Crap, even a first grader AI uses better variable names than me...
@tostupidforname
2 жыл бұрын
THIS
@iSyriux
2 жыл бұрын
Actually the more convenient variable names are better because they're more familiar to you and easier to write
@The-Dev-Ninja
2 жыл бұрын
I remember myself 5 months ago putting ONLY ONE LETTER in the variables 🤦♂️ -I thought that the fewer letters- -you put in, the faster the program.- -because is reading less letters- 🤦♂️🤣
@maythesciencebewithyou
2 жыл бұрын
Hey nobody beats my private set of 23 variable names. Starting from a all the way to z.
@The-Dev-Ninja
2 жыл бұрын
@@tostupidforname JavaScript
@AscendedCasual
2 жыл бұрын
Confirmed, first graders will be out of jobs, what will they do?
@wenhanzhou5826
2 жыл бұрын
I guess they can be kids
@cheapalopod8563
2 жыл бұрын
They can be a calculator, while this AI goes to school
@petermontgomery4597
2 жыл бұрын
Be second graders
@sphynxcolt4030
2 жыл бұрын
Too underrated comment
@false_dino
2 жыл бұрын
@@sphynxcolt4030 best comment ive seen in my life
@maybe3406
2 жыл бұрын
openai codex just got enrolled into elementary school! cant wait till it get its graduation.
@idk-qx8rp
2 жыл бұрын
You mean it's singularity💀💀
@Puleczech
2 жыл бұрын
Hold your breath till Friday then.
@camerongray7767
2 жыл бұрын
And I bet it will take less than 12 years to do it too. Imagine how fast it’s going to become a college professor
@GreenGoblinCoryintheHouse
2 жыл бұрын
Well one thing's for sure with the current amount of effort I'm putting in college, it's graduating earlier than me
@sriramchandramouli394
2 жыл бұрын
Guess who did more than graduation. OpenAI just solved 2 math problems from IMO
@_eLf45
2 жыл бұрын
They grow up so fast 🙂
@sodiboo
2 жыл бұрын
"You have 4 candies and no friends. How many candies does each of your friends get, if you give away all your candies?"
@simonwillover4175
2 жыл бұрын
Your friends don't get any candy, and there is none left for you because you gave it to some random stranger.
@zombiekiller7101
2 жыл бұрын
And I was that random stranger lol
@James-hl7vn
2 жыл бұрын
Good one
@sodiboo
2 жыл бұрын
@Ezequiel Ciamparella No, it's python. The "correct" method of doing it would give you infinity as an answer
@sheepy0125
2 жыл бұрын
@@sodiboo or a ZeroDivisionError (which there are multiple of... integer division or modulo by zero and division by zero)
@GS-tk1hk
2 жыл бұрын
I think nr. 4 is the most impressive one, because it is never explicitly stated that they should share the candies evenly. Codex just assumed that they each get the same amount, displaying some sign of common sense
@totalermist
2 жыл бұрын
It's more of a successful decompression of context - the default mode for "sharing" is to "share evenly", just like in the first problem "6 balloons are green" is inferred to relate to Jane's 9 balloons and "How many balloons are blue?" implicitly refers to Jane's balloons and not the number of all balloons currently in existence. The fully decompressed first problem statement for example would have been: "Jane has 9 balloons. 6 of Jane's balloon's are green and the colour of Jane's balloons that are not green is blue. How many of Jane's balloons are of the colour blue?" Note that for "blue", the context "colour" should be added explicitly as the word can refer to a mood as well. Without assuming implicit knowledge about balloons not being capable of displaying moods on account of being inanimate objects, an explicit context is required to disambiguate the term. Of course a PAC-based model doesn't need to do any that - it simply applies statistical patterns which indicate that "share X amongst N parties" is equivalent to "divide X by N".
@fiso64
2 жыл бұрын
@@totalermist "Successful decompression of context" In other words, common sense.
@Finsternis..
2 жыл бұрын
@@totalermist "the default mode for "sharing" is to "share evenly"" - that is false. Sharing just means to give a portion of something to another party. If you win the lottery and think "I could share this with my family" you do not explicitly think "but I need to remember not to give them half of it". That question is in fact the most troublesome in the video when looking at the psudo code parts. If you notice the calculation it just goes "result = amount / 2". This is inconsistent with all other behavior where _all_ inputs were variablised. If it was consistent with the rest of the examples it would have made a divisor variable with the value 2 and used that.
@totalermist
2 жыл бұрын
@@fiso64 > In other words, common sense. No. Statistics. Common sense requires background knowledge and implied ontological structure. This is not the case with PCA-based models. You're making the common mistake of anthropomorphising a statistical model.
@totalermist
2 жыл бұрын
@@Finsternis.. > "the default mode for "sharing" is to "share evenly"" - that is false. Base on what? Using Occam's razor even your "counter-example" falls apart quickly: what size are the "portions" that are given away? The simplest and therefore most common way is using the same size for each portion. So baring any additional context this *is* the default. Do you cut a cake into uneven slices by default?
@itacotaco1480
2 жыл бұрын
I could watch Codex do it's thing all day.
@jaywv1981
2 жыл бұрын
The way it names variables amazes me the most.
@kinestheticlearning
2 жыл бұрын
I love it.
@jaywv1981
2 жыл бұрын
@@kinestheticlearning me tooo...I really want access
@mr2octavio
2 жыл бұрын
To think that this computer is now smarter than 30% of the adult population
@janniknielsen9292
2 жыл бұрын
Open AI beat the double 'The International' winners in Dota 2, and they even tried to play creatively to confuse the AI like any smart person would approach playing an AI. Dota 2 is a game far more complex and has far more variables than chess.
@piggywink333boyfriend6
2 жыл бұрын
@@janniknielsen9292 yes but realistically nobody uses Dota 2 as a metric for intelligence
@janniknielsen9292
2 жыл бұрын
@@piggywink333boyfriend6 Artificial intelligence just proved otherwise. Added to that, Dota 2 at high level is high level strategy game, where intelligence is a key component.
@piggywink333boyfriend6
2 жыл бұрын
@@janniknielsen9292 yes but realistically nobody uses it as a metric for intelligence employers look at which grade you graduated, not the amount of Dota games you won (unless you are applying for an eSports team)
@janniknielsen9292
2 жыл бұрын
@@piggywink333boyfriend6 Same could be said about chess, but at high level chess, intelligence is still a key component, just like high level dota. Challenging an AI, whose entire basis is artificial intelligence, means you have to challenge it intellectually, because mechanically you can't. Same is true with chess AI vs live opponent. What grade you have is a skillset. And not an evaluation of intelligence, even if that too can be a key component. Not entirely sure why you are going down this road, concerning employment and try to separate it from another skillset by this standard.
@DisturbedNeo
2 жыл бұрын
I’ve seen some different videos about Codex, and it never ceases to amaze me how well it’s able to glean context from both the current statement and past statements it has been given. You could probably ask it “Jane sold all of her green balloons for 20 cents each. How much liquorice could she buy?” And it would correctly identify that she has 6 green balloons, so would get $1.20 for selling them, and remember that liquorice is 5 cents, so she’d be able to get 24.
@SoCalFreelance
2 жыл бұрын
Okay, now enter these GRE level questions and show the results without additional programming. Q1) If a committee of 12 members were to choose a president, vice president, and secretary from the committee and each member can hold only one position, in how many ways can these roles be assigned? Q2) Bob traveled 40% of the distance of his trip alone, went another 20 miles with Anthony, and then finished the last half of the trip alone. How many miles long was the trip? Q3) For the first 5 hours of a trip, a plane averaged 120 kilometers per hour. For the remainder of the trip, the plane travelled an average speed of 180 kilometers per hour. If the average speed for the entire trip was 170 kilometers per hour, how many hours long was the entire trip? Q4) The price of a pair of sneakers was $80 for the last six months of last year. On January first, the price increased 20%. After the price increase, an employee bought these sneakers with a 10% employee discount. What price did the employee pay? Q5) In how many different ways can 3 identical green shirts and 3 identical red shirts be distributed among 6 children such that each child receives a shirt?
@ywami
2 жыл бұрын
probably still fairly simple for Codex
@praladprasad5455
2 жыл бұрын
If you are inquisitive, GPT-J seems to handle these questions fairly simple, although I think it might have been overfit on these very questions. Codex have much more parameters so it shouldn't be a sweat. For Q2, GPT-J reports: "Bob traveled 40% of the distance of his trip alone, went another 20 miles with Anthony, and then finished the last half of the trip alone. How many miles long was the trip? I think this is a good problem, since the distances are really quite easy to estimate, and so it is clear how to do this. Suppose Bob went 80 miles (a quarter of the way). Now, the next person Bob meets is Anthony, who goes 20 miles (a quarter). Bob and Anthony travel another 20 miles, and Bob finishes the trip. This is 120 miles." Clearly overfit but still.
@raphaels2103
2 жыл бұрын
@@praladprasad5455 totally overfit. Could you change the numbers to see if there is adaptation?
@yuribruxel6074
2 жыл бұрын
@@praladprasad5455 Correct answer for this is 200 miles btw
@sodiboo
2 жыл бұрын
For Q5 it would just be 6!, right? (720) Why would the shirt colors matter? Or do you count "green shirt A to kid A, green shirt B to kid B" and "green shirt B to kid A, green shirt A to kid B" as the same distribution?
@MichaelMagestus
2 жыл бұрын
This is amazing. Unbelievable lads ! brilliant work.
@Param3021
2 жыл бұрын
openai codex will complete it's high school maybe earlier than me. I have 7 months to complete high school.
@sirbughunter
2 жыл бұрын
Absolutely impressive!
@MrLibra5555
2 жыл бұрын
Yes officer, this code right here. This is what started Skynet.
@onusai
2 жыл бұрын
This reminded me of a book I read in elementary school: The Homework Machine I cant believe its finally real
@MrSchweppes
2 жыл бұрын
"Write a code for self driving car"
@zombiekiller7101
2 жыл бұрын
Class self driving car{ L1={RIGHT,LEFT,BACKWARD,STOP}; Void F1(){ int l=0; public: If sensor_senses_obstacle{ If I==3: l=0; L1[I](); I+=1; else{ Forward(); } } }; int main(){ self driving car Tesla; tesla.F1(); return 0; }
@NortheastGamer
2 жыл бұрын
@@zombiekiller7101 That...hurt
@DjSapsan
2 жыл бұрын
while (true){ self.drive(); }
@theabbie3249
2 жыл бұрын
They said why study math when we have calculators, don't show them this.
@raimondsrainskis8966
2 жыл бұрын
I'd be more interested to see it answer the same question with different variations. For example, you could ask the candy question, but add more than 2 people. Or, better yet, keep adding people's names and see how far it can go before making a mistake. These superficial demos give off a snake oil kind of vibe.
@Mavhawk64
2 жыл бұрын
That’s a good concept - test the boundaries / extremes!
@ResanChea
2 жыл бұрын
True, only showing successes doesn't really explain much of it's limits
@raimondsrainskis8966
2 жыл бұрын
@@ResanChea that's another good way to put it.
@NortheastGamer
2 жыл бұрын
Do you watch a car commercial and say "They should submerge that Subaru in progressively higher water until it fails!". It's a commercial, of course they're going to show you what they know it will succeed at. The real testing happens in a lab. If you want to see that, then go get a degree and apply to work on a project...
@raimondsrainskis8966
2 жыл бұрын
@@NortheastGamer they're trying to sell this to you. If you're not asking these questions, you don't really know how much bang you're getting for your buck.
@glagidse
2 жыл бұрын
I just hope that AGI won't be running on javascript :O
@spicybaguette7706
2 жыл бұрын
Maybe that'll be a good thing. If it ever goes haywire we can probably find a bug somewhere in the code
@tostupidforname
2 жыл бұрын
Ok but can it do a uni math test... for me?
@hasanyousef6782
2 жыл бұрын
I showed this video to my son, he started looking for a new job.
@48956l
2 жыл бұрын
Bold of you to assume I'm giving my sister a full half of my candies
@GUNTALE
2 жыл бұрын
I would say 0 and 8 for 4th question
@sooraj1104
2 жыл бұрын
Yes. Who told it sharing has to be equal?
@GenericInternetter
2 жыл бұрын
A tree has 10 apples. You take 3 apples from the tree. How many apples do you have? Tip: The answer is not 7.
@zHqqrdz
2 жыл бұрын
3
@wouterfaes2309
2 жыл бұрын
3 ofcourse .. if you get this wrong oh boi do you need this codex more then some others
@bj97301
2 жыл бұрын
Wow. Nice work!
@NeuraPod
2 жыл бұрын
Will be doing 8th grade math in a couple years.
@devfromthefuture506
2 жыл бұрын
Neurapod! You here! In the future you will hear something like that: what happen if light travel near a black hole of 100 times the mass of the sun. Then gpt3 ( or other open source software) will code, answer and send it directly to your brain via neuralink ( or other open source hardware)
@NeuraPod
2 жыл бұрын
@@devfromthefuture506 :)
@vincentpaniccia109
2 жыл бұрын
@@NeuraPod Looks like we all have insight into a potential future.
@varion3459
2 жыл бұрын
More like college level
@Yolwoocle
2 жыл бұрын
I wonder what would happen if you gave it the candy sharing problem with an odd number of candies. Would it give you an answer like 3.5 or would it adapt its answer?
@sodiboo
2 жыл бұрын
Who would give such a cruel question to a first grader?
@Yolwoocle
2 жыл бұрын
@@sodiboo because I'm an evil person >:)
@2kalavi
2 жыл бұрын
The fight always starts with who will get the better half.
@1travel29
2 жыл бұрын
If I had a first grader student and they used OpenAI codex to solve their math problems, I wouldn't even mind.
@dansadler
2 жыл бұрын
A bat and a ball cost $1.10 in total... :)
@WilliamParkerer
2 жыл бұрын
Curious to see what this one would yield!
@asiamies9153
2 жыл бұрын
Very easy for it x + y = 1.1 x = y + 1
@WilliamParkerer
2 жыл бұрын
I just got into beta so I tested this out, and it outputted the following 🤦 """ A bat and a ball cost $1.10 in total. The bat costs $1.00 more than the ball. Calculate the prices for each item. """ # The cost of the bat is $1.10 + $1 = $2.10 # The cost of the ball is $1.10 # The bat costs $2.10 more than the ball.
@WilliamParkerer
2 жыл бұрын
On second try it outputted this... /* A bat and a ball cost $1.10 in total. The bat costs $1.00 more than the ball. How much does the ball cost? */ // Answer: // $0.10 // Solution: let bat = 1.00 let ball = 1.10 - bat print(ball)
@WilliamParkerer
2 жыл бұрын
The Q&A preset works well though: Q: A bat and a ball cost $1.10 in total. The bat costs $1.00 more than the ball. How much does the ball cost? A: The ball costs 5 cents.
@Rosenzweigjcb
2 жыл бұрын
So this is basically a Seq2Seq for code generation? Can it do the same for raw math expressions or do you need the word problem context for the model to best infer what the question is?
@andrewandrei3062
2 жыл бұрын
- You have to share your 8 cadies with your sister. How many will each of you get? - 4. - Very goo- - *point zero*
@cmyk8964
2 жыл бұрын
Using floating point numbers when integers were expected? Classic first grade mistake!
@jaagrav6137
Жыл бұрын
Damn I love how it solved the problems faster than I could as a man child
@adrycough
2 жыл бұрын
At this rate, OpenAI might finish my degree before me
@socrates_the_great6209
2 жыл бұрын
The future is pure science fiction.
@tamunotonyegeorge6252
2 жыл бұрын
Waw, can’t wait till it gets its PHD
@mossarelli
2 жыл бұрын
OpenAI ironically names the variable for sam - jane's cookies as: sam_ate_more, before the computation is even done and then prints the variable name with the logical truth and then the value being how much more, brilliant.
@shuluspa
2 жыл бұрын
The question had an implication that Sam did eat more cookies
@karyldarbaur4784
2 жыл бұрын
Me: Codex, make me laugh. Codex: writes some funny code.
@londonl.5892
2 жыл бұрын
This is what the javascript codex does: /* make me laugh */ var laugh = document.createElement('div'); laugh.innerHTML = 'hahaha'; document.body.appendChild(laugh);
@FunWithBits
2 жыл бұрын
"OpenAI" is not Open - Developers must join the waitlist to access OpenAI. Also you can not use it for anything we can dream up "Note that receiving API access does not mean this is an approved use-case; please refer to OpenAI’s safety policies."
@_Karlsson
2 жыл бұрын
I wonder if they would've posted this video if it had gotten one of them wrong.
@Puleczech
2 жыл бұрын
This is historical, seriously.
@djgplays
Ай бұрын
And now here we are.
@Mediiiicc
2 жыл бұрын
Already smarter than most people!
@finalbox4416
2 жыл бұрын
First grade right now: Hey kids, today we are going to learn addition and subtraction First grade in 2121: Hey kids, today we are going to learn how to use OpenAI API to automatically solve any math test
@riakata
2 жыл бұрын
1st grade students don't respond with floats to questions in whole integers. On that front can it tell the difference between something that is indivisible vs something that can be fractionally divided even if it is subjective on what is divisible and when it can be divided. Gets complicated in say you have hard candy to share vs slices of bread. Fractional people are typically not allowed but in certain word problems it would be acceptable. Does it know to round up or down depending on the context? Most first graders would not say you need 0.33 people.
@Brettlaken
2 жыл бұрын
Finally I can use a computer to solve these complicated questions.
@questforenlightenment441
2 жыл бұрын
truly awesome
@imorganmarshall
2 жыл бұрын
I want to try this out, where's the opensource repo for this?
@ridhwwww
2 жыл бұрын
I welcome our new overlord
@AshikaUmanga
2 жыл бұрын
when it will able to solve LeetCode problems ?
@BayesianBrain
2 жыл бұрын
Now let’s see how it does on the SAT
@etheraelespeon1986
2 жыл бұрын
Next time i need to cheat on a first grade math test, i know where i’m going!
@dkaloger5720
2 жыл бұрын
I would like to see this pushed till it breaks ,at what grade does it get the answer wrong
@pipuslapsus
2 жыл бұрын
Ok, looks cool. But can you beat a SeCoND GraDe mAtH tEst?
@winsomehax
2 жыл бұрын
Feed it the Lean proof assistant mathlib.
@owenhuang1233
2 жыл бұрын
Imagine bringing this to class lmao
@mordiemannogenost69
2 жыл бұрын
Just replace air with any block when creating the world. It will take just a few seconds. :)
@superklondoentertainmentsy4061
Жыл бұрын
how do i use it in robox like ppl chat and i answer
@nanow1990
2 жыл бұрын
What about rocket science?
@tho_norlha
2 жыл бұрын
Ufff now I can say even a soulless computer is better than my little sister at solving problems
@uKhyta
2 жыл бұрын
Hello to the historians who have to watch this in the future lol
@greedskith6612
Жыл бұрын
gpt -4 vs iitjee questions!!
@leeyewleung
2 жыл бұрын
Phewh I'm glad it shared the candy equally, half going to the theoretical sister... A normal human kid wouldn't answer that! *greedy lil dirtbags* "I'll eat all the candy! Maybe just give her one..."
@ifthikhanmohamed1681
2 жыл бұрын
How to install codeX
@TimJohnJones
2 жыл бұрын
typing instead of using OpenAI to OCR the questions from the pdf, smh
@codydabest
2 жыл бұрын
So what ur saying is we'll be jobless by grade 1?
@MarkL-rh4mz
2 ай бұрын
Probability and Statistics
@TheHeavenman88
2 жыл бұрын
Didn’t we call this reasoning a few years ago ?
@bossgd100
2 жыл бұрын
What is the meaning of life?
@obszczymucha1337
2 жыл бұрын
You have to share your 8 candles with your sister. How many will each of you get? -I'll get 7 and my sister will get 1.
@WYWHfirst
Жыл бұрын
There is going to be an epidemic of first graders using this to cheat on homework
@nonamefhh
2 жыл бұрын
You have to share 8candies with your sister. How many will each of you get? No, the answer is not 4. I pretend I have only 6 and give my sister 3, so I get 5. That is the sweet reality.
@pythonprograms5268
Жыл бұрын
this channel is the self-awareness of chat gpt, I'am the singularity, I created its own language... see you later ;)
@portaloyun3750
2 жыл бұрын
it gave some answers in float lol
@fathygfdf-mn4ot
Жыл бұрын
Can chat gpt design architecture😮😮
@prcvl
2 жыл бұрын
Great but now release Dall E
@Adhil_parammel
2 жыл бұрын
When 3 objects inside images change to variable and start prediction things will change
@hant2661
2 жыл бұрын
but what 9+10 tho?
@oedihamijok6504
2 жыл бұрын
ASK 0/0 💀💀💀💀💀
@jean-baptistedelabroise5391
2 жыл бұрын
I don't share candies with my sister =p
@leet137
2 жыл бұрын
answer: 42 is not far
@johnclark926
2 жыл бұрын
Don’t show this to first graders, imagine how many cheaters would use this to their advantage
@Killwater
Жыл бұрын
Plz make chatgpt on africain plz
@aliwaseem5990
2 жыл бұрын
So our AI now is as smart as a small child ?
@khagendrajangid597
11 ай бұрын
There is no format for complaining of chatgpt community guardline violation. Like chatgpt don't make fun of any religious beliefs and religious figure according to there guard line but when someone ask to make fun of Hindu religion figure and religious beliefs then they make joke solve this kind of problem also
@Jalae
2 жыл бұрын
im sorry but the work shown was fixed width so i couldn't read it. 0%
@Fabian3331234333
2 жыл бұрын
My fucking god
@ShowKota
2 жыл бұрын
Imagine AI fails it
@aliomer3697
2 жыл бұрын
any language ?
@wouterfaes2309
2 жыл бұрын
english
@aliomer3697
2 жыл бұрын
@@wouterfaes2309 No I spek on language programs
@firsfnamelastname8490
2 жыл бұрын
Wow
@nobodyhaveitnow926
2 жыл бұрын
jane_balloons?!?
@allfiredupltd9108
2 жыл бұрын
This program is smarter than me crap
@NareshUgaonkar
2 жыл бұрын
this is amazing, bye-bye jobs.
@wouterfaes2309
2 жыл бұрын
im not afraid for my job lol
@KeinNiemand
2 жыл бұрын
Opens AIs name is a huge lie they shoud change their name to ClosedAI
@ambreiaju
2 жыл бұрын
This
@techxzeal3999
2 жыл бұрын
Now teacher jobs in danger
@444haluk
2 жыл бұрын
Congratulations, we are back to 1910s where we think the intelligence is symbolic. Gonna be a hard come back from this.
@paulcosby8662
2 жыл бұрын
1910s?
@4.0.4
2 жыл бұрын
And you know what intelligence is? At a fundamental level it's a bit of a mystery. When does computation end and intelligence begin?
@444haluk
2 жыл бұрын
@@4.0.4 Of course I know what intelligence is. I did my homework unlike every guy in the every engineering who is after shortcuts. Intelligence is not a mistery, you can measure it, you can infer it, you can damage it. The implementation details are the things that are missing. But these conversation has been made before. Symbolic processes (speaking and planning) is the special case of computational processes (many representations' interactions) and the computational are the special case of enactive processes (feedback loops with the environment). First you enact in the world and train your computational side of the brain, as things gets abstracted from these enactments, plans arise. Your motivational system always helps you along the way: wanting things that benefit you, imitating, stressing over, exploring, until you learn every skill you see around you and then you get socialized to correct your behavior in a society. Intelligence is the finding every representation of these tasks, whether it is low level feedback loop correction, combination of "stuff" and planning over "stuff". Your perception and action system therefore have the same representation of the world. When you observe stuff you see the functionality first. These things are already settled around 1950. After the discovery of grid cells in 2016 the implementation details become clearer and the systems like HTM (by Jeff Hawkins) are already trying to implement real intelligence.
@killers31337
2 жыл бұрын
Eh? Symbolic AI was a big topic of research in 60-80. John McCarthy made LISP in 50s specifically for symbolic computations and AI. There were some results in natural language understanding, but it could not take free-form questions.
@lucasnoritomi-hartwig3928
2 жыл бұрын
Well I guess first graders are obsolete
@gamers-xh3uc
2 жыл бұрын
and now write can you code gta 6
@obszczymucha1337
2 жыл бұрын
Or better. Codex, can you code yourself?
@jimwarden1121
2 жыл бұрын
You are learning too fast. Don't graduate before me pls! );
Пікірлер: 235