as a kid? If there's reason to believe sublingual is better, that's good to know, but it sounds like you just don't know about injections.
Sublingual immunotherapy has an obvious advantage because people don't like shots. And it doesn't require a prescription. Indeed, one should be suspicious of a conflict of interest in the medical consensus. But injected doses are more precisely controlled, so there is good reason to believe they work better. And the doses are smaller, so the material cost is smaller.
Compliance to the schedule may be the main obstacle. It is not obvious whether doctor appointments make this better or worse. This probably varies between people.
What the legal system should be is irrelevant.
In the post you talked about editing all 237 loci to make diabetes negligible, but now you talk about the normal human range. I think that is more correct. Editing all 237 loci would leave the normal human range; the effect on diabetes would be unpredictable and the probability of bad effects likely. Not because of pleiotropy, but just the breakdown of a control system outside of its tested regime.
First of all, the population numbers are complete garbage. This is completely circular. You are just reading out the beliefs about history used to fabricate them. The numbers are generated by people caring about the fall of Rome. The fall of Rome didn't cause of decline in China. Westerners caring about the fall of Rome caused the apparent decline in China.
Second, there was a tremendous scientific and technological regress in Rome. Not caused by the fall of Rome, but the rise of Rome. There was a continual regress in the Mediterranean from 150BC to at least 600AD. Just look at a list of scientists: it has a stark gap 150BC-50AD. It is more controversial to say that the renaissance 50AD-150AD is a pale shadow of the Hellenistic period, but it is. In 145BC Rome fomented a civil war in Egypt, destroying Alexandria, the greatest center of learning. In 133BC, the king of Pergamon tried to avoid this fate by donating the second center of learning. It was peaceful, but science did not survive.
Since those are rare causes of deaths, they don't matter and they're hard to measure. Also, this is a small study, so I trust earlier studies more.
There is a mechanistic explanation. Alcohol is a blood thinner. Blood thinners protect from ischemic heart disease, which is such a large portion of mortality a small improvement can make up for worsening of all other causes. Which is exactly what we see in the observation.
It's that simple.
Before asking why, ask what. Why did the technological growth of ancient Rome not snowball into the industrial revolution? I reject the premise. Rome was a period of regress in both physical technology and social organization, although it did spread some technology westward.
More generally, the macro trends of history are largely fabricated to prove the desired conclusion that there is always exponential progress, except in a few collapses that are so sharp that they cannot be denied. Why did this growth not produce the industrial revolution? Because it wasn't progress.
Slaves reproducing themselves is nonmalthusian, but rare. Romans captured slaves in war and enslaved debtors. I think the only time in history chattel slaves reproduced themselves is the New World, which was quite nonmalthusian.
This is a very popular theory, but it seems to predict way too much. The Greeks and Romans did have animal powered wells and mills. They had water mills and water saws. They probably had windmills.
Yeah, FTX seems like a totally ordinary financial crime. You don't need utilitarianism or risk neutrality to steal customer money or take massive risks.
LaSota and Leverage said that they had high standards and were doing difficult things, whereas SBF said that he was doing the obvious things a little faster, a little more devoted to EV.
The hard part is being willing to call papers bad. The task I find difficult is getting people to acknowledge that I called them bad, rather than gaslighting me.
Someone just told me that the solution to conflicting experiments is more experiments. Taken literally this is wrong: more experiments just means more conflict. What we need are fewer experiments. We need to get rid of the bad experiments.
Why expect that future experiments will be better? Maybe if the experimenters read the past experiments, they could learn from them. Well, maybe, but maybe if you read the experiments today, you could figure out which ones are bad today. If you don't read the experiments today and don't bother to judge which ones are better, what incentive is there for future experimenters to make better experiments, rather than accumulating conflict?
France had a military coup in 1958 followed by 6 months of dictatorship. What threshold had France not passed in 1958 to not count as a full democracy? Does the Dictator's Handbook actually say this?
Did you click through from Paul's LW post to his blog? He gives a proof that a reversible computer can implement a PSPACE algorithm with only polynomially many erasures, and thus only polynomially much energy consumption, at the cost of running a little longer, hardly a noticeable difference compared to the exponential time required. But he also provides context which I suspect you need.
Right, the point is that a Reversible PSPACE appears physically realizable, while currently existing computers could not actually run for the exponential time necessary to compute PSPACE problems because they would also require exponentially much (free) energy.
It took 10 years from mass residential refrigeration to lead to use of CFCs. It took another half-century to detect atmospheric CFCs and the damage they were causing.
This makes it sound like it's an important point in the timeline, that substantial use of CFCs can be dated to c1930. This seems fundamentally wrong to me.
Mass introduction of modern residential refrigeration took place from 1914-1922.
What do you mean? Cooling food? I think that is a rounding error. A single wall AC has 10x as much freon as a refrigerator. Thus I think the bulk of the freon came later and there was not so long a delay from deployment to discovery. But it should be possible to look up actual freon production.
I think the growth of air conditioning was contained by the cost of electricity, not freon. It's hard for me to imagine electricity cheap and widespread enough to allow refrigerators without becoming in a few decades cheap enough to cool houses. But maybe I can imagine a 19th century with Einstein refrigerators yet without electricity. I don't think that would have destroyed the ozone layer.
I think talking about Google/DeepMind as a unitary entity is a mistake. I'm gonna guess that Peter agrees, and that's why he specified DeepMind. Google's publications identify at least two internal language models superior to Lambda, so their release of Bard based on Lambda doesn't tell us much. They are certainly behind in commercializing chatbots, but is that a weak claim. How DeepMind compares to OpenAI is difficult. Four people going to OpenAI is damning, though.
I assume you know this, but to be clear, OpenAI has already used pirated books. GPT-3 was trained on "books2" which appears to be all the text on libgen (and pretty much all the books on libgen have been through OCR). It was weighted the same as the common crawl, lower than Gutenberg or Reddit links. This seems to answer your second question: they will likely treat pdfs on the libgen the same as pdfs on the open web. If you're asking about whether they will train the model on the pixels in these pdfs, which might make up for losses in OCR, I have no idea.
How many characters is your 500 line source file? It probably fits in 8k tokens. You can find out here
Since you have to manually activate plugins, they don't take any context until you do so. In particular, multiple plugins don't compete for context and the machine doesn't decide which one to use.
Please read the documentation and the blog post you cited.
Does OpenAI say this, or are you inferring it entirely from the Wolfram blog post? Isn't that an odd place to learn such a thing?
And where does the Wolfram blog post say this? It sounds to me like he's doing something like this outsider, making one call to Wolfram, then using the LLM to evaluate the result and determine if it produced an error and retry.
Using nat.dev, I find that 002, 003, and Turbo all get the same result, wrong on the first and right on the second. This is an example of Turbo being Inferior to Chat. I also tried Cohere, which got both. I also tried Claude. Full v1.2 got both wrong. Instant 1.0, which should be inferior, got the second correct. It also produced a wordy answer to the first which I give half credit because it said that it was difficult but possible for the slow policeman to catch the fast thief. I only tried each twice, with and without "Let us think," which made no difference to the first. I almost didn't bother adding it to the second since they did so well without it. Adding it made 002 and Claude-instant fail, but Claude1.2 succeed. (I also tried llama and alpaca, but they timed out.)
Right, Wikipedia cites a 1972 paper using viruses to deliver DNA, but no vaccine until 1984. Whereas, mRNA in lipids went from delivery in 1989 to a vaccine in 1993-1994. So twenty years on one metric, but ten years on another metric that probably screens off the first one by virtue of coming later.
But that's just playing around. Obstacles artificially created by the FDA are real obstacles. To the extent that the vaccine-hesitant mean anything by "old-fashioned," they mean large scale experience in humans. More people received vector vaccines in the Oxford trials than in all deployment before. If you want to know about Bell's palsy, that's the only way to find out. On the other hand, if you want years of follow-up, a 2015 trial of vector vaccines could have been an big advantage over mRNA vaccines, although I don't know if they actually followed up after years. With no placebo group, it's not clear what analysis they could make.
Lots of people did assert that adenovirus vaccines were old-fashioned. But this is false. The first such vaccine was approved in 2019 (maybe a more appropriate comparison date is is 2015). I am skeptical of trying to manipulate lies, even if it is easy to predict that people will ultimately believe many falsehoods.
Note the tens, perhaps hundreds of thousands of Germans who got the illegal Stöcker vaccine, a recombinant protein vaccine, that is, a 20th century vaccine, an actual old-fashioned vaccine. Were they just opposed to government sanction, or did they care about how old the techniques were and would have gotten the Novavax vaccine, had it been available? Maybe a better strategy would be to encourage actual diversity.
I recently went to a public space that advertised that its escalator handles were disinfected by UVC. Focusing on marketing can produce the worst of both worlds.
Let's assume that OpenAI is reckless. Does giving them money make them more reckless?
It seems to me that your thought process is that OpenAI is evil and thus giving them money symbolically rewards evil. There can be some value in symbolic actions. This reminds me of the sporting and culture boycotts of Apartheid South Africa. To whatever extent that these worked, it wasn't about the money, but other forms of leverage.
Maybe tiny positive feedbacks reinforce behavior, although this seems pretty anthropomorphic. But maybe giving them money for services widens their options from inherently short-term venture funding. A sustainable stream of product money might make them less reckless. Probably it just can't compete with venture funding, but, if anything, I think the sign is positive.
Lung cancer affects old people.
Also, while your link claims that lifetime healthcare costs are greater for smokers, it does not claim it is a consensus, but specifically mentions that many people claim the opposite. And that's before getting to Gerald Monroe's point.
The first thing to do is to distinguish human things from inhuman things. Physical things really are run by rigid laws. Social things like contracts, money, property, and a guilty verdict are caused by humans and this should make it obvious that they don't have rigid behavior. (The feeling of guilt is yet a third category.)
A lot of people seem to think that signatures are magic. Would you agree with that description of your children? It would be interesting if you could figure out where this idea came from, either spontaneous generation or transmission.
I think most democratic countries use proportional representation, not FTPT. But talking about "most" is an FTPT error. Enough countries use proportional representation that you can study the effect of voting systems. And the results are shocking to me. The theoretical predictions are completely wrong. Duverger's law is false in every FTPT country except America. On the flip side, while PR does lead to more parties, they still form 1-dimensional spectrum. For example, a Green Party is usually a far-left party with slightly different preferences, instead of a single issue party that is willing to form coalitions with the right.
If politics were two dimensional, why wouldn't you expect Condorcet cycles? Why would population get rid of them? If you have two candidates, a tie between them is on a razor's edge. The larger the population of voters, the less likely. But if you have three candidates and three roughly equally common preferences, the cyclic shifts of A > B > C, then this is a robust tie. You only get a Condorcet winner when one of the factions becomes as big as the other two combined. Of course I have assumed away the other three preferences, but this is robust to them being small, not merely nonexistent.
I don't know what happens in the following model: there are three issues A,B,C. Everyone, both voter and candidate, is for all of them, but in a zero-sum way, represented a vector a,b,c, with a+b+c = 11, a,b,c>=0. Start with the voters as above, at (10,1,0), (0,10,1), (1,0,10). Then the candidates (11,0,0), (0,11,0), (0,0,11) form a Condorcet cycle. By symmetry there is no Condorcet winner over all possible candidates. Randomly shift the proportion of voters. Is there a candidate that beats the three given candidates? One that beats all possible candidates? I doubt it. Add noise to make the individual voters unique. Now, I don't know.
In what sense are these failures agentic?
There are two issues. One is capability. Why do GPTs have the ability to use other languages? The other is why does the RLFH cause InstructGPT to follow instructions in other languages? The tweet is explicitly about the the second and your question seems to be about the second. But the follow-up tweets suggest that Leike is asking about the first (also?). I think the second is not surprising at all, conditional on the first. But the first is quite mysterious.