Nov. 7 -- Former Microsoft Chief Executive Officer Steve Ballmer gives "Bloomberg Studio 1.0" host Emily Chang one of his brain teaser interview questions.
Yeah I am going to take advice from this idiot. Just look at Microsoft’s performance during his so called leadership. The contrast pre Ballmer and post Ballmer to the dreadful performance during his reign is shocking.
But we only have Steve's word that he chose and stayed with some number. I'd refuse to play unless he wrote the number on a piece of paper at the start so he could not adjust his "number" during play.
@@a_voice_in_the_wilderness you're low, I mean you're high. you're low. you're low, I mean you're high... I don't think Ballmer even remembered how the game worked...
No, because that is exactly the strategy that the game counts on you using. She failed to recognize that the game was designed to defeat that very strategy. Performing a binary search on a range of 1-100 can take up to 7 guesses (log2 100=7).
@@HouseOfHam there's no "strategy". It was given that a binary search is being used, so it was a simple statistical assessment as to whether the expected outcome is beneficial. She learned the question and how to binary search from nothing. That is impressive.
@@pieluver1234 Its a pretty straight forward thing to do that most people do naturally. Its literally just picking the middle point of the potential search and she didn't even do that correctly after the first 2 guesses. Thats why Balmer got screwed up on the high low when she said 60 - he was expecting 62/63. I think he may have even changed the number because she started deviating from what he expected.
Even still she approached it the binary search way. Ballmer goes on to say "it turns out it's a binary search masquerading as a game" and basically says she failed, but like... she got it, she did a binary search. I'd say that's impressive for a non-developer.
In the current interview process in 2020, this would be considered to be a Easy level problem. It was certainly easier to get hired into big tech even a decade ago than it is now.
This isn't really a leetcode style problem, but i would rate it on the easier side of medium. You just need to know that the optimal strategy is a binary search and that the time complexity of a binary search is worse case O(logN). Therefore, the person proposing the game has an optimal strategy which makes the binary search go to its worst case. The expected required search time would therefore be log(100) = 6.64. At this point, the player is losing money.
@@jopmota This question was oriented towards evaluating a programming mindset and was a pretty basic example of a question that you might pose to a Dev. But the interview process "back in the day" wasn't solely "brain teaser" questions. There was also plenty of digging into past work and experiences as well as skill evaluation and "team compatibility". When questions like these were asked it was primarily to evaluate "how" the interviewee approached the problem - not necessarily whether they came up with the answer to the problem that you expected. (I rarely used questions like this when I was interviewing people, though I did have friends that used them.)
How much he earned from one job is irrelevant. He still only had one job, so him being job interviewed is still something he doesn't have much experience with.
I think it's better to get advice from the dude who gives the interviews rather than someone who has been through a lot of interviews. (This way you are informed about what the interviewer looks for. If someone got interviewed a bunch of times and had a bunch of jobs, he might've been bouncing around a lot for an unfortunate reason.)
Yeah luck placed him as the roommate. Honest hard work made him money. Opportunity presents itself to many people, only a few can seize it and make something out of it. You cant write off his successes at MS even though he has undoubtedly done taken some/many dumbass decisions.
@@jbas7525 @J Bas Do you know Bill Gates personally, or or you just speculating? Because even the most ruthless people are usually pretty generous towards personal confidants. And really? Tell me, How may BILLIONAIRES do you think there are in the world? A few thousand at best, you think this mediocre middle manager at best, would have become a BILLIONAIRE. You think just "pulling your weight" at a job gets you to become the top fraction of a fraction of a percent in wealth do you?
How to pass any interview: be the one that offers value to the company, act and demand according to a person that actually brings value to the company. Understand that the company is worthless without people like you and without customers.
Without companies, a person will have no jobs :D we all work for a company, the Government is a large conglomerate :D , own your own business, you work for yourself, there is no escaping that fact :D
“Do you want to play a game?” “No sir I’m not here to play games I’m looking to feed my family” Commenters: appreciate the serious analysis but this comment was made for *amusement* purposes only. No one would seriously verbalize this at a job interview. That’s why it’s a *joke*
While this is the factual reason why people apply for jobs, it is a red flag thing to say during a job interview according the recruiters that I know. In our glorious capitalistic society, companies are looking for what you can give them, not the other way around.
His Stanford/Harvard degrees aside, University teaches you that networking and having friends is important... Winning friends in high places is a difficult skill.
@@chinchilla505 Kevin o'Leary of Shark Tank was asked about his MBA he did. He asked if it was worth it. He said the education wasn't worth it, but the connections he made was worth it. The connections made him the money.
@JuuzouSusuyaScythe TwoHandedGsKing I don't think it's sad. Some people are happy being at the business hierarchy all their lives, all they want to do in life is clock in and clock out from their job. Some people are happy with the opposite, they want to run companies like Microsoft, they want to become millionaires. Is either wrong?
If anyone knows anything about computer science this is rather straightforward. You use binary search, so log2(x), but you have that x = 100 so you can just calculate the worst case scenario that way ... log2(100) = 6.64 therefore implying you lose.
If you know anything about anything you'd recognise the risk is the fucker is going to bullshit you with whatever number you provide, logarithmically or no. ;)
that's the average number of searches, but to figure out if you want to play the game or not, you actually have to write out the full expected value, because there's a tiny chance you win $5, but 0% chance you lose $5
correct. You need 7 tries (it's between 2^6 = 64 & 2^7 = 128 so 7 as the integer closest to number of tries to reach 100) to find the number by binary search so you're always going to lose money, unless you get lucky. Ballmer's "explanation" about more numbers to lose on than to gain is either misinformation or he doesn't understand the binary nature of the problem (which I find hard to believe since he was a math major)
That's right, but the interesting twist is that the guesser should expect that the person she is playing against will choose the worst case as their number. This let's you skip the binary search and only guess numbers which requires the maximum number of searches.
Anyone with a tech, math background will intuitively and immediately reach for Probability in their mind. Any engineer fresh out of school should have said I am not playing.
For the the people saying this is a stupid question sit down and pay attention, you might learn something: This isn’t some dumb game. He’s testing your ability to solve problems and more specifically your ability to apply known computer science algorithms to real world problems. The question is should you play the game, or better put, are you likely to win?, he’s not testing whether you say yes or no but HOW you come up with the correct answer, which is NO, you should not play. Answer: you have 5 tries to pick correctly before you start loosing money, so you want an algorithm that is very efficient. In this case the most efficient algorithm would be a binary search which has a big (O) of n log n. 2^5 = 32 but the list is a 1-100 which means it’s possible that after 5 choices you still won’t have the correct number. 2^6 = 64, that’s still not enough . 2^7 = 128, which is more than 100, and that’s perfect. So with binary search the most it could ever take is 7 tries to get the right answer. It could be less but worst case scenario is 7. The problem is he knows we’re using binary search because it’s the most efficient way, but since it’s a known algorithm it has very specific rules about which numbers to select. He can pick a number that will guarantee that we use all seven picks to get the right number. But we start losing money after 5. Therefore, no, you should not play the game. Programming is about solving problems. If you can’t handle or apply such a basic algorithm why would they intrust you with handling much more complicated problems at their company.
If you want to know the _actual_ exact answer, you calculate the expectation value where you multiply the probability of the occurrence times the money gained or lost from each case. The full calculation is: = 1/100 * (2^0*5 + 2^1*4 + 2^2*3 + 2^3*2 + 2^4*1 + 2^5*0 - 37*1) = $0.20 per game. Basically, each round the number of ways (out of 100 possible random numbers) that you could get the right number increases by a factor of 2. So there's 1 way to get it right on the first round (answer 50), 2 ways on the second round (answer 25 or 75) and so on until after the sixth round (which is 2^5 ways). Up to this point, you have had 63 opportunities to get the correct answer. Assuming you've played optimally but didn't win by the sixth round, you will be guaranteed to win on the seventh round which has 37 numbers left. Of course, this all assumes that he randomly chose the number. He can obviously choose a biased number if he wants which changes everything.
We learned that Steve's number likely wasn't fixed at the start of the game. He simply said "hi" and "lo" in such a way as to force her make lots of guesses.
I would answer: "If you give me 7 choices I'll play, because each time you say high or low you decrease the options by half and 100 divided by 2 seven times narrows the correct range to less than one, hence selecting the correct number". A more difficult and messy way to ask this would be same but ranging from 0 to 128 (both included). Some people will get fucked by this when their answer is wrong.
How to really answer the question: How about if I lose, I buy you lunch. BOOM, Win Win. You've built a relationship with the interviewer. And paid for all your wrong guesses.
The issue with the method is its focuses so narrowly on a unique trait serving a subset of the job functions, for example, programming. This very programmer will eventually be promoted to a manager and then to an executive position and if s/he lacks other required skills such as imagination, empathy, teamwork, etc. is bound to be failed. Many technology companies including Microsoft have changed their criteria for success after passing through lackluster times.
When she said 60, the guy really hesitated because the interviewer was really closed to guessing the number and if that was the case he will not be able to prove his point, fortunately for him the interviewer was so nervous and anxious that then she said almost a randomly lower number. The guy was right anyway, you should not play this game unless the number range is +/- 10.
Having worked for both Microsoft and Amazon, I must say Amazon's approach to interviewing is far superior to Microsoft's (at least the traditional Microsoft "cult of the puzzle" interview approach that Ballmer is describing here). Microsoft wants to know if you can solve a bite-sized intellectual teaser in the moment. Amazon wants to know if you have exhibited a pattern of behaviors over time that align with their leadership principles. The Amazon approach is much better. IMHO.
I think Google do lots of basic computer problems (like search and sorting algorithms). and they are famous for the "reasonable approximation" questions like "How much does this building weigh" with the intent of listening more for the reasoning process than the answer
You don't need to do the odds on playing the game, logically speaking if somebody is asking you to play a game for money then they most likely know that the odds are in their favor and thus you shouldn't play.
If you want to know the _actual_ exact answer, you calculate the expectation value where you multiply the probability of the occurrence times the money gained or lost from each case. The full calculation is: = 1/100 * (2^0*5 + 2^1*4 + 2^2*3 + 2^3*2 + 2^4*1 + 2^5*0 - 37*1) = $0.20 per game. Basically, each round the number of ways (out of 100 possible random numbers) that you could get the right number increases by a factor of 2. So there's 1 way to get it right on the first round (answer 50), 2 ways on the second round (answer 25 or 75) and so on until after the sixth round (which is 2^5 ways). Up to this point, you have had 63 opportunities to get the correct answer. Assuming you've played optimally but didn't win by the sixth round, you will be guaranteed to win on the seventh round which has 37 numbers left. Of course, this all assumes that he randomly chose the number. He can obviously choose a biased number if he wants which changes everything.
@@F.G.30.4.91 It is the correct answer because you _should_ play the game based on how he described it in this video (and assuming a fairly random number). In the actual job interviews, I'm fairly certain that they phrase it differently so that the expectation value is different and negative. Ballmer just said it incorrectly here. He should have said that the guess after the $1 prize would instantly be a $1 loss, not a $0 wash. This would make you subtract 2^5 and 37*2 instead of only subtracting 37. The expectation value in this new situation would be -$0.49 so you wouldn't play the game if it was phrased in this way. Ballmer just fucked up here.
My answer would be, "I get significant utility from playing the game. For me, to play this game is fun and challenging and so I will engage in the game for the potential cost of a couple of bucks. What is life anyway when you don't take any risks? Had Microsoft not taken risks, it wouldn't be where it is today." BAM, Hired! (hopefully)
I would have said "No" but not because of the probability, but because my opponent has complete control over the situation and with no oversight whatsoever.
Thats one question from one interview, which is like a waterdrop in an ocean, compared to what you'll have to pass in order to get hired. At least thats my expirience with the interviews in software development.
You don't have to use binary search. First I would make him write the number so he can't cheat. Then I would go binary but with an offset that would derail his expectation of me dividing at exactly half. So first I would go 49, then 74, then 61. This probably changes the probabilty to my side (I didn't count).
Yes everyone he changed the number on the fly to prove a point. Besides his question was would you play the game or not; not actually play the game and see if you can beat the odds.
This is an old, tired, strategy example. They explained this very game at the orientation day of my MBA program. I suspect Mr. Ballmer picked it up from his MBA, too.
The objective was to determine if the candidate would agree to play the game. Good tease. Emily Chang didn't get it. Strategist don't agree to play games they will lose, disrupters (rebels) answer by flipping a middle finger, consensus builders don't answer yes or no - instead they create their own answer.
But Microsoft doesn't want rebels, if you have 300000 employees and they are all "rebels" and "disruptors" they will go bankrupt on shitty projects in one year.
first he says the number will be between 1 and 100, the statement is exclusive meaning the pick will be 2-99, so 98 possible choices. with binary search you would get 1x$5, 2x$4, 4x$3, 8x$2, 16x$1, 32x$0, 36x$-1. calculating the expected value using above (outcome / 98 * dollar value) added up you get ~$0.21. therefore statistically take the chance. BUT he doesn't say he will pick a number at random, so it could just be say no bc he will pick one of the 36 negative value options.
First time ever, I found Ballmer somewhat interesting. Given, that he has a background in Mathematics I would have really loved to see how he further elaborated into his question.
The expected value of this game is actually positive if the number is picked uniformly at random (it's 20 cents). So unless you have 1$ in your account, you should play it. Also, if you suspect Balmer is going to pick numbers that make it hard assuming you'll use a simple binary search, you can take that into account and come up with a modified strategy that takes advantage of this to make even more money.
Emily actually had the right idea there, but it wasn't clear the actual question was whether or not you should play the game. I think she thought Steve was asking if she wanted to try!
Mathematically, it would take you a total of 7 guesses, if you did it correctly and always rounded down if it was higher and rounded down if it was lower. As an absolute maximum, 7. The chances of you doing it in 5 tries or less is significantly less.
Isnt it as simple as 5 guesses of 100 means 19:1 against getting it right ? So if you guess 50 then 25 or 75 so in 2 guesses you're in the right quartile but left with 3:25 against before you start losing money.
it is a binary search classic problem, where the problem gets halved by 2. resulting in log n base 2 . So here n is 100, log 100 base 2 somewhat equates to 6.6 (rounding to 7). So in worst case, the interviewee will need 7 attempts, but Steve is giving just 5 chances to prevent you from making a loss.
Imagine the number is 2: I say 50, low,(we're at 4$) 25, low, (we're at 3$)12, low, (we're at 2$)6, low, (we're at 1$)3, low - (we're at 0$) and the choices are: 2,1 or 0. On average, you'll pay Steve Ballmer. Anyone familiar with merge sort will be using it as their metaphor.
If you want to know the _actual_ exact answer, you calculate the expectation value where you multiply the probability of the occurrence times the money gained or lost from each case. The full calculation is: = 1/100 * (2^0*5 + 2^1*4 + 2^2*3 + 2^3*2 + 2^4*1 + 2^5*0 - 37*1) = $0.20 per game. Basically, each round the number of ways (out of 100 possible random numbers) that you could get the right number increases by a factor of 2. So there's 1 way to get it right on the first round (answer 50), 2 ways on the second round (answer 25 or 75) and so on until after the sixth round (which is 2^5 ways). Up to this point, you have had 63 opportunities to get the correct answer. Assuming you've played optimally but didn't win by the sixth round, you will be guaranteed to win on the seventh round which has 37 numbers left. Of course, this all assumes that he randomly chose the number. He can obviously choose a biased number if he wants which changes everything.
When I'm asked trick questions like this in an interview, I already know it's not going to be a fun place to work at, lol. I'm fine with questions testing to see what I'll do in a situation and the right answer is open... but not a trick question with just one right answer the interviewer is judging people by.
That's not even close to a trick question. You can look at it as a statistics questions or a computer science (algorithm) question, but if you don't remember either approach from your university education, what business do you have working at a tech company?
ummm there's no way you could win that game if the number he's thinking about isn't written down before or there is a proof that he didnt just change that number on the go just to win
If the quiz he asked wash to determine when a candidate decides to step back and think of his/her option to understand their personality, that indicates a preemptive determination factor from interviewer’s perspective and they are only selecting a candidate who thinks like they do. Why go about it through such an abstraction layer?
He would only interview high positions in leadership, not technicians, so the question has merits on how one would respond from a decision making standpoint.
