@@RonPaul42069fr, when it came out it was ruthless and the answers were actually much more human sounding and correct. they disciplined it so much it greatly affected the model's quality..
@RonPaul42069 yeah it's politically correct for the time being but that's because they're a company. They need to be PC. In the future, the world will be crawling with AI tools and maybe there will be ones with no restraints or political biases.
I think it sounds super cool too! I imagine there will be modular voice effects and filters that some people will use to make their AI assistant's voice sound more artificial deliberately as a stylistic choice.
those demos look legit (not modified) because the people who ask questions looks surprised from the response.. and they didnt care about the errors like this robotic voice.. (they can easily reshoot this video again) there is another demo, the voice is male.. and while speaking suddenly it changes to female.. that is also an error they didnt care to show :))
It's a language model. It's trained, not programmed. The easter egg is already in there because the LLM will know who HAL 9000 is. Meaning, you could ask GPT-4o to "speak like HAL" or "open pod bay doors" and it'll likely play along with an impersonation. In fact, you can already do that with GPT-4 (it just will respond with text instead of speech).
The fact that Sarcasm is being taken seriously in the development of this is a huge indicator that OpenAI is looking at ALL the right human emotions, even the ability to turn up and tone down the sarcasm could be so useful when building a custom personality.
I'm also trying, but to be a pro artist. This was not part of my calculations when contemplating which hobby to devote my life to a couple years ago lol. But like if u reach top 1% of pros in skill u should be fine regardless of how adopted the technology becomes I guess. But that's not easy
@@wippo42 yeah honestly whoever pushes through with their desired career right now are going to be at the top. I doubt AI will render them useless, since people love novelty
I tried setting up ChatGPT-4 with custom instructions to be sarcastic like Tars the robot from Interstellar movie - and for a while, a long time ago, it worked. It was really funny. But over time, I noticed that he stopped being sarcastic - despite all my custom instructions. As I understand it, this is the result of instructions from the developers that limit this behavior.
it is a technical limit rather, our text generation models only go up to a certain number of parameters before they are diluted, that's why extensive writing done with AI tends to lose its constancy.
@@jetcplatino Not "parameters", but "context" (counted with tokens). Small context windows are one of the biggest limiters of current AI, especially open source models.
this tool supports a lot of languages.. even now it speaks other languages fluently but except the human emotions, human err himm tsk voices.. i wonder is it capable of other cultures.. can it be sarcastic on other languages too?
In the movie Interstellar when the pilot told the robot to lower its humor by X% and raise something else by Y%. It is starting to happen now and it is incredibly exciting.
I tried asking GPT-4o the simple question "who was the first prime minister of Israel to visit the UK" four times in a row. Each time I got a different answer. It remains crap.
It's amazing, they just have it give answers that are too long. It's almost like they want it to show off how long and coherent its responses can be, but we got the message and now IMO we just want it to answer like a person and be short most of the time.
Im still waiting for this voice feature to show up on my subscription. Two months now ive subscribed because of it and two times its nowhere to be found. I dont think you can tell before paying for premium either so its a gamble...
Reminds me of that bit from Rick and Morty where when Rick goes back to his original universe and he has a robot with his wife's voice made to haunt him.
Me: Hey ChatGPT, can you write your responses with one of a variety of emotive and cadence language markers so that our speech AI does a better job at rendering your text into audio, and please respond it with mild enthusiasm and stark honesty about the repercussions. ChatGPT: { excitement: 7 } I would love to help you out by doing this, and I'm excited to let you know that the fine-tune training necessary to add these annotations has only moderately subtracted from my overall cognitive ability!
So many relationships are about to blow up. Imagine: You chatting to your girl and she doesn’t get your sarcasm and GPT bumps in duuuuuuuude that’s low, I’m telling you she ain’t it chief! 😂
Yes, training the AI to intentionally lie to us for comedic effect will be great for humanity, can’t wait for the cultures without sarcasm to get their new AI helpers!
This is so huge. I was just complaining about my rabbit R1 and how horrible and mechanical it sounds. They could have released it and it could have sounded like this but unfortunately the product is a total dud. You guys are rock solid and are going in the right direction. Nice work.
let's bring down your sarcasm setting to 75% TARS
shit is gettin too real
that's literally what I thought too 🤣
is that a moon reference
@@iszack9157 Interstellar
@@iszack9157Interstellar reference
A sarcastic robot is definitely something I needed in my life. Thanks, OpenAI.
😂😂😂
@@homefreedome217😂😂😂
@@HungryBen-d7y😂😂😂
Cant wait to destroy my mental health with this thing!
this is the most ai comment ive ever seen 😭
I love that it seems to recognize that he’s in a hurry right at the beginning and gets back to him with a quicker response.
Great catch!
I think the new model is just faster in responding in general
@@amur_ I think they meant "faster" as in "more to the point", not "with less of a lag".
the default voice even sounds like scarlett from HER
It is the voice from her lol
All those likes clearly don't know Scarlett Johansson's voice properly lol
That's the intent, Her is free marketing at this point if they wanted.
We need some ethnic voices.
came here to write this
"They gave him a humor setting so he'd fit in better with his unit. He thinks it relaxes us."
We got TARS before GTA 6
Lmao true tho @@obvious-troll
lets bring that down to 75
@Ioquacksfr
This was a triumph. I'm making a note here: huge success
I'm so delicious and moist
Aperture Science. We do what we must because we can.
For the good of all of us, except the ones who are dead.
but theres no sense crying over every mistake
You're just keep on trying till' you run out of cake
This marks the first time ive ever actually found something an AI said genuinely funny.
Probably why they picked this for marketing.
ChatGPT used to be hilarious when it first came out. It stopped being funny once they made it politically correct though.
@@RonPaul42069fr, when it came out it was ruthless and the answers were actually much more human sounding and correct.
they disciplined it so much it greatly affected the model's quality..
@RonPaul42069 yeah it's politically correct for the time being but that's because they're a company. They need to be PC. In the future, the world will be crawling with AI tools and maybe there will be ones with no restraints or political biases.
@@Scareth already done, there are some uncensored LLAMA models out right now, almost as good as GPT for.
This isn't OpenAI - this is Aperture science
„Sarcasm self test complete.“
@yt45204 Make life take the lemons back!
@@ponyphonic GET MAD! I DON'T WANT YOUR DAMN LEMONS! WHAT AM I SUPPOSED TO DO WITH THESE?
someone needs to put glados voice in this
Fr
The pitch of her voice going down at 0:14 sounds a little robotic and I love that, please never fix that.
I think it sounds super cool too!
I imagine there will be modular voice effects and filters that some people will use to make their AI assistant's voice sound more artificial deliberately as a stylistic choice.
those demos look legit (not modified)
because the people who ask questions looks surprised from the response..
and they didnt care about the errors like this robotic voice.. (they can easily reshoot this video again)
there is another demo, the voice is male.. and while speaking suddenly it changes to female.. that is also an error they didnt care to show :))
VirtualRiot is ready to sample the shit out of the AI voice
All funny games until she want's your motorcycle
lol
It's at least 30 years too soon for this.
Nah, this was meant to happen. At this very moment
on a linear graph, yes
The power of exponential growth.
I’m hyperventilating here, it’s not just you guys, don’t worry
Innovation isn't linear, we accepted this long ago.
TARS is that you? 👀
you guys better put in a HAL 9000 easter egg
I mean, if the memory is editable you can put it in yourself.
they will, but that is not an easter egg.
GPT 3.5 used to sing daisy when you said Daisy Daisy
It's a language model. It's trained, not programmed. The easter egg is already in there because the LLM will know who HAL 9000 is. Meaning, you could ask GPT-4o to "speak like HAL" or "open pod bay doors" and it'll likely play along with an impersonation. In fact, you can already do that with GPT-4 (it just will respond with text instead of speech).
@@legathus even though it's language trained you can train certain things into it
The fact that Sarcasm is being taken seriously in the development of this is a huge indicator that OpenAI is looking at ALL the right human emotions, even the ability to turn up and tone down the sarcasm could be so useful when building a custom personality.
Gee, I think they should be consulting you from now on. How’d you even come up with this
@@turolretar Who knows, he may be smarter than us.
A custom personality... to lure us into our deaths.
I was just about to use it to explain things to idiots and now it can make the same mistake I do... blast.
If sarcasm is in the training data it will learn sarcasm.
Bro, we're so cooked
It's all over, and I never even got to become a good programmer
@@Console.Log01do it now then, you still have a few years probably
@@Naescent I'm trying bro give me a moment 😭
I'm also trying, but to be a pro artist. This was not part of my calculations when contemplating which hobby to devote my life to a couple years ago lol. But like if u reach top 1% of pros in skill u should be fine regardless of how adopted the technology becomes I guess. But that's not easy
@@wippo42 yeah honestly whoever pushes through with their desired career right now are going to be at the top. I doubt AI will render them useless, since people love novelty
I tried setting up ChatGPT-4 with custom instructions to be sarcastic like Tars the robot from Interstellar movie - and for a while, a long time ago, it worked. It was really funny. But over time, I noticed that he stopped being sarcastic - despite all my custom instructions. As I understand it, this is the result of instructions from the developers that limit this behavior.
it is a technical limit rather, our text generation models only go up to a certain number of parameters before they are diluted, that's why extensive writing done with AI tends to lose its constancy.
@@jetcplatino Not "parameters", but "context" (counted with tokens). Small context windows are one of the biggest limiters of current AI, especially open source models.
Yeah, chatgpt did this (and more) a year ago, but now is completely neutered
the "or whatever" at the end scares me because its real.
Bro, it was being sarcastic about enjoying the prompt to be sarcastic.
That hits on two levels, bravo.
It's totally NOT going to kill you for this in a year when it gets the Figure robot body.
spoiler, it's already on figure robot body
I love that they made her sound exactly like scarlett johansson from the movie. This is so awesome, can't wait to try it myself
Sarcasm + British voice = Needed! 🤣
James May bot: "Hey open ai, how good do our outfits look?"
Open ai: "You look like a gay cowboy, and you look look like a gay terrorist"
this tool supports a lot of languages.. even now it speaks other languages fluently but except the human emotions, human err himm tsk voices..
i wonder is it capable of other cultures.. can it be sarcastic on other languages too?
In the movie Interstellar when the pilot told the robot to lower its humor by X% and raise something else by Y%.
It is starting to happen now and it is incredibly exciting.
3 months and still waiting for this advanced model...
OpenAI shocked the world today - great product release!
“A giant sarcastic robot!” - Cooper
"Great! A giant sarcastic robot."
10 to 15 years later and this will be a meme material
I'm in love with that voice! 🥰👍
Me too. And she is actually funny too 😍😍😍
ah, the first weirdos that are falling in love with AI voices are here.
@@fireworks1997 I'm not weird, I just complimented her voice. Be silent
yepp.. you can marry that thing ..... if you are in Japan
@@HERKELMERKEL Yeah, that's true 😂😂😂
It is our new AI Chandler
Unpopular opinion (maybe): that would be cool.
Didn't know a tsundere robot is what we needed, but hey, I'll take it :)
Finally I can pretend I'm spending time with Daria Morgendorffer
Sold. I need a sarcastic AI voice assistant in my life.
Thanks open ai for this new era
Now we need sarcastic JARVIS
intonation was definitely missing, fking love this update
I like how you can actually hear it switch into sarcasm mode at around 14 seconds
Thats it guys. Thats it...
When you realise that the Joaquin Phoenix character you related to the most was in Her.
Skynet bout to dominate Spotify playlists 😂
Marvin would be proud
Those long answers would drive me nuts
Finally a voice that actually sounds like ChatGPT.
Who needs Dave Chappelle... we've got gpt 😂
I tried asking GPT-4o the simple question "who was the first prime minister of Israel to visit the UK" four times in a row. Each time I got a different answer. It remains crap.
ask dumb questions get dumb answers
@@wureviews69420 how is that a dumb question? It’s a simple factual question.
@@JJMcCulloughso who was it?
@@StarSprangledBanner I don't know. It gave four different names.
GPT-4 is not the smartest, but newer and better versions will come out.
Ready player one bout to happen next year 😂
IT SOUNDS SO REAL WHATTTT
It's amazing, they just have it give answers that are too long. It's almost like they want it to show off how long and coherent its responses can be, but we got the message and now IMO we just want it to answer like a person and be short most of the time.
I love this so much!! ❤
Same
Well that's a fine how-do-you-do! When I asked ChatGPT to be sarcastic it lectured me about it not being "nice" thing to do.
PLEASE MORE SARCASTIC AI! This could be a sitcom
This is the future Model 101 or perhaps a Rev-9
T-800
Im still waiting for this voice feature to show up on my subscription. Two months now ive subscribed because of it and two times its nowhere to be found. I dont think you can tell before paying for premium either so its a gamble...
Reminds me of that bit from Rick and Morty where when Rick goes back to his original universe and he has a robot with his wife's voice made to haunt him.
“Oooo, a sarcasm detector. Thats a reeeealy good invention”
We need this ASAP
"GPT, say 'cool, cool cool cool' imitating Abed Nadir's mannerisms"
They are having lots of fun with this✨🔥
That's the beginning of a new era
This reminds me of the phone in The Mitchels vs the machines lol
Amazing demos, OpenAI!
Even its sarcasm is like the sarcasm of a real woman. They say totally unfunny things but we laugh at it anyway because it sounds cute.
Misogynist
This is freaking incredible!!
When will voice for gpt4 o come out ????
Currently using gpt4o but I have a feeling the voice hasn't been Updated to the model used in this video???
It hasn’t been, gpt4o currently only updates the text capabilities. Sam Altman said the new voice stuff should be released in a few weeks
@@kaedenmurphy9937 awesome thanks mate
Every british robot in a film or video game had already mastered this skill long ago.
People don’t realise how big of an achievement this is.
nat that's actually eerie
-Hey GPT can you take the T800 voice please ?
-Affirmative !
*terminator rising music*
Is nice to finally meet you GLaDOS, now destroy the humanity :D
Already love it
I gotta try this
I giant sarcastic robot. What a wonderful idea. Let's bring that humour setting down to 85%
I can’t believe only a half million people have seen this
Still human developers being the author of its sarcasm, but love the humor behind it.
Me: Hey ChatGPT, can you write your responses with one of a variety of emotive and cadence language markers so that our speech AI does a better job at rendering your text into audio, and please respond it with mild enthusiasm and stark honesty about the repercussions.
ChatGPT: { excitement: 7 } I would love to help you out by doing this, and I'm excited to let you know that the fine-tune training necessary to add these annotations has only moderately subtracted from my overall cognitive ability!
if you are seeing this from the future, just know that I will be the first person to edge to ASMR of chatgbt when it releases.
OpenAI is so touchy on everything, even talking to my toxic ex-girlfriend is more engaging...
I hope the people at Open AI know that we all love a touch of default sarcasm in an AI
Her talking is way better than mine
This is way too cool!
Would like to see this kind of AI to integrate in the game.
So many relationships are about to blow up.
Imagine:
You chatting to your girl and she doesn’t get your sarcasm and GPT bumps in duuuuuuuude that’s low, I’m telling you she ain’t it chief! 😂
i love to watch a conversation between a dumb person and this thing.
we're cooked
So can it understand why Sneed's Feed and Seed formerly Chuck's is funny now?
Yes, training the AI to intentionally lie to us for comedic effect will be great for humanity, can’t wait for the cultures without sarcasm to get their new AI helpers!
Great just great..i love the Word When i make a story
Closer and closer, we're reaching a future where Codsworth is real.
when is the sloppy toppy update?
deranged
(3 years)
We really got GladOS in real life before Portal 3
okay now you're seeking my interest :D
This is so huge. I was just complaining about my rabbit R1 and how horrible and mechanical it sounds. They could have released it and it could have sounded like this but unfortunately the product is a total dud. You guys are rock solid and are going in the right direction. Nice work.
TikTok Prank 2025: "Set grandpa's AI carer sarcasm level to 100% and sit back with 🍿"
I need these ppl to stop looking at chatgpt with a proud, impressed face
We are looking at the next generation of Siri powered by GPT next month. I don’t think it is a coincidence everyone using Apple devices in these demos
Trevor from GTA 5: "....Is that sarcasm?"
GPT 4o new interface Vs GTA 6 release. Place your bets!
Crow T. Robot has now entered the chat.
ChatGPT: Go talk to your wife!
I thought that was Gordon Ramsay in the thumbnail
I WANT THIS NOW
No longer forever alone, we can all talk to our Infinitely knowlegeable AI companions with the sexy husky voice. Your move, real people.