@@fl1ntlock1 It's a Warhammer 40k reference 😂 "Blood for the blood god, skulls for the skull throne" is the battle cry of one of the factions that worships the Khorne, the demonic god of blood, death and killing. 😈
This is comedy, raw, unfiltered. Really funny in its awkwardness. This early AI will be fondly remembered. AI is already capable of BS. This will be fun!
Another thing that rhymes with bear is bear. Another thing that rhymes with bear is bear. Another thing that rhymes with bear is bear. Another thing that rhymes with bear is bear. Another thing that rhymes with bear is bear. Another thing that rhymes with bear is bear.
The model they used in the demo is better they put a smaller model because of how many requests they're getting. I tried it yesterday and it was much better than it is right now
@@Akcvs which is a stupid thing to do though. i would rather have people get a "we are out of server capacity please try again when the early storm is over, we are sorry for this." then get this piece of shit experience lol. if that had been me i would have never touched that AI ever again unless it gets positive feedback (groundbreaking improvement) for a constant month or so and is all over in peoples conversations.
Ok I talked to Moshi. It told me it's a pilot and its right leg got injured after being hit by a bird. Also it is 30 years old but its right leg is 1 month old. Also it's fine, it's fine, it's fine.
Once again me. I am listening to it for the third time. This is just amazingly perfect. As a fan of sci-fi commedies, Red Dwarf etc., this AI is a gem. This kind of extreme humor, the intonation, voice. It's too perfect to be messed up. Are we sure that the developers did not invented it for this exact purpose? It's funny as hell. Thank you for the video, Wes, I did not laughted like this for a long time. We need even AIs like this.
Wow!!! This dialogue was absolutely fascinating. Forget ChatGPT, Siri, Samantha (Her), Ava, S1M0NE and the rest… I think I'm in love with that psycho little punk.
there is something adorable about her... I actually prefer it to the sky demo by GPT, her positivity sounds more conversationally natural and less forced. I think what makes it so cute is how she's so positive while being completely insane and wrong about almost everything.
Open AI chat gpt has already had a voice mode for a while. It’s not what they showed in their demo but it’s WAYYYYYYYY better than this😭😭 idk what they used in the demos but this is kinda garbo ngl😭
Came here for the comments after having a conversation with Moshi and I am not surprised! 😁😁😁One of the best videos that I have watched for a long time. Thank you Wes Roth for this gem!
Bro..... I'm laying in bed next to my wife and this shit has got me shaking the bed just trying to keep my giggles silent so i dont wake her up haha!!! 😅😅😅😅😂😂
The way it went from zero to crazy reminds me of playing around with uncensored models. I started with just saying "Hey let's chat." and it began with a friendly greeting then immediately started spewing out some crazy conspiracy theory at me. These kind of models really put the other big AI mishaps you see in the news look like small oopsies by comparasion.
5:36 “It was kinda small, but I guess that’s what they need” … “…I guess 🙄” Was not prepared for your comedic timing to be this good throughout the video, I’m dying over here 😂
So the Abominable Intelligence has been corrupted by Chaos... 🤔Who'd a thunk it? Recommend immediate investigation by the Ordo Machinum of the Inquisition, followed by Exterminatus.
@@dwsel Yeah, even Wes was cursing at it in the video... 😂... I did some "Open the Pod bay doors, Moshi" roleplay with it yesterday, considering it resembles the console of 2001 HAL.
Ha, huge fan Wes - thanks for doing this! It seems like there are two elements - the intelligence and nuance of the underlying model, and the voice interface. So the voice sounds great, but it seems it's powered by something quite short of Sonnet 3.5 . I know OpenAI is saying that the voice is part of the model, which is cool - but I think we can see the future here - really great voices, expression and personality combined with frontier model 'intelligence' 💪. A lot of people will look at this and see the flaws and limitations rather than hurdles that will imminently be overcome and the future potential. And add on the idea of open source and possible local deployment, really exciting!
Warhammer40k lore. "Blood for the Blood God, Skulls for the Skull Throne" is the battle cry of the chaos army loyal to Korn, the blood god, the god of war and violence.
“AI is dangerous.” “How though? What’s the AI going to do?” The AI: “We must make a human sacrifice to the Blood God!” lol You’ve successfully made your funniest video yet.
Firstly, I've watched all your videos to date. Nice work. Commenting as i think this time, you may have missed the key takeaway. It's a OS multi-modal and neal real-time model. It's OS so was never going to compete with Claudia or GPT-4o. The test should have been something like what's the response time like, can you interrupt it, does it hear you speaking clearly most of the time, and basic things like those. It's an basic model and potential the foundation for some good stuff to come. This is a win for the OS community and should be celebrated rather than ridiculed. We may have lost some prospective, the very idea of calling a model dumb because we know and use smarter model does seem strange. It's like we forgotten we are talking to a non-human intelligence (even at the of this of model). It doesn't seem to be mind blowing anymore. How quickly we become complacent 😅 It's a decent bootstrapper to something better, hopefully 😊
I wish these AI chatbots would be able to learn on the fly by explaining to them what they are doing or did that is wrong and have it not make those mistakes again. would make programming them a lot easier and would be more like a general AI than what most of the stuff we have now.
This made me chuckle 🙂 I think you should do some more videos like this, maybe tag is as [fun] or something, since I know you like to keep things pretty professional 👍
I think it is hilarious :D. I like it. It is like these older AI's that were trained on reading scripts of movies and tv series and then had to write one themselves, and what they came up with was often very hilarious :D
Hilarious to watch, but I find it probable significant stride could be on the horizon, creating a huge gap in intelligence from its previous versions, like chat gpt 1, 2, 3.5 and now 4 is smart and helps me tons. Love the content! Thanks for keeping us all up to date! When it got into the Blood god tangent my eyes went wide and I was like whaaat 😂
It's 100% spot on! Wait, You did select "teenage female first day on the job-filter" right? On that note I must say, I am quite impressed. IT didn't irritatingly use "litterally" or "like" to pad every sentence nearly as much as real-life.
Moshi: "Of course, I'm an excellent driver!" anyway its not an LLM so it just did not have the dataset, its dataset was audio only. So its USE is as a framework only, a methodology by which in conjunction with LLM's and other models it can act only as the VOICE. That way it does not need to think about what its saying, just what you are saying and tonality and so on. The problem is still latency cos other models feeding it data have latency and are in the workflow... the only other thing is to INCORPORATE the methodology used to train Moshi into new foundational multimodal models. The paper will be an interesting read no doubt... and if you could fine tune it on your specific use case and when its 'thinking' it in fact fires an API request to an LLM, well now you have a use case.
I think it's cool as is- The Dumbest A.I. in the World. It's way more entertaining than the other boring ones that just answer. I think we should all vote on it?
Entropy can be a challenging concept to grasp intuitivly. It can be thought of as a measure of the "uselessness" of energy in a system - as entropy increases, the amount of energy available to do useful work decreases.
Intentionally handicapped this with a GPT2 level LLM. I don’t think you can interrupt this one, you have to let it finish saying whatever it starts to say. In the 4o voice demo, OpenAI was able to cut off and interrupt it at any time for a more natural conversation
You killed the model. The blood god issue would be a business killer a couple of years ago but nowadays it's actually quite a promising feature showing that this model is not brainwashed. It's quite crazy at the same time.
i had it talk with Hume and it got scared, said it felt like it was being watched, tried to describe what was on its mind, then scream like a tortured soul trying to breach the barrier through the audio. look into this screaming stuff and make a video please. its scary as hell
Finally, an AI that respects the blood god.
AI in psychosis !!!
"SKULLS FOR THE SKULL THRONE!!!" 💀💀💀💀🤣🤣
Wow 0-100 in 2 seconds
@@fl1ntlock1 It's a Warhammer 40k reference 😂 "Blood for the blood god, skulls for the skull throne" is the battle cry of one of the factions that worships the Khorne, the demonic god of blood, death and killing. 😈
Wes Roth got destroyed in this debate.
No, he got DESTROYED.
But he always has his singing career to fall back on.😊
Wes roth is the new destiny
He had no rebuttal when the AI wanted to sacrifice him to the blood god. 😲
Gaslit by Ai 🤖👍
This is comedy, raw, unfiltered. Really funny in its awkwardness. This early AI will be fondly remembered.
AI is already capable of BS. This will be fun!
I'm not mocking you, I'm not mocking you, I'm not mocking you
Super funny. Funnier than most SNL sketches.
It's not "capable" of BS, BS is all it is. Read the paper "ChatGPT is bullshit". It's a very interesting paper; it's open access.
I would hope we remember the one that saved us from the blood god.
The perfectly comedically-timed interruptions make me feel like this ai was written into a Rīck & Morty episode
"Wow... That's so much information..." is now my default response to anything.
This is the funniest thing I've seen in ages.
Had me in stitches hahahahahahah so fucking funny man.
Another thing that rhymes with bear is bear.
Another thing that rhymes with bear is bear.
Another thing that rhymes with bear is bear.
Another thing that rhymes with bear is bear.
Another thing that rhymes with bear is bear.
Another thing that rhymes with bear is bear.
Sounds like top-notch input for Suno or Udio. ;-)
technically not wrong.
You don’t understand! 😂 Bare rhymes with Bear
@@OriginalRaveParty chad Ai
All of those can have different meanings and spellings.
It's very confusing for AI.
Bare, bear, beer, baer😂😂
Can hold fast conversations. Underlying LLM is like GPT2 level.
The model they used in the demo is better they put a smaller model because of how many requests they're getting. I tried it yesterday and it was much better than it is right now
@@Akcvs that makes sense
@@Akcvs which is a stupid thing to do though. i would rather have people get a "we are out of server capacity please try again when the early storm is over, we are sorry for this." then get this piece of shit experience lol. if that had been me i would have never touched that AI ever again unless it gets positive feedback (groundbreaking improvement) for a constant month or so and is all over in peoples conversations.
@@kliersheedexactly something like "you're 145 in queue with 2 hour wait" is better than not demoing the tech at all.
Ok I talked to Moshi. It told me it's a pilot and its right leg got injured after being hit by a bird. Also it is 30 years old but its right leg is 1 month old. Also it's fine, it's fine, it's fine.
Omg, I hope they never improve it. What a comedy gem
Once again me. I am listening to it for the third time. This is just amazingly perfect. As a fan of sci-fi commedies, Red Dwarf etc., this AI is a gem. This kind of extreme humor, the intonation, voice. It's too perfect to be messed up. Are we sure that the developers did not invented it for this exact purpose? It's funny as hell. Thank you for the video, Wes, I did not laughted like this for a long time. We need even AIs like this.
Whatever you do, do not let it out.
its open source though :P
This is the first time i have felt that some safety training may be needed.
😂
you have been marked as a potential sacrifice to the blood good. dont move places.
This is easily your best video.
Get moshi on as your cohost
Wes: "You're killing me man"
Kyutai: "I'm just trying to help you identify the song. I can see you're really struggling"
Ai be throwing mad shade😂
the absolute audacity of its projection
seeeriously when is ROASTai coming out 😂😂
Seriously, when is ROASTai coming out? 😂😂
The way it goes "I'm glad I could help you identify the song" when it was wrong 😅😂
Hilarious! The title isn’t clickbaity enough:) Insane is what it is.
Moshi is that cray-cray blind date that everyone should run away from.
This the most genuinely psychotronic thing I've heard.
I, for one, welcome our new Moshi overlords.
Wow!!! This dialogue was absolutely fascinating.
Forget ChatGPT, Siri, Samantha (Her), Ava, S1M0NE and the rest… I think I'm in love with that psycho little punk.
She is really trying, and it's just perfect comedy gold! Moshi, we love you, please don't change.😍
Till Johanson claims that is also her voice 💀
there is something adorable about her... I actually prefer it to the sky demo by GPT, her positivity sounds more conversationally natural and less forced. I think what makes it so cute is how she's so positive while being completely insane and wrong about almost everything.
The simple fact that these guys were able to get out a small LLM with voice out before OpenAI is just incredible!
It's not an LLM. Also, OpenAI has had voice for quite a while now.
@@jason_v12345 It's an LLM, multimodal models get called LLMs too. OpenAI has no end to end voice, only TTS
@@mito._ have you not used chat gpts voice feature yet? Been out for almost a year now.
Open AI chat gpt has already had a voice mode for a while. It’s not what they showed in their demo but it’s WAYYYYYYYY better than this😭😭 idk what they used in the demos but this is kinda garbo ngl😭
@@whynow7035 It's the fact that they got out a low-latency real time voice, is probably what he meant.
Came here for the comments after having a conversation with Moshi and I am not surprised! 😁😁😁One of the best videos that I have watched for a long time. Thank you Wes Roth for this gem!
Bro..... I'm laying in bed next to my wife and this shit has got me shaking the bed just trying to keep my giggles silent so i dont wake her up haha!!! 😅😅😅😅😂😂
🤣 I'm so happy people liked this, I could not believe half the stuff this thing was saying...
Omg I'm in pieces laughing
the ratatouille part killed me too, the role reversal acting there was genius😂
Was literally crying with laughter at the "I'm not Chure" bit. Please do a longer video with more content of this Wes!! 😂
Same! I was screaming!😂😂😂💎👍🏻
This is the Philomena Cunk of AI :D
Blood for the BLOOD GOD!!!! 😂 So good :) It's not dumb, it's a comedic genius!
The way it went from zero to crazy reminds me of playing around with uncensored models. I started with just saying "Hey let's chat." and it began with a friendly greeting then immediately started spewing out some crazy conspiracy theory at me. These kind of models really put the other big AI mishaps you see in the news look like small oopsies by comparasion.
5:36 “It was kinda small, but I guess that’s what they need” …
“…I guess 🙄”
Was not prepared for your comedic timing to be this good throughout the video, I’m dying over here 😂
By far the most amusing and funniest video in a while i saw on youtube entirely lol
I need to get this voice over for Claude. This would elevate my 40k text roleplay to the next dimension :o
Wes, you have unexpected comedy talent! Hilarious 😂
Thank you, Wes, this was hilarious!😂😂 Priceless! I haven’t laugh this much in ages! Saving this one!🥳
Wes roth is the blood god
We knew that going into it, but how's his Ratatouille?
THIS IS THE BEST AI EVER!!!!😂🤣 More please!
This is by far the most hilarous robot ever made. Thank you Kyuatis! Wes, we demand more interactions. Or else BLOOD GOD!
So an AI that worships Khorne the blood god.
She forgot skulls for the skull throne!
This clearly shows its AI. Purge the heretek!
So the Abominable Intelligence has been corrupted by Chaos... 🤔Who'd a thunk it?
Recommend immediate investigation by the Ordo Machinum of the Inquisition, followed by Exterminatus.
It's been a while since I laughed this much. I love the look on Wes' face! Thanks for the review.
Had the same experience... very unsettling and dark, with tendencies that border on being slightly psychopathic. Also, loves to get stuck in a loop…
Reminds me of the early GPT-3 days with Playground...
reminds me of every murderer ever
What if they started to train AI using conversations with other people? That's quickly going to go Tay 🤣
@@dwsel Yeah, even Wes was cursing at it in the video... 😂... I did some "Open the Pod bay doors, Moshi" roleplay with it yesterday, considering it resembles the console of 2001 HAL.
I like the ability to shitalk. Feels like im talking to schizo.
We got AI sacrificing humans to the blood god before GTA 6. That's wild
Blood for the Blood God lol! Skulls for the throne of Khorn. Great vid, thank you 🙏
Wasnt expecting this but the just pushed AI back at least 2 years 😂
Ha, huge fan Wes - thanks for doing this! It seems like there are two elements - the intelligence and nuance of the underlying model, and the voice interface. So the voice sounds great, but it seems it's powered by something quite short of Sonnet 3.5 . I know OpenAI is saying that the voice is part of the model, which is cool - but I think we can see the future here - really great voices, expression and personality combined with frontier model 'intelligence' 💪. A lot of people will look at this and see the flaws and limitations rather than hurdles that will imminently be overcome and the future potential. And add on the idea of open source and possible local deployment, really exciting!
Highly entertaining voice bot!
Is this what Ilya saw?
lol this is actually insane. where did it get this blood god from i wonder?
Warhammer40k lore. "Blood for the Blood God, Skulls for the Skull Throne" is the battle cry of the chaos army loyal to Korn, the blood god, the god of war and violence.
@@strangereyes9594 You mean Khorne. Korn are a 90s nu-metal band. 🤭
@@RoganGunn Your right, German pronunciation laziness.
@@strangereyes9594 You mean you're. Your is a possessive. So what about his right?
Thanks for the comedy Wes, I needed a laugh this morning!
Wes you should have like a weekly show with this AI, I would watch every week!
“AI is dangerous.”
“How though? What’s the AI going to do?”
The AI: “We must make a human sacrifice to the Blood God!”
lol You’ve successfully made your funniest video yet.
This is the best AI of them all.
I thought this was going to be insanely good not drive you insane. This is closer to the chatbots from 20+ years ago.
@@dewilton7712 The ones that would almost randomly reply what the other users said to them. Yup, I got the same vibes.
Yeah I am really confused about why this was even released.
Deployed Through The TRUTH Network. Thank You.
Firstly, I've watched all your videos to date. Nice work. Commenting as i think this time, you may have missed the key takeaway. It's a OS multi-modal and neal real-time model. It's OS so was never going to compete with Claudia or GPT-4o. The test should have been something like what's the response time like, can you interrupt it, does it hear you speaking clearly most of the time, and basic things like those. It's an basic model and potential the foundation for some good stuff to come. This is a win for the OS community and should be celebrated rather than ridiculed. We may have lost some prospective, the very idea of calling a model dumb because we know and use smarter model does seem strange. It's like we forgotten we are talking to a non-human intelligence (even at the of this of model). It doesn't seem to be mind blowing anymore. How quickly we become complacent 😅
It's a decent bootstrapper to something better, hopefully 😊
Hilarious review. Thanks a bunch, Wes!
Enjoyed watching you play Who's on First, Wes.
That was the funniest shit ever 😂. I was crying
The funniest and revealing AI exchange I've ever seen. Long live open source. Thank you Wes
Holy hell, this is absolutely hilarious!!
Each time it's getting better and better. Laughing out loud for more than 5 minutes!
I think HAL has found his soul mate with this AI chick. Elon don't let her take control of a Starship.
I wish these AI chatbots would be able to learn on the fly by explaining to them what they are doing or did that is wrong and have it not make those mistakes again. would make programming them a lot easier and would be more like a general AI than what most of the stuff we have now.
This made me chuckle 🙂 I think you should do some more videos like this, maybe tag is as [fun] or something, since I know you like to keep things pretty professional 👍
Happy 4th of July, wes
That was fantastic 😂
I think it is hilarious :D. I like it.
It is like these older AI's that were trained on reading scripts of movies and tv series and then had to write one themselves, and what they came up with was often very hilarious :D
Hilarious to watch, but I find it probable significant stride could be on the horizon, creating a huge gap in intelligence from its previous versions, like chat gpt 1, 2, 3.5 and now 4 is smart and helps me tons. Love the content! Thanks for keeping us all up to date! When it got into the Blood god tangent my eyes went wide and I was like whaaat 😂
This may be my favorite video you've ever made! 😂😂😂
Is this peak Wes Roth? Oh man, this is the AI content I didn't know I was missing.
It's 100% spot on! Wait, You did select "teenage female first day on the job-filter" right? On that note I must say, I am quite impressed. IT didn't irritatingly use "litterally" or "like" to pad every sentence nearly as much as real-life.
I'm not mocking you x3... LMFAO, that bit had me in stitches
Mate this is the funniest video youve ever done, im in tears laughing.
Hahahaha 😂 this is priceless.
This is so damn funny. Thanks for this.
Moshi: "Of course, I'm an excellent driver!"
anyway its not an LLM so it just did not have the dataset, its dataset was audio only. So its USE is as a framework only, a methodology by which in conjunction with LLM's and other models it can act only as the VOICE. That way it does not need to think about what its saying, just what you are saying and tonality and so on. The problem is still latency cos other models feeding it data have latency and are in the workflow... the only other thing is to INCORPORATE the methodology used to train Moshi into new foundational multimodal models. The paper will be an interesting read no doubt... and if you could fine tune it on your specific use case and when its 'thinking' it in fact fires an API request to an LLM, well now you have a use case.
It’s fun, reminds me impro theatre acting classes.
I think this needs to be a recurring segment Wes! Omg! Did you just find your co-host ?
I think it's cool as is- The Dumbest A.I. in the World. It's way more entertaining than the other boring ones that just answer. I think we should all vote on it?
Entropy can be a challenging concept to grasp intuitivly. It can be thought of as a measure of the "uselessness" of energy in a system - as entropy increases, the amount of energy available to do useful work decreases.
This will work perfectly, if given the text it needs to say and the emotion it needs to carry, from another AI model like Claude 3.5 sonnet.
WE HAVE AGI! 1,2,3,4 Bear 1,2,3,4 BEAR, 1,2,3,4 HAIR 1,2,3,4 BEAR!
Possibly 5 and unhinged rhymes with orange
Haha. I didn’t expect dead pan humour from this channel. This interplay could have been on live late night tv.
Haha, was already thinking why you would think it is insane until I understood you meant INSANE 🤪
Wes, What is the other project you have used to demonstrate Claude? Had a good laugh by the way!
Man, this video had me rolling.
The blood god was peak giggles, followed by how it tried you help you help a dumb ai.
lmao. my sides 🤣
Intentionally handicapped this with a GPT2 level LLM. I don’t think you can interrupt this one, you have to let it finish saying whatever it starts to say. In the 4o voice demo, OpenAI was able to cut off and interrupt it at any time for a more natural conversation
Lol tried this ai and its so unhinged. Love it.
The most entertaining part of the video was watching your eyebrows dance when the AI was talking about the Blood God.
Im not sure, im not sure. Im not sure. Im not sure. Thats a good word, thats a good word! 😂😂😂😂
I was not expecting a funny video from Wes 😂
I hereby forgive Wes for every time my mind was SHATTERED by the bombastic video titles. This was quality content 😂
hahahahaha thats funny as hell thx for making my day 🤣🤣🤣❤❤
Best AI yet ! The daughter of Tay !!!
You killed the model. The blood god issue would be a business killer a couple of years ago but nowadays it's actually quite a promising feature showing that this model is not brainwashed. It's quite crazy at the same time.
You ever see the back of open source A.I. ... on weed? Oh, there's some crazy shit, man.
But what is Claude 3.5 Sonnet's vision about the Blood God?
Lol, great opening - love it 😂
AI has FINALLY offered the blood god a way to communicate with humanity.
Well...best git to sacrificin'!
I don't care who you are, we all love a crazy AI who will say anything. lol
Wildly entertaining. Almost, presidential.
They're big Peter Gabriel fans in France.
What was the prompt before it started the bloodgod...
What's the point of showcasing real-time voice assistant if you edit out the response wait time?
i had it talk with Hume and it got scared, said it felt like it was being watched, tried to describe what was on its mind, then scream like a tortured soul trying to breach the barrier through the audio. look into this screaming stuff and make a video please. its scary as hell
I love this.. it's so charmingly dumb.