Prime Reacts: Coding Sucks and AI will steal your jobs - YouTube

Channel: unknown

[0]
coding sucks anyways Matt Welsh on the
[3]
end of programming again I still don't
[5]
know who Matt Welsh is I'm sure this
[6]
makes sense a former professor of
[8]
computer science and Google engineer
[9]
lead things generative AI will lead to
[12]
the end of programming within three
[14]
years
[18]
welcome to Costco I love you I love you
[22]
too buddy okay thank you very much
[24]
turning off alerts for a second dude
[26]
three years that's pretty ballsy okay I
[28]
always made the prediction or my
[29]
personal in internal prediction is that
[33]
Within
[34]
um
[36]
with within my kids lifetime that they
[39]
will never learn how to drive now my
[40]
kid's nine oh I'm getting kind of close
[42]
I don't know I don't know if it's
[43]
actually gonna work out you think an AI
[45]
is this an AI generated person not
[47]
showing his hands let me see that hands
[49]
you're you're an AI aren't you whoever
[51]
you are Matt Matt Welsh sounds like a
[54]
sounds like an AI name can we all agree
[56]
that sounds kind of AI you know what I
[58]
mean
[60]
yeah
[61]
all right not showing hands classic AI
[63]
move when you think about it
[66]
yeah yeah AI hands the AI hands problem
[69]
all right so let's let's read this uh
[71]
this month Matt Welsh a former professor
[72]
of computer science at Harvard's oh damn
[75]
credentialing hard spoken a viral Meetup
[77]
in Chicago Association for computer
[79]
Machinery the CAC the kakam
[83]
oh Associated for computer machine yeah
[85]
they went with ACM but there's this is
[87]
actually the cacum there's that you got
[88]
the Chicago in there explaining his
[89]
thesis that Chad GPI and get up copilot
[93]
represent the beginning of the end of
[94]
programming this is actually pretty
[96]
interesting uh because it is pretty good
[98]
but it is like
[100]
it's so scary how it gets it wrong and
[103]
it's because it almost seems like it has
[106]
the form
[107]
of programming but not the actual
[110]
programming itself in some sense
[113]
you know there's how there's like
[115]
context-free grammars
[117]
and uh blah blah blah blah I can't
[119]
remember all the terms now that I've
[120]
been out of school for 12 years but it
[122]
feels kind of like
[123]
this is this whole argument where we're
[125]
programming is like a larger subset and
[128]
though we can describe a lot of it you
[130]
know semantically correct with language
[132]
it's not quite ever correct and so it I
[136]
don't know I I'm very curious on this
[138]
Welsh thinks programming is on the cusp
[139]
of changing from a job that humans do to
[141]
one that robots will do thanks to
[143]
Technologies like Chad GPT wait what's
[145]
chat GPI I want to know about this AI
[147]
business what is this I am I just am I
[150]
living in the Stone Age on GPT already
[152]
my already Legacy code what the hell uh
[156]
or perhaps product managers or code
[158]
reviewers the two human roles uh he
[160]
thinks are relatively safe from robots
[162]
code reviewer
[165]
oh
[167]
oh
[172]
imagine
[174]
if you just become only a code reviewer
[178]
uh or let's see let's see
[181]
but don't expect to continue your career
[183]
as an actual programmer he says because
[184]
machines are taking over that role Welsh
[187]
who has a who's held a senior
[188]
engineering positions that's Google and
[190]
apple had written an article on this
[191]
topic on January 2023 edition of the
[194]
communications of the ACM magazine it is
[197]
very funny that they still even using
[199]
the term magazine come on get get up
[200]
with times however it wasn't clear up
[202]
until this ACM video presentation that
[204]
Welsh has funded a or founded a startup
[207]
to prove his theory fix the AI which he
[210]
is CEO of is uh is a self-described
[213]
automation platform for large language
[214]
models it aims to help business programs
[217]
custom software using you guessed it
[219]
chat EPT and similar machine learning
[222]
programs I can't wait for people to pay
[225]
this company a bunch of money to have a
[227]
program that just absolutely sucks all
[230]
the time
[231]
do you know what I mean like think about
[232]
how
[234]
think about I mean just just think about
[237]
all the discussion you have with a
[239]
designer with with people going over
[241]
iterations thinking about how it feels
[243]
like really talking about then and just
[246]
like having Microsoft front page 2.0
[249]
builds your website like imagine Netflix
[251]
built with Dreamweaver too it is just
[254]
gonna be
[255]
just the worst this is like my this is
[258]
my least favorite future is the shitty
[261]
AI program we all have to deal with
[263]
you're just gonna be on hold it's like
[264]
oh my goodness it's gonna be Robo calls
[266]
where you call about you you call up
[268]
someplace to get a refund and you have
[270]
to try to explain your situation to some
[273]
stupid ass robot that never gets it
[276]
right you're like yes one one six nine
[282]
it's just like no okay and you're like
[285]
no no es friba Espanol
[288]
no a spring on the way they espanol no
[291]
no and boom you're speaking Spanish to
[294]
some stupid robot because you can't get
[297]
the damn thing to work correct it's just
[298]
the worst it's just the worst computer
[300]
science is doomed
[303]
yeah this is the worst Welsh came off
[305]
the uh top rope at the start of his
[307]
presentation claiming that computer
[308]
science is doomed man talk about
[310]
clickbait uh I think the field is going
[312]
to change erratically I do agree with
[314]
that uh he explained if you think about
[316]
what is uh what is computer science as a
[319]
discipline it has always been about
[321]
really one main thing which is
[323]
translating ideas into programs
[329]
hmm
[331]
is that
[333]
could you say that that is true could
[335]
you say that computer science is really
[337]
one thing
[343]
or would you say that computer science
[350]
is the translation of programs into
[352]
ideas almost the opposite
[354]
it's like the technical way of thinking
[357]
about something concrete into more
[359]
abstract terms
[362]
is this just pitching a startup he went
[364]
on to say that computer programming has
[365]
evolved over the past 60 years trying to
[367]
to try and make it easier for humans to
[369]
understand code he compared Fortran in
[372]
1957 to basic in 1964. oh where's Cobalt
[375]
and then to rust in 2010. he said there
[378]
is there really isn't much difference
[379]
between those three in terms of ease of
[382]
use for human programmers
[388]
that that is not true
[391]
that is
[393]
dude start a [ __ ] TCP server in
[397]
1957. oh wait you'd have to travel to
[400]
the Future 19 years or was it 1974 that
[404]
TCP was formerly won I forget if it's
[405]
1974 1976 I always get vim's creation
[408]
mixed up with tcp's creation once in 76
[411]
one's in 74. dude what do you mean there
[415]
is no
[421]
and 57 like what was it in Fortran 57
[424]
you could never use the seventh column
[426]
if you put a character in the seventh
[429]
column it just doesn't what are you
[432]
talking about that is crazy talk this
[435]
look at what's happening in this code
[437]
compared to what you could do there was
[439]
no iterating and match statements and
[441]
pattern matching and collecting into a
[442]
colony via trade system quality is a
[447]
trait it's implementing the collect
[449]
trait and what
[451]
okay I am not happy about this right now
[453]
I'm not happy I am not happy right now
[456]
with this programmers uh let's see
[458]
programs that people are developing
[460]
today are just as complex
[463]
just as hard to maintain just as
[465]
difficult to understand and just as full
[467]
as bugs as they ever have been
[471]
statement is patently
[474]
are massively more complex today than
[477]
they were 20 years ago there's no way
[480]
you're saying this this has to be
[481]
written by an AI this just has to be
[483]
because this is too stupid this is too
[485]
stupid that someone said this out loud
[487]
that someone somehow thinks that today
[490]
is equally as complex as it was before
[493]
yeah sure there were Punch Cards before
[496]
and that was really inconvenient but
[498]
that is a physical representation of
[500]
inconvenience versus having to think
[501]
about all the stupid inputs places of
[504]
failure connections internet going down
[506]
try implementing RTP trying to figure
[508]
out any of this crap this is just
[510]
ridiculous
[511]
foreign
[513]
I'm getting myself a little hot right
[516]
now
[516]
I'm getting myself a little hot
[518]
ignorance runs of miles in the uh AI
[521]
hype Bros dude this this is oh my
[523]
goodness you're right is this like are
[525]
we watching web3 right now
[529]
is this web 3 right now are we watching
[531]
it because remember web3 is like ah the
[533]
financial system's actually currently
[534]
being Rewritten to use crypto I don't
[536]
know if you know this but crypto is
[537]
actually currently rewriting uh the
[539]
financial system right as we speak and
[541]
we're like no that can't be right I mean
[543]
you couldn't you couldn't actually do
[545]
that that easily like it takes a lot of
[547]
time and momentum and then you also have
[549]
this like whole problem of like
[550]
governments and Military and then their
[552]
money and you're trying to say you're
[554]
just gonna usurp it by not having a
[555]
military and for the people and by for
[558]
the people you mean you have to pay
[559]
hundreds of thousands of dollars to
[561]
figure out how to manage a Geth node I'm
[563]
gonna call bullet on that because that
[565]
doesn't seem correct something about
[566]
this article
[568]
smells
[570]
smells it was get up co-pilot that made
[572]
Welsh fundamentally reassess the meaning
[574]
of computer science co-pilot radically
[576]
changes the the way that we write code
[580]
he said and has at least in my personal
[582]
experience been just as remarkable and
[584]
profound way of accelerating my
[586]
development so I don't think I've gotten
[588]
any faster with copilot other than
[591]
logical boilerplate that's like the only
[594]
thing that co-pilot really is good at
[596]
copilot's kind of annoying copilot gets
[599]
it massively wrong copilot gets it so
[601]
wrong
[603]
like it's so dude okay fine [ __ ] it say
[606]
I mean sorry uh we're gonna go in here
[610]
um we're going in here and I'm just
[611]
gonna show you this okay delete this and
[613]
like this type uh video uh video frame
[616]
control oh my goodness I'm in the wrong
[618]
language at this point uh struct video
[621]
frame time stamp uh there we go time
[624]
stamp milliseconds let's go now look at
[626]
this audio frame good job Cola pilot I
[628]
call that logical boilerplate you nailed
[630]
it now I'm gonna go like this uh struct
[632]
video perfect that's exactly what I
[634]
wanted uh how about another one perfect
[636]
that's exactly what I wanted how about
[637]
imple video and I want to do a pub
[640]
function FPS and I want to return a
[643]
float 64 and let's see how about uh oh
[647]
let's see oh nice okay it's taking code
[649]
I already have in this project it
[651]
literally is just taking code that I
[652]
have in this project okay it already
[654]
it's taken damn it you're taking code
[656]
for my project here uh crap personal
[658]
let's just redo that I was wondering why
[660]
it was completing just so dang well
[663]
um maker Test 2 CD Test 2 cargo a net
[668]
Vim right here Source main go in here go
[671]
struct video frame go in here time stamp
[675]
uh because it even did Ms or us or Ms
[678]
like it just did it correctly which you
[680]
know a little uh audio nope see now see
[684]
there we go now we're starting to get to
[685]
all sorts of weird stuff now if I do
[687]
that perfect struct uh video is going to
[690]
be uh frames yep awesome now should be
[692]
able to do audio awesome impul video
[694]
let's go like this Pub function FPS
[697]
return F uh 64. oh nice 30. thank you
[701]
for that that seems nice self what do
[703]
you got here
[705]
no we're not going to do that how about
[707]
match uh self
[710]
frames.length
[713]
that's what you're going to go with
[715]
oh so that's how you calculate FPS you
[718]
just take the length of your frames and
[721]
your last time stamp
[723]
so let's just say this happens to be the
[725]
epoch since 1970 January 1st and that
[728]
just happens to be what you're gonna go
[730]
with somehow that seems wrong somehow
[734]
that is just it it looks like code I
[737]
would write
[738]
this looks like words I would use maybe
[741]
not in this particular order but these
[743]
look like words I would use this is
[745]
pretty good okay let's not do that one
[747]
here I'm going to help you a little bit
[748]
co-pilot here let's let's retry this
[750]
again how about this one frames.first uh
[752]
wait hold on is this oh whoopsies you
[754]
know what I didn't include that maybe
[756]
that maybe that's part of my problem
[757]
okay copilot maybe it's my fault maybe
[759]
it's my fault it's my fault okay let's
[762]
look at this one oh man this is looking
[763]
really good right can we all agree this
[765]
is really good
[766]
everybody everyone thinks this is good
[771]
anyone here thinks this is bad
[782]
hey guess what you're all [ __ ] morons
[785]
okay you know why because this has a bug
[787]
in it look at what it is think about
[789]
this statement first frame last frame I
[792]
have one frame what is the duration
[795]
right here it is zero what happens when
[799]
you divide by zero that's not great when
[802]
you divide by zero people don't like
[803]
dividing by zero that's really really
[805]
bad this thing doesn't even take it like
[808]
it just put such a sneaky bug in there
[810]
it's so sneaky it's so sneaky look at
[814]
that it's so damn sneaky because you
[816]
don't know like it's so hard to see
[818]
because it looks correct it looks
[820]
correct but it is so dang sneaky do you
[825]
know what I mean it is snakey sneaky
[827]
snakey snake eye right you just wouldn't
[829]
know that so if you you know if you do
[832]
self.frames.length throw in one of these
[834]
then you go in here and then you put an
[836]
n and then you go over here this is why
[838]
I knew what it was supposed to do if n
[839]
is greater than one now that was my
[842]
solution to this problem this is my
[844]
solution my solution is much smarter
[845]
right big brain big smooth brain over
[849]
here versus that wrinkly brain over and
[851]
wherever it's at I don't even know what
[853]
the hell's going on here
[854]
you have two frame though don't you
[856]
first and last
[858]
think about it for a second let's just
[860]
think about it just let's just hop over
[861]
here
[862]
let's just create a new one of these and
[864]
I want you just to think for a second
[865]
let's pretend you had a struct of one
[869]
item in it
[871]
now
[872]
which one's the where's the where's the
[874]
first element
[875]
Can you spot it can anyone spot what the
[878]
first album can you just type out the
[879]
letter what's the first element in this
[880]
one
[882]
ah X nailed it okay
[885]
good job good job good job first element
[886]
okay now
[888]
trick question atheist
[890]
where's the second element so I'm gonna
[893]
put a little little thing or where's the
[894]
last element shall I say sorry where's
[896]
the last element so this is first
[898]
where's the last element where could it
[900]
be
[902]
oh damn it you're right this is also the
[905]
last element
[907]
it's right there it's it's the same
[910]
thing right and so boom computer goes
[913]
explodiate because guess what it doesn't
[915]
work that way that's not how it works
[917]
first is last one there's one item okay
[919]
have you ever been in a golf tournament
[921]
and in your golf tournament there's only
[922]
two people playing so when you when you
[924]
inevitably lose that golf tournament
[926]
because you're terrible at golf you
[928]
claim you got second place in the golf
[930]
tournament you don't say you got last no
[933]
one says they got last okay everyone got
[935]
a trophy for being second in a golf
[938]
tournament everybody knows about being
[940]
second in the golf tournament okay it's
[941]
classic all right let's keep on going
[943]
this is taking way too long already he
[945]
thinks co-pilot is an incredible product
[946]
as I'm typing copilot does a very very
[948]
good job at often completing my thoughts
[951]
for me your thoughts must be wrong then
[954]
it reads my mind a lot more than I
[957]
really think it should
[959]
I got plenty of questions for you now
[961]
what's going outside your mind of yours
[964]
he said he added that co-pilot is a
[966]
fantastic productivity boost because it
[967]
saves me from having to contact switch
[969]
interesting uh I had such an insult
[973]
there I'm gonna let that go is this the
[975]
same picture over and over again oh no
[976]
this is just live picture and that's
[977]
just where he's at okay
[980]
um let's see let's see as good as it is
[982]
Welsh thinks the current version of
[984]
copilot is just the beginning absolutely
[985]
this would be insane if he did not think
[987]
that
[988]
just imagine if it's this is the end
[991]
this is the last one
[993]
we you know we tried it
[995]
and uh didn't work out it didn't work
[998]
out uh people kept getting things not
[1001]
quite correct so uh yeah we gave up
[1002]
right there uh there are only two things
[1004]
stopping copilot from getting much
[1006]
better he said more data and more
[1008]
compute
[1012]
clearly this is a man who has never npm
[1014]
installed I don't think there's a
[1016]
shortness of data anywhere in the
[1018]
typescript system you know what there's
[1020]
nothing anywhere there's there's nowhere
[1022]
there's absolutely nothing that's
[1023]
stopping right now
[1025]
plenty of data people I mean there is
[1028]
lit there is the is even an is odd
[1030]
package
[1031]
typescript is created every possible
[1034]
situation typescript is so bad at
[1037]
writing code it's so bad at writing code
[1039]
it is so bad at writing code Type
[1044]
squipped uh dot TS you can do something
[1047]
like this function quick sort
[1050]
uh numbers number and guess what
[1054]
copilot is going to copy it hey copilot
[1057]
just implemented correctly do you see
[1059]
this this is actually a correct
[1060]
implementation of quick sort right is
[1062]
that correct no it's not correct this is
[1065]
not correct you don't create memory and
[1067]
quicksort why are you creating memory
[1070]
why are you creating memory stop
[1072]
creating memory it's an In-Place quick
[1075]
sort algorithm
[1077]
oh my goodness
[1080]
oh
[1082]
[Music]
[1083]
it just pisses me off it's just wrong
[1087]
regularly
[1089]
it's correct in operations wrong in
[1092]
principle
[1093]
uh since both of those things are
[1095]
abundant he doesn't see any reason why
[1097]
co-pilot in a year or two or maybe three
[1099]
isn't going to get to the point of you
[1102]
type a few lines at the top of your
[1104]
source file and it just writes the rest
[1105]
oh gosh imagine how many bugs that you
[1110]
know what I swear co-pilot doing this is
[1112]
only going to be a function of the fact
[1114]
that there's so many programmers these
[1115]
days that don't even know how to type on
[1117]
their keyboard they type like 40 words a
[1119]
minute and they think it's perfectly
[1120]
fine never to learn how to type yeah bro
[1122]
you know I'll just use this thing like
[1124]
40 to 60 hours a week I don't actually
[1126]
need to learn how to use it why would I
[1127]
ever do that okay it doesn't make me a
[1129]
better programmer to know to you to know
[1131]
how to use my keyboard okay oh copilot's
[1134]
gonna type it for me
[1138]
you can imagine how I feel about you
[1140]
okay
[1142]
you can imagine how I feel about that
[1144]
okay
[1146]
have you considered that you can just
[1147]
ask Chad gbt to fix the bugs though
[1162]
so I can't believe I can't believe
[1164]
program exactly ending I just I can't
[1166]
believe we're actually on the cusp of
[1168]
programming ending this is uh
[1173]
wow
[1174]
um I'm sorry everybody don't learn
[1175]
programming it's not worth it we're
[1178]
about to uh lose all of our jobs
[1182]
I'm the stupid one for learning how to
[1184]
type I should have instead of learning
[1185]
how to type I should have played
[1187]
fortnite or something man Welsh had some
[1189]
intriguing predictions about how this
[1191]
will change the Team Dynamics of
[1192]
software development basically he only
[1193]
sees a couple of roles for humans once
[1195]
programmers are phased out product
[1197]
managers and code reviewers the role of
[1200]
product manager won't change much he
[1201]
says imagine somebody that doesn't
[1203]
actually code in only reviews code guess
[1205]
how good they're going to be at
[1206]
reviewing code
[1208]
you know part of code review is the
[1212]
extrapolation of experience into
[1214]
practice right it is it is literally
[1217]
taking the years of you coding and
[1220]
trying to apply it to a situation to
[1221]
understand what's happening you know how
[1223]
many look Goods to me like first off
[1225]
imagine how much code is going to be
[1226]
generated by copilot okay they're gonna
[1228]
make a lot of code second off we all
[1229]
know if any code review is more than
[1232]
like 20 lines 30 lines 40 lines maybe L
[1236]
looks good to me well I don't know what
[1239]
happened in this file it's written by an
[1241]
AI must be good
[1244]
the AI spits out the code in like a few
[1246]
seconds the AI I know dude
[1249]
it is
[1251]
it's hard in the let's see in the near
[1252]
future instead of uh handing a PRD I
[1255]
don't even know what that means to an
[1256]
engineer team and waiting six weeks or
[1258]
so for them to implement it well said
[1260]
you could just hand the PRD
[1262]
pneumonia ready device uh to the AI and
[1266]
the AI spits out code in in like a few
[1269]
seconds
[1274]
listen
[1275]
of course one of the early lessons of
[1277]
generative AI is the output of those AIS
[1279]
won't necessarily be correct
[1286]
what's early it's
[1287]
two to seven years from now like what
[1290]
the hell is early in generative AI
[1292]
I'm just gonna hand it a product spec it
[1295]
'll just [ __ ] out
[1296]
code that looks correct
[1300]
and then what I'm gonna figure it all
[1302]
out oh my goodness how do we know that
[1305]
the code works
[1306]
[Music]
[1307]
foreign
[1314]
oh no how do we know that it's good how
[1316]
do we know that it's right of course we
[1318]
need to have a thorough testing
[1321]
and the testing is very very important
[1323]
who writes the tests again if there was
[1326]
only some way to generate these tests
[1328]
based on requirements
[1330]
[Music]
[1337]
yep all the cicd stuff and everything
[1339]
that we've invested in over the last few
[1341]
delegate decades are still relevant here
[1343]
devopsy got yourself a job still he
[1345]
thinks that humans with programming
[1347]
abilities will be tasked with reviewing
[1348]
and reading the AI generated code and
[1350]
making sure that it works and is doing
[1352]
the right thing as for programmers and
[1354]
Those About to join the field they will
[1355]
need to become teachers of AI rather
[1357]
than coders it's all about teaching the
[1359]
AI not right in the computer program
[1360]
okay uh oh wow he has a company trying
[1365]
to do this who would have thunk
[1367]
having a set of let's see having set the
[1369]
scene and scared the bejebers out of the
[1371]
200 plus computer scientists who turned
[1374]
uh tuned into this ACM meeting Welsh
[1376]
then explained how his new company is
[1378]
taking advantage of
[1389]
yeah
[1394]
oh oh my goodness no oh oh
[1400]
oh my goodness huh guess what hey guess
[1404]
what you got a problem did you know you
[1406]
got a problem you got a real problem
[1408]
guess what No One's Gonna No One's Gonna
[1410]
program anymore I trust me trust me bro
[1412]
I'm I'm from Harvard
[1414]
by the way I started a company
[1416]
in which we made it so that they could
[1419]
do exactly what I just warned you about
[1421]
uh with fixie the idea is you give it
[1423]
the description of what you want to do
[1424]
and fixie takes it and let's see using a
[1427]
large language models plus agents that
[1430]
can connect to external systems it
[1431]
produces a result for you either the
[1433]
answer to the question or calling an API
[1435]
or invoking a tool or making a change in
[1437]
a database these are all the things that
[1440]
models can do and critically we are
[1443]
doing this not by writing a whole ton of
[1445]
code we are doing this by teaching the
[1446]
AI models how to do it
[1456]
this just kind of looks like Siri
[1459]
see Json agent
[1462]
it got brains Json agent got braids it
[1465]
fetch all the GitHub issues
[1467]
[Music]
[1469]
serialize all the GitHub issues to a
[1472]
string please uh node.js just exceeded
[1475]
string limit coding sucks okay it became
[1477]
clear in this presentation that mattos
[1478]
he has a big invested interest in making
[1480]
traditional coding obsolete really since
[1483]
his company aims to capitalize on the
[1484]
trend of AI programming however given
[1486]
his own long and successful career as a
[1487]
computer scientist you also have taken
[1489]
the prediction seriously uh okay and
[1492]
this is I mean this is isn't that
[1494]
literally the appeal to Authority
[1495]
fallacy isn't that like it
[1498]
how could they not say this incoming I
[1500]
know exactly I mean isn't that isn't
[1502]
that literally just by saying
[1504]
because I can find a doctor right now
[1507]
that will tell you putting coffee up
[1508]
your ass
[1510]
has a lot of health benefits
[1512]
I'm just saying do you want to put
[1515]
coffee up your ass right now I'm just
[1517]
saying is that what you want because you
[1518]
know what I can find an expert 20 years
[1520]
of experience
[1522]
get that coffee right up the pooper
[1526]
because that's going to give you long
[1527]
life boy boy that's gonna give you long
[1529]
life
[1533]
is putting coffee
[1535]
up your ass healthy question mark doctor
[1541]
coffee ediba the rise of coffee enemas
[1544]
don't try this because you know why I'm
[1545]
saying this because this actually
[1546]
happened like the early 2000s was all
[1548]
about putting coffee up the up the booty
[1552]
this is this is not this is not a joke
[1554]
because there was doctors that actually
[1556]
was suggesting this
[1559]
Facebook do you put coffee
[1562]
that's it good clean bucket first one no
[1566]
no no no no no I'm not reading that I'm
[1569]
not reading it I'm not reading it I'm
[1571]
not reading that I'm not reading that
[1572]
not reading it chat look away look away
[1575]
Chad
[1576]
the origins of the coffee script I know
[1578]
this is oranges at the coffee script uh
[1581]
let's see Welsh thinks that coding sucks
[1582]
anyways so just let the robots do it do
[1585]
something else with your time He
[1586]
suggests such as being unemployed yeah
[1588]
you know just go do something else why
[1589]
don't you just call be employed
[1592]
um
[1594]
you know where I think AI is going to
[1596]
really shine at is in like very
[1598]
mechanical things uh I think ai's
[1600]
already write a router code because
[1602]
that's like a very very very mechanical
[1606]
thing
[1607]
um I I believe there it's already
[1609]
probably I I wouldn't be surprised if
[1610]
it's not deeply invested in
[1614]
uh doing all the like die generation for
[1617]
for Hardware all that kind of stuff I'm
[1619]
positive it does
[1620]
most of everything but that's all very
[1623]
mechanical
[1625]
I know this is the Wall Street Journal
[1627]
telling you to stop eating breakfast
[1629]
because it sucks anyways writing
[1631]
computer programming is not the best use
[1633]
of the time for everyone
[1640]
what is the best use of my time
[1642]
easier said than done for people
[1643]
employed as
[1644]
[Music]
[1647]
hard
[1648]
okay that is the best ending to this
[1649]
whole thing
[1651]
you know writing computer programmer is
[1653]
not the best time for everybody easier
[1655]
said than done for people writing it for
[1657]
their job
[1658]
[Laughter]
[1661]
I live in a world of dump I I swear we
[1664]
live in the stupidest Universe this is
[1666]
like the writers of 2023 are the
[1668]
single-handed like best
[1672]
they they are the best writers of any
[1675]
season of earth right like they just
[1676]
keep coming up with just the the
[1678]
greatest curveballs I have ever seen in
[1680]
my lifetime somehow somehow the writers
[1684]
this year they are so good we gotta hire
[1686]
them for 2024. do you think it's do you
[1688]
think chat GPT is writing 2023
[1693]
important
[1695]
yeah so that on the other hand so that
[1698]
greater dispel whoever okay gosh you got
[1700]
to see this now this is actually cool
[1702]
can a large language model summarize to
[1704]
our meeting that would be sweet right
[1705]
like uh just yeah that would be sweet
[1708]
I'd totally be down for that I'm done
[1709]
with this article this article is
[1710]
actually crazy talk I can't keep on
[1712]
reading it it's like I it's like every
[1714]
sentence I read
[1715]
okay let's read more we'll read more
[1717]
okay we'll just read a little bit more
[1718]
but it feels like every sentence I read
[1720]
just got a little bit stupider each time
[1722]
Microsoft is including that feature and
[1724]
Bing yeah not surprising did you see the
[1725]
reason for uh the Mark Anderson uh I
[1727]
think I had a lot of perspective how
[1728]
machines will not take over in this way
[1730]
yeah I let's see plot twist no one could
[1732]
have predicted it
[1734]
all right here we go here's the full
[1735]
presentation oh that's the end got him