Timothy Johnson

• Timothy Johnson 1 Mar 2024 18:04 UTC

  1 point

  0

  on: On the abo­li­tion of man

  when we’re talking about our influence on future generations, we’re almost always talking about the former, Bob-instead-of-Alice, type case.

  
  I admit I haven’t read Parfit yet, but can you give a concrete example of what type of influence you mean here?
  
  I think Lewis would disagree with this claim, or at least that type of influence is not what he has in mind. The example that he uses at the beginning of Abolition of Man is about a particular school textbook, and public education is the prototypical example of “changing a particular person’s values.”

• Timothy Johnson 4 Jan 2022 0:54 UTC

  1 point

  on: More ac­cu­rate mod­els can be worse

  For a more realistic example, this phenomenon of a more accurate model being worse is a common issue in database query optimization.

  When a user runs a SQL query, the optimizer uses statistics about the data to estimate the cost of many different ways to execute the query, then picks the plan with the cheapest estimate.

  When the optimizer misestimates the cost and chooses a bad plan, the typical solution is to add more detailed statistics about the data. But occasionally adding more statistics can cause the optimizer to choose a plan that’s actually worse.

• Timothy Johnson 19 Nov 2021 8:28 UTC

  3 points

  on: Does any­one know what Marvin Min­sky is talk­ing about here?

  For your second question, this paper describes the square root law, though in a somewhat different setting: Strong profiling is not mathematically optimal for discovering rare malfeasors | PNAS. (Incidentally, a friend of mine used this once in an argument against stop-and-frisk.)

  It doesn’t give a complete proof, though it describes it as a “straightforward minimization with a Lagrange multiplier”.

• Timothy Johnson 25 Oct 2021 19:02 UTC

  3 points

  on: Post­mod­ern Warfare

  This distinction reminds me of the battles in Ender’s Game.

  As I recall, Ender was the overall commander, but he delegated control of different parts of the fleet to various other people, as most modern militaries do.

  The bugs fought as a hive mind, and responded almost instantly across the entire battlefield, which made it challenging for the humans to keep up in large-scale battles.

• Timothy Johnson 24 Oct 2021 17:18 UTC

  4 points

  on: An­a­lyt­ics for per­sonal search history

  You might be interested in reading this paper, which tries to build a vision for just such a tool: Towards a Dynamic Multiscale Personal Information Space (mit.edu).

  As far as I know, nothing like it exists yet.

• Timothy Johnson 4 Oct 2021 20:11 UTC

  3 points

  on: Blink: The Power of Think­ing Without Think­ing (book re­view)

  As I understand it you’ve split this idea into two parts:
  1. Good—People can develop their intuition to make correct predictions even without fully understanding how they can tell.
  2. Bad—People frequently make mistakes when they jump to conclusions based on subconscious assumptions.

  I’m confident that (2) is true, but I’m partly skeptical of (1). For example, Scott reviewed one of John Gottman’s books, and concluded that it’s “totally false” that he can predict who will get divorced 90% of the time: Book Review: The Seven Principles For Making Marriage Work | Slate Star Codex.

  Another example which you’ve probably heard before is from the Superforecasting book: Book Review: Superforecasting | Slate Star Codex. Tetlock discovered that most pundit predictions were worse than random in the domains in which they were supposed to be experts.

  I believe that intuition works well in “kind” learning environments, but not in “wicked” environments. David Epstein distinguished these in his book, Range: Why Generalists Triumph in a Specialized world (davidepstein.com). The main difference between the two environments is whether there’s quick and reliable feedback on whether your judgment was correct.

  Chess is a kind environment, which is why Magnus Carlsen can play dozens of people simultaneously and still win virtually every game. This also explains the results of the Bechara gambling task. Politics and marriage are wicked learning environments, which is why expert predictions are much less reliable.

An­drew Yang on “The Priests of the De­cline”

• Timothy Johnson 1 Oct 2021 3:08 UTC

  1 point

  on: San Diego Less Wrong

  Did you really mean 6 AM? I was thinking of joining, but I don’t usually go anywhere on Saturdays before 10. :P

• Timothy Johnson 25 Sep 2021 7:55 UTC

  5 points

  in reply to: aphyer’s comment on: Book Re­view: Homo Deus

  I understood this sentence to mean that animals are intelligent insofar as they can pick up subtle body language cues.

• Timothy Johnson 20 Sep 2021 4:37 UTC

  2 points

  in reply to: DirectedEvolution’s comment on: This Can’t Go On

  There’s a few one-star Amazon reviews for the book that suggest McAfee’s data is incorrect or misleading. Here’s a quote from one of them, which seems like a solid counterargument to me:

  
  “However, on the first slide on page 79, he notes that the data excludes impact from Import/​export of finished goods. Not raw materials but finished goods. He comments that Net import is only 4% of GDP in the US. Here he makes a (potentially) devastating error – (potentially) invalidating his conclusion.
  
  While Net imports is indeed around 4% of GDP, the gross numbers are Exports at approx. +13% and Imports at approx. −17%. So any mix difference in finished goods in Export and Import, can significantly change the conclusion. It so happens that US is a major Net importer of finished goods e.g. Machinery, electronic equipment and autos (finished goods, with materials not included above in the consumption data). Basically, a big part of US’ consumption of cars, washing machines, computers etc. are made in Mexico, China etc. They contain a lot of materials, not included in the graphs, upon which he builds his conclusion/​thesis. So quite possibly, there is no de-coupling.”

• Timothy Johnson 7 Sep 2021 3:10 UTC

  10 points

  on: I read “White Frag­ility” so you don’t have to (but maybe you should)

  Thanks for the review—I appreciate that you spent the time to sift through these ideas.

  When you say that we need to “be willing to make uncomfortable, difficult changes,” do you have some specific changes in mind?

• Timothy Johnson 21 Jul 2021 18:56 UTC

  1 point

  in reply to: Slider’s comment on: Equiv­a­lent of In­for­ma­tion The­ory but for Com­pu­ta­tion?

  That depends what you mean by “computational problems”. In its usual definition, a Turing machine takes a single input string, and provides a single bit as output: “Accept” or “Reject”. (It may also of course run forever without providing an answer.)

  For example, the question, “What is the shortest path through this list of cities?” isn’t possible to encode directly into a standard Turing machine. Instead, the Turing machine can answer, “Is there a path through this list of cities with length at most k?”, for some constant k.

  If you don’t like this, there are ways to modify the definition of a Turing machine. But for the purposes of studying computational complexity, all reasonable definitions seem to be equivalent.

• Timothy Johnson 20 Jul 2021 0:59 UTC

  7 points

  on: Com­po­si­tion­al­ity: SQL and Subways

  I actually think SQL is not so easily composable. For me the gold standard is the Kusto query language: Tutorial: Kusto queries in Azure Data Explorer & Azure Monitor | Microsoft Docs.

  Kusto is a replacement for SQL, with syntax based on the Unix command line. I think it combines the best of both.

  For example, if you want to take a SQL query and adjust it to join with another table and add another column or two, you may have to make changes at several different locations. This isn’t just annoying to write—if your queries are in a version control system, it’s also annoying to review all of the lines that were changed.

  In Kusto, you can do this by just adding one line at the end of the query to pipe it to the new join. That makes interactive data exploration much faster.

  (Full disclosure: I work at Microsoft, and use Kusto for my job every day. I don’t work on Kusto itself, but it’s the one part of our internal tooling that I actually love.)

  Edit: After reviewing your definition again, I think maybe the distinction I’m describing is not part of composability. It should probably be called something like “extensibility” instead—the ability to take an existing solution and make small modifications.

• Timothy Johnson 20 Jul 2021 0:43 UTC

  1 point

  in reply to: Slider’s comment on: Equiv­a­lent of In­for­ma­tion The­ory but for Com­pu­ta­tion?

  Turing machines can only have a finite number of states. So you can do all computation in a single step only if the problem has a finite number of possible inputs. At that point, your program is basically a lookup table.

  Nontrivial computational problems will have an infinite number of possible inputs, so this strategy won’t work in general.

• Timothy Johnson 19 Jul 2021 18:30 UTC

  2 points

  in reply to: deepthoughtlife’s comment on: Equiv­a­lent of In­for­ma­tion The­ory but for Com­pu­ta­tion?

  Sorry, you misunderstood my point. Perhaps I’m being a little pedantic and unclear.

  For the cities example, the point is that when the problem domain is restricted to a single example (the top 100 cities in the US), there is some program out there that outputs the list of cities in the correct order.

  You can imagine the set of all possible programs, similar to The Library of Babel—Wikipedia. Within that set, there are programs that print the 100 cities in every possible order. One of them is the correct answer. I don’t need to know which program that is to know that it exists.

  This is why the OP’s suggestion to define “fundamental units of computation” doesn’t really make sense, and why the standard tools of computational complexity theory are (as far as I know) the only reasonable approach.

• Timothy Johnson 17 Jul 2021 22:46 UTC

  1 point

  on: Equiv­a­lent of In­for­ma­tion The­ory but for Com­pu­ta­tion?

  One problem with quantifying computation is that any problem with only a single instance can be trivially solved.

  For example,

  • How much computation does it take to write a program that computes the one millionth digit of pi? Easy—the program just immediately prints “1”.

  • How much computation does it take to compute the shortest path passing through the 100 largest cities in the US? Again, it’s easy—the program does no “real work” and instead outputs the list of the 100 largest cities in the correct order. (I don’t know the correct order, but such a program definitely exists.)

  For this reason, it doesn’t really make sense to talk about computation in terms of problems with only a single instance, or even a finite number of instances.

  Instead, computational complexity theory focuses on how algorithms scale for arbitrary input sizes, because that’s the only type of question that’s actually interesting.

  So can you clarify what you’re asking for? And how is this related to “qualitatively understanding things like logical uncertainty”?

• Timothy Johnson 22 Jun 2021 20:31 UTC

  4 points

  on: The Point of Trade

  Working for a salary is also fundamentally a form of trade. In my case, I trade my time and expertise in writing software, and my company pays me.

  Purchasing items is also trade—I trade the money I earned to buy goods and services from others.

  To completely eliminate trade, I think people would need to be able to magically wish anything they desire into existence. (If you’ve watched The Good Place, I’m imagining something like what Janet is capable of.)

  At that point, scarcity no longer exists. And since the main point of economics is to solve problems of scarcity, economics doesn’t have much purpose either.

• Timothy Johnson 21 Jun 2021 5:47 UTC

  3 points

  on: Why did the UK switch to a 12 week dos­ing sched­ule for COVID-19 vac­cines?

  Also, note that the UK recently decreased the time range from 12 weeks to 8 weeks: UK reduces gap between two covid vaccine doses from 12 weeks to 8 weeks (livemint.com).

  I’m not clear on the reason, but they have vaccinated 80% of adults with 1st doses, so perhaps they’ve decided that the need for 1st doses is dropping: Vaccinations in the UK | Coronavirus in the UK (data.gov.uk).

• Timothy Johnson 10 Jun 2021 2:26 UTC

  3 points

  on: Bad names make you open the box

  Thanks for writing this so clearly—I’ve bookmarked it to my list of favorite software engineering posts to share with others.

• Timothy Johnson 4 Jun 2021 18:57 UTC

  2 points

  on: Good Macro, Bad Micro

  Isn’t this just a normal consequence of scale? If you had to maintain 500 cars, you’d probably be more systematic. With just one car, the chances of it breaking down on any given day are pretty low.
  
  My own strategy is that I take my car to get inspected every time the maintenance light comes on, which is about 5,000 miles. The rest of the time, I don’t worry about it.