our_summary.txt

Yay. [LAUGHTER]Hi. My name is [inaudible]. masters student and I'm interested in
computer vision.the way they formulate them,  intrinsically relied on camp-
exponential search which, um,no matter how much compute you have, you're never going
 to, Lisp was- is- uh,had a lot of ideas that But of course, um,  it didn't last long.
 And for some reason,  which I don't quite understand,it killed off neural nets
research even though they fields, many of the techniques that we're gonna look at,  
for example, maximum likelihood,came from your statistics or games came from economics
, optimizations, fields, many of the techniques that we're gonna look at, for example,
maximum likelihood,   came from your statistics or games came from economics, optimizations,
gradient descent, hence from- was, you know,  [LAUGHTER]. and you can fool the
computer, um,  as- of- save our-our face recognition system to think that you're
actually, you know, someone else.  Um, and ProPublica this organization realized
whoa, whoa, whoa, whoa. Okay. Any questions about this?Maybe I'll pause for a 
moment. machine learning is going to be such a kind of  an important building 
blockof- that can be applied to any of the models that we kind of develop. 
its main driver of a lot of su- successes in AI because it allows you to,  in software
engineering terms,move the complexity from code to data. So examples like are linear
classifiers,  deep neural networks, um,and most of these models are the ones that 
people in machine learning um, use. In natural language or generating an image,um, 
you are, uh, can be cast in this way as well. blanks with numbers so that, um,  every 
row and column and three-by-three sub-block has the digits 1 through 9. Yeah.  
True. Right, um. [inaudible]Ah, ah, [LAUGHTER] um, how about if I say Alice is, ah, 
happy.  [LAUGHTER] Okay. How do you implement different, um, baselines or oracles?   
Which I'll explain later.We'll actually award some extra credit for students who help 
answer,um, other student's questions. [inaudible]The final project, ah, you can put 
on GitHub. [inaudible] simplify the output in terms of saying that  the substitution 
tells us we considered [inaudible] deletion peoples who considereda substitution or 
vice-versa by saying like an empty character.  And building up on your what you, um, 
what was said. Either we're doing, um, substitution,or, um, otherwise it's like the 
same character. Or we have to insert- each like to- to strings that we have at one 
point-  Uh-huh.[NOISE] So I can insert, um, insert where into-   [OVERLAPPING]So I 
can insert an "s", right? [NOISE] Okay, so which one of these?  Door "a" door "b" or 
door "c"? Yeah.So, recurrences are- are,I guess, one word I haven't really used, It's 
gonna show up, in you know,  search problems, MDPs, and, you know, games.um, is n 
minus 1, right?  Because remember insertion into s is deletion from t,that's why this 
is n minus 1.  [NOISE] Every point recurses three times. Don't try to be, you know,  
too slick at once.And so these are data points, you want to, let's say,  predict 
housing price from,you know, square footage or something like that. So if I actually 
had these two points, what should w be?  Okay, it doesn't matter. We'll, we'll compute 
it.So one principle, which is maybe another general takeaway is,  abstract away the 
details. It's, it's actually fine. And eta is just going to be this, uh, step size to, 
um,   keeping things under control.