Less Wrong is a community blog devoted to refining the art of human rationality. Please visit our About page for more information.

Open thread, May. 1 - May. 7, 2017

1 Post author: Thomas 01 May 2017 03:55PM

If it's worth saying, but not worth its own post, then it goes here.


Notes for future OT posters:

1. Please add the 'open_thread' tag.

2. Check if there is an active Open Thread before posting a new one. (Immediately before; refresh the list-of-threads page before posting.)

3. Open Threads should start on Monday, and end on Sunday.

4. Unflag the two options "Notify me of new top level comments on this article" and "

Comments (29)

Comment author: WalterL 01 May 2017 07:53:08PM 1 point [-]

I am constantly vexed by people who want time estimates before they give requirements. They teach us to say 'no' to this, in college, but it turns out you can't do that in the real world. So we are placed in the position of estimating how long it will take to code 'something', without enough detail about what the 'something' is to allow us to make an educated guess.

It's maddening.

Comment author: Screwtape 01 May 2017 08:45:13PM 3 points [-]

My standby is "From what you've described, it would take about $X days" where X is an estimate that assumes that they want everything associated with what they've said. (So, if they asked for messaging, I assume they want user icons, some kind of friends list or department groupings, and probably file transfer too.) There is a genuine use case for this! If your business guys are non-techies, they might know that adding some feature will probably earn Y dollars, but don't know how much it would cost to add. In those cases, they're really looking to estimate if it's worth it; if they're doing things right, the next step if it looks good is to actually sit down and write up a solid spec.

Of course, lots of places aren't like that. At my job, my estimate is continued with "but I could be pretty far off in either direction. Can you tell me more specifically..." and then from there try and guide them into giving me actual requirements. I write down the main points of what they're asking for, say it back to them when they're finished and make any revisions, and then email them that. If they agree that's what they meant, then I start working on it.

This still doesn't quite work. I've had my spec change radically two or three times a sprint (two week block) or been informed they assumed certain things were just obvious parts of what they asked for. (Like the messaging = file transfer thing. That was uh, non-intuitive, though I should have seen that coming since most major messaging systems have it.) In the end, the most important thing is to make sure you get paid for it. If I'm getting money to sit at a keyboard and think, I'll get over throwing out half the code I write.

If you want more advice on constructively dealing with it, I'm sure there are people here with more experience than I have. If you're looking for commiseration, you have mine.

Comment author: Viliam 04 May 2017 05:17:24PM *  2 points [-]

if they asked for messaging, I assume they want user icons, some kind of friends list or department groupings, and probably file transfer too.

If they want to edit something in a web application, I assume they will also want the ability to edit the data on a smartphone without internet connection.. and then upload the changes later when they get online, and of course have all the edit conflicts magically resolved correctly without any user intervention. Did I mention it is supposed to work on all kinds of smartphones and notebooks?

(Ok, I am lying, I actually didn't assume this... and it actually happened.)

Comment author: WalterL 02 May 2017 02:58:05PM 2 points [-]

Thanks for the advice, but your username makes me reluctant to take it :).

Comment author: Screwtape 02 May 2017 06:02:20PM 7 points [-]

My advice is always of the utmost sincerity, my dear Walterwood. The safest road to awful software is the gradual one- the gentle slope, soft underfoot, without sudden turnings, without milestones, without signposts. The duty of planning tomorrow's work is today's duty; though its material is borrowed from the future, the duty, like all duties, is in the Present. The truth is that wherever a developer has a meeting with a manager, there, whether they like it or not, a transcendental relation is set up between them which must be eternally enjoyed or eternally endured.

Put simply; to decide what the best use case is, you must ask what use the Management wants to make of it and then do the opposite. We know that is ultimately what they will decide. Only by our incessant efforts is the demand for infinite, or unrhythmical, change kept up.

Should you continue to find reason to distrust me, perhaps at some meetup or another I might take the opportunity to have you for dinner. >:)

Comment author: Lumifer 02 May 2017 06:18:16PM 0 points [-]

I might take the opportunity to have you for dinner

A delicious ambiguity :-)

Comment author: gjm 02 May 2017 07:36:44PM 0 points [-]

And one that calls back to the original source.

Comment author: Daniel_Burfoot 03 May 2017 01:02:01AM 0 points [-]

LessWrong: kind of an odd place to find references to Christian ethical literature.

Comment author: Screwtape 12 June 2017 05:44:14PM 0 points [-]

You'd think that, but rationalist spaces are pretty much the only places where people recognize what I'm referencing.

Comment author: Lumifer 02 May 2017 01:59:45AM 1 point [-]

I have a favourite phrase: "It depends" :-)

Comment author: Elo 02 May 2017 01:09:05AM 1 point [-]

support the idea of picking a number, multiplying by 20. Then considering if you should multiply by 20 again.

Comment author: TheOtherDave 01 May 2017 08:04:39PM 1 point [-]

Back when this was a big part of my professional life, my reply was "everything takes a month."

Comment author: Thomas 01 May 2017 03:56:47PM 1 point [-]
Comment author: Oscar_Cunningham 03 May 2017 07:39:55AM 0 points [-]

So in the comments of that page I was trying to prove that the number of possible molecules is infinite. I was trying to find some way of arranging an arbitrarily large molecule so that it didn't collapse gravitationally. I now think that a spherical shell with a constant thickness will work. The mass is proportional to the square of the radius, but it acts as though from a point at the centre, so the gravitational force on any point of the outside does not depend on the radius. Since small plastic spheres can exist, so can large ones.

Comment author: simon 08 May 2017 08:15:29AM 1 point [-]

This would in fact turn into a black hole since a black hole's radius is proportional to its mass. Also, it would collapse due to compressive forces since the force between any pair of hemispheres is proportional to r^2 but the surface to bear the force is only proportional to r.

Comment author: Oscar_Cunningham 08 May 2017 09:45:41AM *  0 points [-]

Yeah, I had already begun to worry about that second point, that the compressive pressure might be greater than I had imagined, but I couldn't see an easy way to calculate it. Your first point, that a black hole will form, is an easier way to see that this arrangement will fail. In fact I think the two points are different ways of looking at the same thing; for a black hole to form the shell must be pushed beyond the limit of its compressive strength.

What about a circular ring rather than a spherical shell? The mass is proportional to the radius, so it looks like it has a hope of working. But no, using the results of the paper here (behind a paywall, sorry) the pressure inside the ring rises with log(r), so it will eventually break.

:-(

Can we prove that all sufficiently large structures collapse?

Comment author: simon 08 May 2017 04:38:53PM *  0 points [-]

Sketch of proof: you proved that a stick collapses (compression scaling as Log(L)).

Well every connected object is either a stick, a curvy stick, or one of those things plus some extra atoms. So - prove that making things curvy or adding atoms doesn't help (enough). So, e.g. thickening the stick in the middle won't save it since you'd need infinite thickness.

for a black hole to form the shell must be pushed beyond the limit of its compressive strength.

hmmm... we've been talking as if in a space without dark energy. But with dark energy, a sufficiently large shell could be balanced by the antigravity of the dark energy within it, the acceleration caused should scale linearly with the radius. So that would be able to be under no stress at all. But I'm not sure it wouldn't form a black hole at large radius. As the interior gets bigger and bigger, eventually you get a cosmological event horizon forming so the interior forms a white hole - so light can't leave the shell to the interior. Since it's balanced, for symmetry reason's I'd expect the same to apply on the exterior. So you have this shell black hole between an interior and exterior both 'outside' the black hole.

Of course, this shell would actually be crushed by the stress in the radial direction, it's only not under stress circumferentially. But, now that we've got this example, we can extend to an ultralight aerogel (space-o-gel?) that balances the dark energy everywhere. I'd expect this to look externally the same as the shell example, so it should also eventually form a black hole. These are just guesses though - not actually calculated.

Edit: I'm now very suspicious of this analogy between the sphere and space-o-gel and will have to think about it more.

Comment author: Oscar_Cunningham 08 May 2017 10:21:11PM 0 points [-]

I agree with what you've said. I'll have to think about it for a while.

I did look at the case of a spinning ring and it seems stable. But I'm only using the Newtonian approximation, and you have to spin it at a speed that increases like sqrt(log(r)) so eventually that's invalid.

Comment author: simon 08 May 2017 11:42:12PM *  0 points [-]

I changed my mind on the space-o-gel though, at least for now.

Nice idea with the spinning ring. With relativity it should be fine as long as light itself isn't pulled in when going in parallel with the ring.

Comment author: Thomas 03 May 2017 08:42:03AM 0 points [-]

This physics does not impress me, but the mathematical idea behind which you imply, impresses me very much!

I want to say, forget about the real molecules, but this spiral structure per se, is an interesting concept.

An arbitrary large mas can be glued together without collapsing into a black hole?

Very interesting, indeed!

Comment author: madhatter 03 May 2017 02:44:44AM *  0 points [-]

1 is the smallest.

Comment author: Thomas 03 May 2017 08:45:20AM 0 points [-]

Could be. Seems possible, likely even.

Comment author: Oscar_Cunningham 03 May 2017 11:19:11AM *  0 points [-]

Perhaps 9 is smaller if you count cryonics (or some other non-mystical method) as reincarnation.

Comment author: Thomas 03 May 2017 05:58:04PM 0 points [-]

I am my yesterme reincarnation. But I don't remember everything. And haven't died last night.

It's a bit vague, now I see that.

Comment author: dropspindle 06 May 2017 02:56:31PM 0 points [-]

It doesn't seem like there's been any discussion on caloric restriction or intermittent fasting since 2014, and even then it didn't seem like any consensus was achieved. Have there been any more studies in the intervening years? Has anyone else started or stopped or failed or whatnot?

Here's Gwern's write up: https://www.gwern.net/intermittent-fasting

(I just noticed that their post was modified in May 2017, so SOMETHING new must have happened...)

Comment author: MrMind 09 May 2017 09:58:10AM 0 points [-]

There has been a publication, recently, about the long-term effect of IF: the conclusion is basically that it's as good as any other method of caloric restriction.

Comment author: madhatter 03 May 2017 09:41:21PM 0 points [-]

Hi all,

A friend and I (undergraduate math majors) want to work on either exploring a variant or digging deeper into the model introduced in this paper:

http://www.fhi.ox.ac.uk/wp-content/uploads/Racing-to-the-precipice-a-model-of-artificial-intelligence-development.pdf

Any ideas?

Comment author: MrMind 04 May 2017 07:20:26AM 1 point [-]

Some ideas:

  • different enmities for different teams
  • team capabilities drawn from a normal distribution
  • utility that does not depend from safety or enmity
  • only some team know their own / others capability
  • the probability of winning the race depending on capability, not simply the highest capable team being the winner

etc.

Also, what it says in note 2.

Comment author: madhatter 04 May 2017 02:11:31AM *  0 points [-]

One idea I had was to introduce a parameter indicating the actions of a governmental regulatory agency. Does this seems like a good variant?