How I Fail – Ian Goodfellow
veronikach.com
veronikach.com
In general, I don't know how I feel about these articles, they seem to be trending among my Twitter friends. I know that these interviews are meant as an encouragement. To show that even the great people of our field had to struggle, so we should not give up when we are still struggling.
However, these stories can also be understood negatively - if you didn't make it, you probably did not work or try hard enough. What makes this effect stronger is the selection bias - we do not see articles of the other 99% that work their tails of and do not become big in their fields. With a society that focuses on being the best, 'making it', etc. it leaves a lot of people feel inadequate, incompetent, or even depressed, even when they are big net contributors to society.
E.g. think of the 50-year old professor a moderately successful academic career. He/she is educating hundreds of students and having a big positive impact on their lives. But they are seen by others and themselves as failed academics who were not good enough to make it.
(I don't know what the solution is.)
So while there are certainly some people who, as Pickard said "made no mistakes but still lost," there are others who didn't put in sufficient effort and failed as a result.
I'd imagine that there are more in the latter camp than in the former. In your example, the professor is only a failure if their goal was something other than moderately successful academic.
Doing it successfully, defined as a measurable outcome, requires that you have a pretty accurate assessment of both the landscape, your abilities and the path to achieve the goal - with some factor considered for deviation from the charted path.
Depending on the scope of goal, the confidence interval on probability of success probably decreases exponentially with respect to time.
I agree, but for longer term goals, an extremely large amount of luck involved. Sometimes the right opportunities and persons just line up. I often realize that I'd be in a completely different (and probably worse) place if my 22 and 26-year old self didn't encounter the right persons and opportunities.
My current career path was largely unplanned, but is better than I hoped for or planned when I was in my early to mid-20ies.
Is it possible that stories like this are supposed to affect everyone differently?
-#1 Take it as encouragement => succeed, and be "happier", and share the success.
-#2 Take it as encouragement => Fail (and possibly move on to something more appropriate for you, and you are all the wiser for it)
-#3 Take it negatively => Quit at something and move on to something else. (see #2)
-#4 Take it negatively => Succeed anyways despite the negativity (see #1)
Maybe I missed an idea here, but how could it hurt anyone to share this information?
-#1 Take it as encouragement => But fail excessively the rest of your life?
-#2 Take it negatively => Give up and never try again?
It seems the positive outcomes outweigh the negative. And the negative outcomes don't seem reasonable to blame on an article.
"This account has been suspended. Either the domain has been overused, or the reseller ran out of resources."
Here's an archive link
https://web.archive.org/web/20180505190115/http://www.veroni...
Wow, what an amazing story.
And worth pointing out that Goodfellows work on unsupervised feature learning is clearly seen in his invention of GANs.
Edit, and:
I think it’s hard to extract value from negative results in machine learning because it can be so hard to tell what caused the negative result. A negative result might point to something very fundamental wrong with an idea, but it might also just be the result of a very small software bug, the wrong idea of the hyperparameter values to try out, too small of a model, etc.
So, so true. I read comments on HN on how important negative results are in ML and I look at all the things I’ve tried and someone else made work, and all I see is reporting negative results just discourages ppl from trying things.
This is a great interview.
I guess in this circumstance it actually was their loss.
Why? The impression I've gotten from a lot of time in industry is that the average Stanford student/graduate is, well, average.
I dunno, as someone that went to a public school with a 50% accept rate it seems difficult to imagine “academic superhumans” struggle with anything, ever.
I know a person who went to a comparable institution in my country, who was considered something of a prodigy by his peers (and me) both before being selected and after joining the institution, and who struggled a lot with his academic responsibilities. In fact, I know a few of them. One of them had to drop out of the institution after repeatedly trying to finish his academic programme, and after two years of extensions.
People struggle with things, in general. Everybody has their struggles, everybody has their demons. Even "superhumans". What we should be aiming for is not the lack of struggles — for that is merely the absence of effort — but perseverance in the face of it and growth from going through it.
What I'm saying is that, in my experience, the belief doesn't hold up. If it were possible to reliably identify "academic superhumans" and concentrate them in one place, we'd have absolutely indisputable evidence of it. The fact that we don't suggests that perhaps your beliefs about how Stanford's admission practices correlate with quality of graduates may be mistaken.
They also know how to present their answers well, and I think teachers subconsciously give them more credit when they're not exactly right. As a funny anecdote, once I let a top copy my homework because he had just come home from a family emergency. He actually got a higher score on that homework than I did, despite the content of our answers being the same.
I think an undergrad friend of mine put it best: "Cambridge doesn't take the top 1%, it takes a certain type of person from the top 10%"
There's 424,000 students got into university [0] and Cambridge enrols 3,480 out of 17,000 applicants [1]. So they don't take even 1% and only the top 4% of students even bother applying.
Edit: I suspect that 3,480 includes overseas students which makes the the percentage of UK enrolments is even less.
Further [1]:
> In 2006, 5,228 students who were rejected went on to get 3 A levels or more at grade A
Perhaps being surrounded by the top 1% makes you feel very normal/average, but there's certainly nothing normal about the Cambridge applicants and your figure of 10% definitely isn't correct.
I got accepted at Imperial with predictions of ABC (+ no extra curricular) to do Maths and Computer Science there's no way Oxford or Cambridge would have looked at me without 4 A's + a significant amount of extra curricular activities.
[0]: https://www.theguardian.com/education/live/2016/aug/18/a-lev...
[1]: https://en.wikipedia.org/wiki/University_of_Cambridge#Admiss...
I went to the college with the highest number of state school students. After participating in outreach events to get more people from a variety of backgrounds to apply to Cambridge I honestly think the biggest bottleneck is people from state schools not applying. They hear all the stories about strange interviews or being surrounding by posh people they have nothing in common with and just don't bother.
I was the only person in my (rather large) school year to get into Cambridge, and only a couple of people applied. I was told all sorts of strange things about the application process which turned out not to be true. I imagine it must be far easier to apply if you go to a school where almost everyone does and the staff are well versed in what to get ready.
> This account has been suspended. > Either the domain has been overused, or the reseller ran out of resources.
I mentioned that in a comment here and, for reasons I don't understand, was immediately downvoted. I deleted the comment in case there is some sort of HN 404 taboo I'm not aware of.
Never do this. If someone doesn't like your comment, then you are just letting them win by removing it. That's exactly what they want: for the comment not to exist. Don't give them the satisfaction; people who engage in censorship practices don't deserve it.
It's good to reflect on whether a comment that you posted that is downvoted really is problematic in some way, and if you see a real problem with that reflection to delete or edit it, but you shouldn't delete a comment when you don't understand a downvote because, frankly, lots of good posts get stray early downvotes, and that usually gets corrected over time.
In the meanwhile here is the post: (link: https://web.archive.org/web/20180505190115/http://www.veroni...) web.archive.org/web/2018050519…