@startClosed @name("Documentation: Start Here!") documentation = "This model contains the information necessary to support a forecasting competition to predict the ratings of upcoming movies on IMDB, Metacritic, and Rotten Tomatoes. To participate in the competition, write a function matching this signature: ``` fn( time: Date between 2024-04-01 and 2024-06-01, movieUrl: Metacritic movie ID like \"boy-kills-world\" or \"challengers\", scoreType: One of [\"imdb\", \"metacritic\", \"rottenTomatoes\"]
start = Date.make(2024, 4, 1) end = Date.make(2024, 12, 1) fnPlot(fn) = Plot.distFn(fn, { xScale: Scale.date({ tickFormat: "%B %d" }) }) percentDiffOverTime(time, yearlyDiffDist) = { diff = toYears(time - start) (1 + yearlyDiffDist) ^ diff - 1 }
@name("My Estimate of B") export myEstimateOfVarB = normal(2, 5)
@startClosed standardRange = PointSet.support(PointSet(Sym.uniform(0.01, 10M - 0.01))) @startClosed missingSupport = MixedSet.difference( standardRange, PointSet.support(PointSet(Sym.uniform(0, 1M))) ) @startClosed
/* Describe your code here */ binomialSample(trials, probability) = trials == 0 ? 0 : List( trials, {|| sample(Sym.bernoulli(sample(Dist(probability))))} ) -> sum
/* How long does it take to read different books? How much does that cost, in counterfactual value, assuming that reading time is counterfactually valuable? This is a very simple table of estimates. Inspired by a previous Guesstimate model. */ @hide cost(wpm, valuePerHour, words) = { wordsPerHour = wpm * 60 words * (valuePerHour / wordsPerHour)
/* Describe your code here */ population = 12 @doc("Mean ounces of liquid people drink per day") ouncesPerPerson = 60 to 130 @doc("Of the liquid that people consume, how much comes from tap?")
// Generated mostly with Anthropic's Claude // Flight details departureAirport = "SFO" destination = "London" flightDuration = 10 to 12 // hours // Ticket price ticketPrice = 500 to 600
/* Describe your code here */ // import from the old Relative Values UI, I didn't write this code -- berekuk //Trying having one file here, to see if that makes editing easier. Will later move this around with some script or similar. ss(t) = SampleSet.fromDist(t) animalModule = { //Sentient Welfare
poundToKilogram = 0.453589 @startClosed caloriesBurned(minutes, met, weight) = minutes * met * (weight * poundToKilogram) / 200 @startClosed showcase(minutes, weight, mets) = { foo = inspect(mets)
@name("Charging efficiency") @doc("Power lost when charging") chargingEfficiency = 0.7 to 0.9 // Calculate cost @hide costToCharge(batteryCapacity, electricityRate, chargingEfficiency) = { loadInkWh = batteryCapacity / 1000 costPerkWh = electricityRate chargeCost = loadInkWh * costPerkWh / chargingEfficiency
import "hub:ozziegooen/helpers" as helpers /* Repurposed from this Guesstimate model: https://www.getguesstimate.com/models/187 Different dangerous activities feature different expected chances of dying. It's hard to estimate exactly how bad dying is, but one metric we can use is to instead consider how many hours in expectation risky events will cost you. For instance, something that has a 1/4 chance of killing you, can be
/* A replication of this Guesstimate code: https://www.getguesstimate.com/models/10465 This is a model of the expected value, in QALYs, to a random US citizen, due to Transformative AI. I expect that due to TAI, there's some probability (around .01% to 10%) that humans alive today will have a long-term positive outcome, meaning that they are sentient and experiencing welfare until roughly the end of the universe. There's also a chance that there will be an s-risk, and humans now will experience this same time, but
export round(num, n) = { asString = String.make(num) splitString = String.split(asString, "") if List.findIndex(splitString, {|r| r == "e"}) != -1 then { // Handle scientific notation parts = String.split(asString, "e") decimalPart = parts[0] exponentPart = parts[1] roundedDecimalPart = if List.findIndex( String.split(decimalPart, ""),
import "hub:ozziegooen/helpers" as helpers a = helpers.interpolate( [{ x: 0, y: 20 }, { x: 10, y: 80 }, { x: 15, y: 30 }], "linear" ) exampleWithInterpolation1(t: [Date(2023), Date(2050)]) = { yearsDiff = (t - Date(2022)) -> toYears Sym.lognormal({ mean: a(yearsDiff), stdev: yearsDiff }) }
/* Eventually, I'd want to organize my smoothie information using this data, from Holden, about Power Smoothies. I started this, but didn't finish, it would be more work to get right. https://powersmoothie.org/power-smoothie-recipes/ */
/* This is a test of the Code Formatter. Note: This functionality has since been fixed */ /** Summary: Scale.log() has errors the x-axis ticks, sometimes. */ logTickErrors = [ /**
/* Quick work on prioritization, going forward */ distChanges = [ { name: "Default to Symlog/log, when skewed", discussion: "Plot default should change between linear and either symlog or log, depending on the skew.", timeCostHours: 1 to 4, benefit: 1,
/* Experimental, in-progress model of AI safety compute, over time. */ startYear = 2023 endYear = 2080 yearRange = [startYear, endYear] transformative_ai_timelines(t: yearRange) = { dist = mx(logistic(30, 10), logistic(30, 30), [0.9, 0.4]) -> truncateLeft(0)
/* Simple Shapley Value Calculator Some math from GPT-4 plus this blog post: https://www.aidancooper.co.uk/how-shapley-values-work/ Inspiration from Nuno Sempere's Shapley Value Calculator https://shapleyvalue.com/ Exported Functions: