Asking someone to solve an obscure mathematical problem (the Two Egg problem seems to be in fashion) doesn't do much except test whether the candidate has memorized the answer.
When not in an an interview I like playing with these kinds of problems, but if I look it up I'll probably find interview prep where it shows the answer as well.
Could you summarize just the problem please?
If an egg breaks when dropped from floor n, then it would also have broken from any floor above that. If an egg survives a fall, then it will survive any fall shorter than that.
The question is: What strategy should you adopt to minimize the number egg drops it takes to find the solution?. (And what is the worst case for the number of drops it will take?)"
Hmmm, that's an interesting question. It's the Price is Right closest without going over solver...
Practical answer, drop from the first floor and then the second. You're probably not going to make it to the third.
I don't want to put my actual answer in case others want to solve it. But I will say there's an obvious solution, but if you apply one of the hard problems of computer science you might eke out a bit of an edge.
EDIT: Actually I don't think that's right. I think the answer varies based on your expectations, and should be tuned after each run if you run the same test multiple times. I do have a good starting point in mind though.
That's The Price Is Right, not Wheel of Fortune ;-)
I was definitely picturing The Price is Right at least.
After thinking about it on the ride home today, I was not quite on the right track before. Again don't want to post my solution because the entire point in asking was so that people could see the question and work it through themselves.
The answer should be a form of binary search, but since you only have two eggs, you'd have to do a sequential search from the bottom floor. You have one chance at a mistake, so I would almost say start with the 2nd floor, but I know it will break on the first floor.
I don't see how that in any way gauges programming what so ever.
yeah, sorta. essentially you partition the 100 floors into some number of groups (i forget what partition length is optimal) and drop the first egg at each segment until it breaks, then go linearly from the previous one
Since you have 2 eggs you can start on the 33rd floor. It breaks you can use the 2nd egg from 1st to 32nd floor.
If it doesn't break, use the first egg on the 66th floor. If it does break, the answer is between 33 and 65, if it doesn't the answer is between 66 and 100. That cuts the maximal testing time time from splitting at 1/2 (1st egg at flr 50), to 1/3rd.
To optimize further, since in the bad case we'd have to search up to 34/35 floors, you can mathmatically find a number where you split the drops at an interval (IIRC the "real" optimal answer is 14, then 13, then 12, etc), which means you're doing less explicit checking.
At 14 - i, you have 10 drops to find a range of 13 - i floors in which the optimal path is found. So instead of a worst case of 35, you have a worst case of around 17.
I find the best way to approach this sort of problem is to first find one optimization, and then use the same process to find if there are further optimizations.
It does relate to programmer's ability to amortize worst case running time.
Once you realize that 1 egg requires linear search you just have to realize that with 2 eggs you have to decrease the level increase by 1 on each success.
Of course those must be some crazy genetically modified eggs.
I would guess the interviewer probably rates people who have memorized the solution far higher than people who have never encountered it and solved it with their brains.
But since we’ve gone there, many high power companies also have a lot of chuckleheads running around.
As an employer I don't care how great your skills are, if you can't walk me through your thought process for 45 minutes you're not going to be overly successful. If you can't incorporate feed back or engage when pressed to change your design you're probably not going to be successful.
Do your recruiting right and you shouldn't need the white board to simulate whether the candidate can code at all (I think white-boarding with the intent of looking for named algorithms is a waste of everyone's time), instead you want to see if they can synthesize information, engage with other people, and otherwise display the soft skills that tend to be much more useful metrics for employee success than coding aptitude.
Granted, my industry is known for not having deep coding problems to solve, this strategy might not be as useful in verticals that require more technology chops.
The post you're responding to makes a great point that this isn't the normal working environment for 99% of actually-writing-code developers. I discuss and iterate designs with my coworkers/leads/managers/architects just fine, but a surprise algorithmic problem you have to simultaneously solve, draw on a whiteboard, and constantly present to an audience while being timed isn't something I have been great at. I bet many others aren't either. I got good at it after a few interviews last time I was looking for a job; but if I hit the interview trail again today, I'm confident that I would flounder in that setting for a while.
Imagine if, instead, the candidate worked out a project -- maybe at home, or maybe in an interview room for some time -- then you take time to review and understand their work and then go through this back-and-forth process of discussing their work. This would also benefit the candidate's understanding of how you collaborate to problem solve in your work environment.
The problem with this approach is that it requires the interviewer to actually invest time to understand the candidate's work.
Anyway, I hope you see why many see it as a problem that the process for a candidate to succeed at a job interview is to practice interviewing skills rather than demonstrating their engineer skills.
As an investor, I don't care how great your skills are, if you can't come up with your new business idea on the spot and walk me through your thought process as you're doing it, you probably suck.. right?
Nah.
Unless you already have the answer, the first step is to come up with the solution. This could involve time spent alone just thinking, or brainstorming with coworkers, or researching the problem on the net, etcetra. What the right steps are depend on the problem, available resources, your background, even your preferred way of working.
The second step is to present and dissect the solution. I dare say this is the point at which the majority of engineers have no problem whatsoever talking about their thought process.
Do you work in an industry where engineers genuinely have to walk somebody through their thought process while they're trying to think up a solution?
If a problem comes up at a meeting and you don't have a solution on the spot, you take note and discuss it later or add it on the agenda for the next meeting. I think that's how people work.
I don't care about where, when, and how you came up with the idea. I care that you can present and discuss it once it's ready for that.
I think this is why many prefer the take-home test. They can focus on the solution. Then, at the interview, they can focus on the talk about the solution. It's also a great equalizer as e.g. people coming to work in a domain they have less experience with can take more time to research the solution space.
The complaints about homework taking up too much time may be valid, but I'd be most happy to trade 2 hours of interviews for about 75 minutes of homework and 45 minutes of interview.
Is there any scientific evidence to support that, or is that just industry hokum?
And communism probably redistributes wealth if it is administered correctly.
If companies want to hire based on general intelligence (not personality, team work, communication and so on) then they should administer IQ tests, which are based on a century of rigorous research. That would at least give you a fighting chance of measuring G.
Why do you think white boarding can measure general intelligence?