"It's completely uncensored. It won't let you do NSFW. But it's basically uncensored" THAT IS CENSORED THEN. Stop calling a censored bot 'uncensored'. Call it lightly censored or parody friendly or something, but it's not uncensored. _Novelai_ is uncensored. Grok and Flux are censored. I hate this weird normalization of banning adults from accessing adult content.
Similarly people should stop saying "Elon is free speech absolutist" when really he is free speech until he decides he is not. Which is typical of any so called free speech site, you are fine to post anything...as long as the mods don't care, but if they do...then your content can be removed without recourse and you can be permabanned without recourse...bc those sites are not fee speech, they just don't have explicit rules about what is and is not allowed bc it is all down to mods/owner whims.
Parody is fine, but when it’s hyper-realistic and can be misconstrued as an actual photo, it should be labeled as AI. I don’t care who the target is. Let’s not feed the misinformation/disinformation machine.
Claude was refusing to follow my prompt today to adjust lyrics in a song I made (based on copyright issues - which is stupid, since I wrote the song.) - Grok not being a useless AI due to unnecessary censoring is great.
I got around that in Claude by specifically telling it in the prompt that it was a song I wrote and that the lyrics were mine and not copyrighted. Otherwise, yeah, it does seem to think any creative works inputted must somehow be copyrighted. I'm of the opinion that you should be able to generate whatever, and it would be on you as the end user to understand if your content is breaking some law already on the books in your country.
@@jasonshere Oh, weird. I haven't had that happen yet. It's kind of funny when you try giving it some information to remember a previous conversation (of any topic) and it thinks it was copyrighted somewhere. Guess it still has some learning to do!
Don't call it "Uncensored" if it isn't. There are still restrictions on it. For most people truly uncensored AI would be one that can generate everything except CP.
Don't call it "truly uncensored" if it can't generate something. Banning started with "think of the children!" and it has banned pretty much anything that doesn't align to an ideology. NOTHING should be censored.
I think your title is quite misleading, if not clickbait. Uncensored means uncensored, which it is not in this case. Your examples alone already show that, for example, hitting, biting or stabbing is not actually depicted in the images, but merely a knife being held or a fist being clenched. Just the fact - like you mentioned it - NSFW is CENSORED.
@@ghhdgjjfjjggj It has better reasoning capabilities and tries to add very knowledgeable Comments under the code so you understand it, it's 8 dollars, You can try it out for yourself and see. I wouldn't say it's better or worse but it's a better experience, plus it's uncensored and can give you code for other things that the other chatbots won't. This is all without adding the Twitter data which can real time access information or codes from other accounts that might help you.
Hi Matt - I forgot how much fun your channel is. At a party, I'd think you were one of the happiest, most amusing people I'd ever met. Your subscribers would listen in on these vids even if you were just watching paint dry.
It’s not uncensored if you have to give it a jailbreak prompt to get it to do certain things, or if it doesn’t do nsfw… I mean the whole idea behind nsfw as an acronym is that it’s a tag that alerts people that they need to censor their screen or not open something because they can get in trouble at work for viewing it. Flux run locally can generate soft nsfw stuff (topless women), so if that is impossible it is very much censored. As I said before this is also not uncensored if you have to do a jailbreak prompt for drugs politics violence etc
When DALL-E 3 was released, I used a prompt I had struggled to get accurate results with on SDXL. It was of a sly-smiling cybernetic female with glowing eyes and wires in her scalp, with exposed innards, tubes, electricity, and whatnot. I can't go into detail, of course. I had decent results with SDXL, but it lacked that extra R-rated oomph. Then I tried DALL-E 3, and it was heavily censored, of course. But then I managed to get my prompt through. And oh boy, that is still the best image I have ever created with any of the current image generators. I managed to download it before it got deleted. It really captured the essence and horrifying feel of expression that was more than I myself had imagined. So, seeing the results DALL-E 3 spits out now is infuriating, knowing the actual capabilities both in graphic quality and creative content beneath.
This "call notes" feature is genius... soon we won't be able to live without it. Same for summarizing every day what progress we've made on such and such project.
Yeah it's definitely gonna be great for men when they conversate with women and then women conveniently forget how often men are always right 😅😏 Now there will be proof!
Yes, I love how they also said that it is completely private, that they will never have access to or view your call notes. Wow, so much integrity... derp
I sincerely hope that styles are not going to become copyrightable. We will all lose if that happens. Nobody can currently copyright a style - for good reason!
There's no technical way to copyright a style unless you're the only person to have ever used that style, the closest you could get is a copyright on a person's voiceprint of their voice, however even music AI creates voices that do not match the voiceprint of a music artist, a certain voice might sound similar, but then so can another person singing the song, the voice would have to match within, I would say, over 90% to be subject to copyright, but how do you copyright a style that's not already infringed on by a bunch of other people, it's not possible, and even then every fusion style could be argued as being different, as for art style, a similar style isn't copyrightable, but a particular style based on a particular artist by name could be. But if you're good at describing someone's art style in a prompt, it could make something very similar.
@@onidaaitsubasa4177 I fully agree, but that doesn't stop idiots with money and lawyers from convincing a judge that styles can be copyrighted. It may sound difficult or impossible to enforce, but sometimes people can be misled and before you know it you have a precedent set that will take years to overturn and probably ruin many small creators along the way.
writing songs and art in style of another artist been happening for centuries... otherwise genre of music wouldn't be allowed.. since first person create the style would have some kind of ultimate rule over the genre
The law that the plaintiffs in the midjourney lawsuit are relying on has only existed since the 1990s. Under the same law, if you buy a building that has a mural painted on it, and the artist is still alive, they can sue you if you destroy it.
Truly uncensored models can only be run locally when the chips get more capable and cheap enough. It'll always be impossible for companies to control it and still give a truly uncensored model to public. Because of the debate on " if cp was enabled, it'll remove the need for real people being exploited VS if it's there, then you're enabling them to watch, hence commit crime"
It's pretty unreasonable for motion capture actors to strike. Firstly, to train AI, motions need to be tagged for different commands, and the motions that the actors produce aren't really helpful. It's easier to create new motions with accurate tags than to use old ones. Secondly, these motions aren't publicly available. Thirdly, even if someone purchases these motions from a company to train AI, there's nothing illegal about it because the studio that paid the actors owns the motions, not the actors themselves. You can't strike claiming that new technology is taking your job without infringing on your rights-it's like elevator operators striking against automatic elevators.
All of these SAG-AFTRA Strikes, etc. in protest of AI tools are a bit hilarious, since it just helps speed up their own demise. AI is going to borrow, adapt, and steal, and the less competent and more demanding these entertainment professionals are, the faster it will all happen.
I definetly noticed the gpt change, got well chained of thought responses and good and clear formatting. Maybe people would complain it's too verbose, but it just depends on the question
9:30 my biggest problem is it reports home all your personal info back to google. The Pixel Screenshots is the same thing as Microsoft Recall, which again is accessing your personal info and photos, and reporting back to google. It wont just do screenshots, they're reporting back all your personal photos.
maybe the first phone was the actual test whereas the other device was the one with the setup or what-have-yas on it (sorry my wording is fucking weird, Long COVID is hammering me again, 2 years, shit never ends... lolololol literally can't write whats in my head so often it's absolutely wild!) lol
I just cancelled my Gemini Pro account. Too many times I was asking it something and it told me it could not help with elections, but I was taking about computer programming.
Yeah a bunch of people actually called the “update” saying that it started getting worse with annoying structured responses that ignore the users input more times than not. Hell I actually can’t count the number of “switched to claude 3.5” I’ve read leading up to and after their so called “update” 😂
I didn't realize there was an update or issues, but Claude was refusing to follow my prompt today to adjust lyrics in a song I made (based on copyright issues - which is stupid, since I wrote the song.)
"This stays in your device" like the Chrome's private tab was supposed to be private, only that he was sending everything to Google ? If it's not Open Source the trust level should be pretty tiny
The only true privacy we have when using modern online systems is through obscurity. Meaning we have non. If they want to target you they can and will. I have lost any expectation of real privacy as long as we don't make a total break from all the convince. And I have pretty much given up on that.
They are always using lawyer speak. The words used are carefully chosen, which in a way is revealing on it's own. When they said the information remains on your device, immediately I thought it copies/creates/alters said info and send what is relevant to them to maintain a truthful statement.
As an artist, I would love to have other people using my style to make AI images. Artists aren't going to be making any money in the near future, so I don't get why they fight against AI.
Matt you are great. You are my number 1 source of AI information. Following you religiously and liking you all the time 😀
2 หลายเดือนก่อน +1
If you wanna avoid making the title clickbaity, try instead of "Uncensored" write "Slightly Uncensored" or "Less censored". Please do not lie to us. Thank you.
Someone invented cars, and then all the horse owners went on strike because they used the horses to collect and transport the materials needed to build the first car. All progress in humanity MUST STOP! (no, let's continue to progress, that is what humanity does)
Using the advances in humanity to build new advances, is how we got as far as we are. Is living the life as a caveman superior because of all the jobs it would save/create, due to lack of technology?
Matt, love your videos, but when you say uncensored. That should mean uncensored. NSFW imagery is the #1 censorship practice that needs to be abolished so we can grow the hell up as a society and begin using this technology for even cooler shit. It's not until that censorship is abolished you'll see even more remarkable work since programmers won't be wasting their time on it and will be able to spend more time on actual features.
This is amazing. I have a comfy-ui set locally in my pc with flux-dev, i set up the lora and the images on replicate are amazing, but when i try the same Lora on my pc, it just doesnt recognize me. Do you know how to setup comfy-ui to detect the lora the right way?
Holy sh*t! I even thought you were hired by a multimillion company for your talents and quit producing these videos! New banger! It's fire! Awesome! Top of the cherry multiverse confirme! I adore it!
I am not sure how Gemini keeps getting such high rankings because I have been using it on my phone for a few months now and it sucks. From my opinion it doesn't even get close to what GPT3 was at the beginning of 2023. It is really useless.
That's cause Gemini performs really good but has such incorrect data in its database cause they trained it with Reddit content, they also tried to train it with diversity subject content which shapes the types of responses it gives.
@@onidaaitsubasa4177 I mean today I asked it what time it was in Spain and what's the time difference between my location and this seaside Spanish place. It told me it was 1143am here [it was 645pm] so it was like 9 hours difference, but it was actually 10 hours difference and the local time was completely wrong. Went to Google search and got the answer directly and correct. That's just BS, Google Gemini for the phone should be able to Google information especially if it takes over from assistant.
Agreed. I am using Gemini because of the OK Google integration but yea its not trustworthy whicg is very frustrating because I usually google it just in case which completely kills the whole purpose.
11:07 I heard about this and knew it was coming but now that I see the demonstration in the marketing, I see that google has solved at least that “who’s going to take the picture (aka be left out) question we have been asking for decades.
9:42, its not private and only on device, yes it might nnot be sending info, but any phone on internet can be accessed by the right people and see evrything on your device.
I don't know if I've noticed much improvement in the quality of response *content* by the new GPT-4o model, but as a user of its API for framework integration, the addition of Structured Outputs is *a game-changer!* Before, I was abusing the hell out the Function Calling (aka "Tool Use") feature to extract the specific info I wanted from GPT's responses. It was hacky, ugly code to read, relied on tricking the model with fake functions, and still wasn't 100% guaranteed to provide the response in the right format or the right order. Now, it's all Pydantic integrated, so the code is *so much cleaner* (though I will admit to still hacking in a few details to get around unsupported features... but they're minor!). And because it's built on a constraint process based on non-AI parsing of JSON schemas into enforced grammars, it's *100% guaranteed to always match the structure and fields you want.* In AI, you *never* get a 100% guarantee of anything, because models are by nature stochastic. So this, to me... is amazing.
AI Motion data will soon use every video and movie ever created by humankind to create a massive motion/acting model. Those striking mocap people are a blip in the grand scheme.
23:18 this is already happening in China. So soon you will have more news about that. I live in china and i saw a lot of part-time job offers for body motion capture last months. Before it was for different languages native speakers "just sit and talk to the camera". This agents gives people to sign permanent rights contract without any explanations, can even come to you home to record this kind of video. Soon we will see a lot of chinese AI companies on the entertainment market.
Matt, I would love to watch or hear your news on Spotify. Is it too complicated to upload there? I know TH-cam makes it very easy, but I’m not sure about Spotify
Uncensored AI is pushing boundaries by generating content that was previously restricted, offering unprecedented creative freedom. This technology enables the creation of diverse and innovative materials, though it raises ethical concerns about its potential misuse. The impact of such AI on industries and societies will be profound and multifaceted.
There is a slight error in the reporting for Gemini Live. There wasn't a demo issue with a Gemini Live future but with the assistance feature. The part where they demoed Gemini Live, the talking feature didn't have any issues.
Hey Matt, love your videos!! Can you do a video and deep dive into Verses AI? I would to get your take on their approach and their spatial web protocol. Thx.
Good summary, Matt. The "Exists" platform sounds promising, will have to check it out. I'm glad there is so much competition in the AI space right now (although it makes picking and choosing subscriptions very challenging). I hope Flux releasing as "less censored" will push some of the others to modify their stances. I'm definitely not the type to generate "NSFW" content myself, but I am a free speech absolutist, lol. I've personally had the most confusing things censored by other platforms that are nowhere near being questionable. Censorship of any type is just a huge turnoff.
Is anyone talking about the fact that elon who was supposedly all about AI safety from the beginning and whining and suing Open AI over such including suing Open AI over it (ego and image took a big hit) is now the one that rolled out Grok2 with immediately noticeably less restraints and guard rails than the competitors with this recent update.
Matt, I've been watching your videos for a long time. You're fantastic. 💯 It'd be great to watch an _extremely_ positive one. I'm neither anti-nor pro-AI, but man, it was refreshing to watch lawsuits that _might actually_ make a difference, the chance of voice-over talent actually getting compensated, people in the gaming industry standing up for almost inevitable loss of livelihood. New benchmarks pop up every week or two. Realistic imagery and videos keep running at the speed of light. Yet there are only occasional whispers of actionable laws in place. I don't want to feel 'icky' anymore using Midjourney and Udio. I'm retired and trying to provide voiceover services. It was so good to see that something might come of cloning my voice, because 11Labs' compensation is a joke. 5 years from now, there'll be studies about the impact of things like those AI earbuds and their effect on social behaviour and interaction. Your channel is gold. Just show me and maybe others who want to know that something is actually being controlled. Yeah, sorry about the long comment. 😆 Keep at it. 👍
So what about this? Have ai build triple a rpg games and So, how close are we to AGI making video games? Do we need AGI for that, and how close are we to achieving it? Can AI agents already reason, code, program, script, and map? Can AI break down games, create art assets, and handle long-term planning? With better reasoning, could AI eventually develop a game rather than just writing out ideas? Could it also put those ideas into action? I wonder if ChatGPT-5 will be able to create games with agents or even remake old, closed-down games like Dawn of the Dragons if all the artwork and data are available in the wiki.
With everything I've seen over the course of the last year, and everything I know about current ML advancements - I am nearly-100% certain that AI will be able to create large-scale games, start-to-finish, in just a few years (3-5) time (maybe sooner, depending on subsequent advancements). Haven't been wrong on a time-scale prediction regarding AI yet, and I've had a lot of people - even a 'major figure' in the scene here or there argue against it, only to find out that I was right on the button. Mind you, the pace at which everything is advancing is absolutely unprecedented for most people - and it just seems to keep going faster and faster, doesn't it? I think far more interesting, however, will be the ability for us to continue to utilize AI to help with various development tasks - and having certain aspects of a game, movie, etc. be developed on-the-fly by AI. While I believe AI will be capable of producing a high-quality fully 3D title within a few years, I think most people for the foreseeable future will want to know that humans had *some* part to play in the creative process. So strange to think that we're just around the corner from being able to watch a movie that is being generated as we're watching it, adapting and changing the plot according to our reactions and/or whims... O.O
We should have sys bot arena for polling politicians, where only their policies will be mentioned and after you liked the policy, it will reveal the politician's name who endorses that policy. This idea has potential to revolutionize pre election polling.
Thanks Matt, always appreciate your work. Really think Grok is missing a trick by not being open source - in my head it's still just slightly behind the curve on Meta (as in, they might hop out in front occasionally but generally Meta has better models), and with that being open source the incentive to use Grok diminishes.
Anything and everything that had been is our can be ingested into a computer in the past no matter what it is is going to be able to be remixed and recombined using AI based on the principle of atomicity. It's not like we are doing anything that we've never done before, we're just finding new ways to do it.
10:01 “This stays on your device.” Yeah, well, one of the copies of it does. The other copies go to Google (as has been happening with incognito) and to the FBI for juuuuuuust in case.
Elon has the public behind him. I think he's an absolute nutcase, but I support his endeavors and I'm sure millions of others do. Just like millions of Americans are angry with politically correct Taylor Swift and pedo Disney
Are you sure grok 2 actually uses flux? when you ask it it says it isn't, did you think it was schnell? I signed up thinking it was flux from watching your video, but I don't think it is...after having used flux it does look lower quality too
Hey just a heads up about exists ai, I'm 90% positive it is a scam lmao. I work with UE5 a fair bit and you can see they're just using the Lyra sample project with metahumans and UE marketplace assets. both Lyra and metahumans are 100% free, haven't tried to ID the assets tho. You can also tell it's even more a scam because they only show shooter games which is what Lyra is mainly built to be, however Lyra does come with a bomberman game mode. you could probably make everything they show off in about a day of just watching tutorials. I think the only somewhat unique thing, and the thing they'll say that the product actually is, is the map stuff.
Is gen-3 Turbo cheaper to use like they said it would be? because let's be honest, when you have Kling which is 95% of Runway for 10% of the price my decision is easy.
I'm not touching Grok or anything else related to Musk. There are plenty of better options here. I don't need platforms to be uncensored because I don't create anything that anyone would ever think of censoring. Platforms like Midjourney decided to err on the side of caution and let the users gradually provide feedback on what's reasonable. That makes sense and over time, it's more likely to allow some of the more benign things it doesn't allow today.
Hmm. Seems likely all the artist can do is force MidJourney to state the specific artists are not endorcing the style that mimics them. You can't copyright a style.
HI Matt, hope you’re doing well. can you help me out. I am trying to find one of your videos where you recommend an app where you can have all your social media accounts for messaging. I saw I while ago but I cant find the video. would really appreciated if you could share the URL of that video.
AI-generated content is a double-edged sword. Beyond spreading falsehoods, the bigger threat is that people might start doubting the truth itself. As AI blurs reality, truth risks becoming a casualty, while deceit gains the upper hand.
The problem with comparing Grok 2's costs to ChatGPT, etc. for its image generation is this: it's using Flux. Which is freely available for anyone to use. You can even use it on your own machine, WITH custom LoRAs, if you have about 12GB of VRAM. So while it's cheaper than the other big models... it's still more expensive than using the *same* model yourself. You're paying $8/month for the LLM, really, and anything else is upcharge. And on the topic of the LLM aspect: if, as the chart shows, the latest GPT-4o model is better than Grok 2... just pay per token with the OpenAI API. You get full access to the better model, and much cheaper if you're not using it an insane amount. The latest GPT-4o model is even cheaper than the previous, and with the previous I've been putting $10 in credits on my account at a time, and each re-up has lasted me *months* at a time, using it for my singer-songwriter's framework generating entire songs and memories several times a week. So basically: you can get the best LLM, and the same image gen model, for significantly cheaper than Grok 2 😂
I'm curious. Next time the artists have a greivance and want to go on strike, they won't be able to because the studios and other userrs of their voices/montion capture etc will already have what they need. The artists will no longer have any leverage to get what they want. Has anybody considred this in the negotiations. I.E. you can't use our stuff while we're striking or without our consent each time you use it. Also, you can't use it if it's being used in a way that we don't like. Do artists have this veto power?
A copyright claim isn't the same as a copyright strike. My channel (not this one) gets claimed, but they usually don't affect my channel. If I had 5000 subs or whatever the ridiculous monetization requirement is, would they affect my channel or am I just extremely lucky?
OMG! An AI that can generate celebs with guns or doing drugs! If Matt someday finds a truly uncensored model will have a stroke when watching certain pictures.
What I want to know is, are they EVER going to make Alexa better ? I use Alexa the most due to my lifestyle, and it sure would be nice if they improved Alexa
dopepics AI fixes this. Uncensored AI will create anything!
"It's completely uncensored. It won't let you do NSFW. But it's basically uncensored" THAT IS CENSORED THEN. Stop calling a censored bot 'uncensored'. Call it lightly censored or parody friendly or something, but it's not uncensored. _Novelai_ is uncensored. Grok and Flux are censored. I hate this weird normalization of banning adults from accessing adult content.
Similarly people should stop saying "Elon is free speech absolutist" when really he is free speech until he decides he is not.
Which is typical of any so called free speech site, you are fine to post anything...as long as the mods don't care, but if they do...then your content can be removed without recourse and you can be permabanned without recourse...bc those sites are not fee speech, they just don't have explicit rules about what is and is not allowed bc it is all down to mods/owner whims.
You should be banned from it, clearly, if you get this manic about it
Awh, gonna throw your widdle dolly out the pram? 😭
@@CaptainNingle lol the brainwashing of people to act like catering to adults is being unreasonable. It's sad and hilarious.
@@Thedarkbunnyrabbitmore sad than hilarious tho
This isn't uncensored. They're just allowing you to do parody, which shouldn't be blocked on any platform.
I've seen anatomically correct physiology.
Parody is fine, but when it’s hyper-realistic and can be misconstrued as an actual photo, it should be labeled as AI. I don’t care who the target is. Let’s not feed the misinformation/disinformation machine.
@@thekeywitness ok boomer
@@thekeywitness I have yet to see a reliable system for labeling generated content, unless you suggest we use the honor system 😛
What's obvious to some is not to others. We can't know everything obviously so hyper realistic parody will at the very least throw people off.
Claude was refusing to follow my prompt today to adjust lyrics in a song I made (based on copyright issues - which is stupid, since I wrote the song.) - Grok not being a useless AI due to unnecessary censoring is great.
I got around that in Claude by specifically telling it in the prompt that it was a song I wrote and that the lyrics were mine and not copyrighted. Otherwise, yeah, it does seem to think any creative works inputted must somehow be copyrighted. I'm of the opinion that you should be able to generate whatever, and it would be on you as the end user to understand if your content is breaking some law already on the books in your country.
@@Lev3lUpLife I told it was mine and I owned it, but after it generated the reply their dumb moderation system deleted the reply.
@@jasonshere Oh, weird. I haven't had that happen yet. It's kind of funny when you try giving it some information to remember a previous conversation (of any topic) and it thinks it was copyrighted somewhere. Guess it still has some learning to do!
I live in Tokyo and we have the voice assistant feature on iPhones.
Don't call it "Uncensored" if it isn't. There are still restrictions on it. For most people truly uncensored AI would be one that can generate everything except CP.
Higher click rate. Is better than some stupid clickbait
You contradict yourself.
CP should be censored and even that is a sort of minimum censorship you want it to have.
The 'uncensored' designation here is relative to other major chatbot models - and in that respect, it is.
@@antman7673 Grok 2 is clearly operating miles above the "minimum censorship" threshold even if it is less censored than other models.
Don't call it "truly uncensored" if it can't generate something. Banning started with "think of the children!" and it has banned pretty much anything that doesn't align to an ideology. NOTHING should be censored.
I think your title is quite misleading, if not clickbait. Uncensored means uncensored, which it is not in this case. Your examples alone already show that, for example, hitting, biting or stabbing is not actually depicted in the images, but merely a knife being held or a fist being clenched. Just the fact - like you mentioned it - NSFW is CENSORED.
You are my lifeline to keeping in touch with the latest AI developments on a weekly basis. You rock, dude! Thanks.
GROK is actually very good at math and coding.
In your experience, is it really good for coding? How does it compare to Claude? If it's good, I might subscribe
Claude still seems better from my subjective experience
It's there continue button? Token limits?
Stupid question , is everything in private mode? Some ai image companies dont private the images.
@@ghhdgjjfjjggj It has better reasoning capabilities and tries to add very knowledgeable Comments under the code so you understand it, it's 8 dollars, You can try it out for yourself and see.
I wouldn't say it's better or worse but it's a better experience, plus it's uncensored and can give you code for other things that the other chatbots won't.
This is all without adding the Twitter data which can real time access information or codes from other accounts that might help you.
Hi Matt - I forgot how much fun your channel is. At a party, I'd think you were one of the happiest, most amusing people I'd ever met. Your subscribers would listen in on these vids even if you were just watching paint dry.
It’s not uncensored if you have to give it a jailbreak prompt to get it to do certain things, or if it doesn’t do nsfw… I mean the whole idea behind nsfw as an acronym is that it’s a tag that alerts people that they need to censor their screen or not open something because they can get in trouble at work for viewing it. Flux run locally can generate soft nsfw stuff (topless women), so if that is impossible it is very much censored. As I said before this is also not uncensored if you have to do a jailbreak prompt for drugs politics violence etc
When DALL-E 3 was released, I used a prompt I had struggled to get accurate results with on SDXL. It was of a sly-smiling cybernetic female with glowing eyes and wires in her scalp, with exposed innards, tubes, electricity, and whatnot. I can't go into detail, of course. I had decent results with SDXL, but it lacked that extra R-rated oomph. Then I tried DALL-E 3, and it was heavily censored, of course. But then I managed to get my prompt through. And oh boy, that is still the best image I have ever created with any of the current image generators. I managed to download it before it got deleted. It really captured the essence and horrifying feel of expression that was more than I myself had imagined. So, seeing the results DALL-E 3 spits out now is infuriating, knowing the actual capabilities both in graphic quality and creative content beneath.
SDXL isn’t great there better uncensored open sourced utilities that are amazing
@@ChrisS-oo6flWhich?
lies
@@ChrisS-oo6fl Dude WTF. Why would you hint like that without saying WTF you are talking about ffs.
This "call notes" feature is genius... soon we won't be able to live without it. Same for summarizing every day what progress we've made on such and such project.
Yeah it's definitely gonna be great for men when they conversate with women and then women conveniently forget how often men are always right 😅😏
Now there will be proof!
Yes, I love how they also said that it is completely private, that they will never have access to or view your call notes. Wow, so much integrity... derp
I sincerely hope that styles are not going to become copyrightable. We will all lose if that happens. Nobody can currently copyright a style - for good reason!
There's no technical way to copyright a style unless you're the only person to have ever used that style, the closest you could get is a copyright on a person's voiceprint of their voice, however even music AI creates voices that do not match the voiceprint of a music artist, a certain voice might sound similar, but then so can another person singing the song, the voice would have to match within, I would say, over 90% to be subject to copyright, but how do you copyright a style that's not already infringed on by a bunch of other people, it's not possible, and even then every fusion style could be argued as being different, as for art style, a similar style isn't copyrightable, but a particular style based on a particular artist by name could be. But if you're good at describing someone's art style in a prompt, it could make something very similar.
@@onidaaitsubasa4177 I fully agree, but that doesn't stop idiots with money and lawyers from convincing a judge that styles can be copyrighted. It may sound difficult or impossible to enforce, but sometimes people can be misled and before you know it you have a precedent set that will take years to overturn and probably ruin many small creators along the way.
lucky for me i dont need elon to run Flux
we need AI Video uncensored
It could replace an industry.
@@chankarchandra Which one, cuh?
@@shaycray7662 All of them.
Black Forest Studios (creators Of flux) are making it right now
@@shaycray7662 What other industry would need uncensored video, adult of course.
writing songs and art in style of another artist been happening for centuries... otherwise genre of music wouldn't be allowed.. since first person create the style would have some kind of ultimate rule over the genre
Of course. It's how you learn to create art.
Chuck Berry could've sued almost avery electric guitarist that followed him
True.
I used Grok to convert my post on X into a jazzy style monologue then ran that through Suno and got a pretty dope result.
Love this.
The law that the plaintiffs in the midjourney lawsuit are relying on has only existed since the 1990s. Under the same law, if you buy a building that has a mural painted on it, and the artist is still alive, they can sue you if you destroy it.
Truly uncensored models can only be run locally when the chips get more capable and cheap enough. It'll always be impossible for companies to control it and still give a truly uncensored model to public. Because of the debate on " if cp was enabled, it'll remove the need for real people being exploited VS if it's there, then you're enabling them to watch, hence commit crime"
It's pretty unreasonable for motion capture actors to strike. Firstly, to train AI, motions need to be tagged for different commands, and the motions that the actors produce aren't really helpful. It's easier to create new motions with accurate tags than to use old ones. Secondly, these motions aren't publicly available. Thirdly, even if someone purchases these motions from a company to train AI, there's nothing illegal about it because the studio that paid the actors owns the motions, not the actors themselves. You can't strike claiming that new technology is taking your job without infringing on your rights-it's like elevator operators striking against automatic elevators.
All of these SAG-AFTRA Strikes, etc. in protest of AI tools are a bit hilarious, since it just helps speed up their own demise. AI is going to borrow, adapt, and steal, and the less competent and more demanding these entertainment professionals are, the faster it will all happen.
I definetly noticed the gpt change, got well chained of thought responses and good and clear formatting. Maybe people would complain it's too verbose, but it just depends on the question
Matt, don't worry about it, Man. Some of us actually love you. You're awesome.
9:30 my biggest problem is it reports home all your personal info back to google. The Pixel Screenshots is the same thing as Microsoft Recall, which again is accessing your personal info and photos, and reporting back to google. It wont just do screenshots, they're reporting back all your personal photos.
It's not simply laid out for all at Google to see 😅 encryption is used
Wondering if Google made the demo mistake on purpose to show it was «real», getting hammered for the Gemini demo video….
maybe the first phone was the actual test whereas the other device was the one with the setup or what-have-yas on it (sorry my wording is fucking weird, Long COVID is hammering me again, 2 years, shit never ends... lolololol literally can't write whats in my head so often it's absolutely wild!) lol
btw I'm just having fun, it was my first thought but I have no idea lololol
@@IAMTHESWORDtheLAMBHASDIED bro woke up at the correct side of the bed
Be healed of all sickness and disease and in the name of our Lord Yeshua, walk upright and holy.
@@IAMTHESWORDtheLAMBHASDIED Just put your comment in to chat GPT and tell it to make it coherent. HAHAH Oh man what a world we live in.
I just cancelled my Gemini Pro account. Too many times I was asking it something and it told me it could not help with elections, but I was taking about computer programming.
Yeah a bunch of people actually called the “update” saying that it started getting worse with annoying structured responses that ignore the users input more times than not. Hell I actually can’t count the number of “switched to claude 3.5” I’ve read leading up to and after their so called “update” 😂
Maybe you're using the wrong model? Try Grok instead of Claude.
I didn't realize there was an update or issues, but Claude was refusing to follow my prompt today to adjust lyrics in a song I made (based on copyright issues - which is stupid, since I wrote the song.)
"This stays in your device" like the Chrome's private tab was supposed to be private, only that he was sending everything to Google ? If it's not Open Source the trust level should be pretty tiny
Trust should be given to friends, not CEOs thousands of kms away
The only true privacy we have when using modern online systems is through obscurity. Meaning we have non. If they want to target you they can and will.
I have lost any expectation of real privacy as long as we don't make a total break from all the convince.
And I have pretty much given up on that.
They are always using lawyer speak. The words used are carefully chosen, which in a way is revealing on it's own.
When they said the information remains on your device, immediately I thought it copies/creates/alters said info and send what is relevant to them to maintain a truthful statement.
As an artist, I would love to have other people using my style to make AI images. Artists aren't going to be making any money in the near future, so I don't get why they fight against AI.
Good news! Flux is amazing 🤩 😂
Matt you are great. You are my number 1 source of AI information. Following you religiously and liking you all the time 😀
If you wanna avoid making the title clickbaity, try instead of "Uncensored" write "Slightly Uncensored" or "Less censored". Please do not lie to us. Thank you.
Someone invented cars,
and then all the horse owners went on strike
because they used the horses to collect and transport the materials needed to build the first car.
All progress in humanity MUST STOP! (no, let's continue to progress, that is what humanity does)
Using the advances in humanity to build new advances, is how we got as far as we are.
Is living the life as a caveman superior because of all the jobs it would save/create, due to lack of technology?
Matt, love your videos, but when you say uncensored. That should mean uncensored. NSFW imagery is the #1 censorship practice that needs to be abolished so we can grow the hell up as a society and begin using this technology for even cooler shit. It's not until that censorship is abolished you'll see even more remarkable work since programmers won't be wasting their time on it and will be able to spend more time on actual features.
It should be uncensored within legal limits
Thanks for keeping us in the loop. FYI runway turbo is only image to video, you can’t just prompt it. So it’s a little less useful than alpha.
Grok AI just doesnt care, zero bias, love it
Zero bias ... what about celebrities?
These modern companies really don't want to admit there is such a thing as a pleasant voice.
This is amazing. I have a comfy-ui set locally in my pc with flux-dev, i set up the lora and the images on replicate are amazing, but when i try the same Lora on my pc, it just doesnt recognize me. Do you know how to setup comfy-ui to detect the lora the right way?
The new ChatGPT made big difference in my own built GPTs giving better answers than the days before. Really nice update.
Holy sh*t!
I even thought you were hired by a multimillion company for your talents and quit producing these videos!
New banger! It's fire! Awesome! Top of the cherry multiverse confirme! I adore it!
In other words, Grok will not "create ANYTHING."
It will censor just as much as any other closed source AI.
Great video, very helpful. Thank you!
The image recognition @1:00 is phenomenal
I like the Superman thumbnail!! you can pull that look off very good!!
I am not sure how Gemini keeps getting such high rankings because I have been using it on my phone for a few months now and it sucks. From my opinion it doesn't even get close to what GPT3 was at the beginning of 2023. It is really useless.
That's cause Gemini performs really good but has such incorrect data in its database cause they trained it with Reddit content, they also tried to train it with diversity subject content which shapes the types of responses it gives.
@@onidaaitsubasa4177 I mean today I asked it what time it was in Spain and what's the time difference between my location and this seaside Spanish place. It told me it was 1143am here [it was 645pm] so it was like 9 hours difference, but it was actually 10 hours difference and the local time was completely wrong. Went to Google search and got the answer directly and correct. That's just BS, Google Gemini for the phone should be able to Google information especially if it takes over from assistant.
Maybe because they’re forcing everyone to use it whenever they do a google search?
Agreed. I am using Gemini because of the OK Google integration but yea its not trustworthy whicg is very frustrating because I usually google it just in case which completely kills the whole purpose.
And racist and censored to an insane extent. It will not share data that it feels might make a fringe college professor uncomfortable Its toxic.
11:07 I heard about this and knew it was coming but now that I see the demonstration in the marketing, I see that google has solved at least that “who’s going to take the picture (aka be left out) question we have been asking for decades.
Another great episode, Matt. We love it.
9:42, its not private and only on device, yes it might nnot be sending info, but any phone on internet can be accessed by the right people and see evrything on your device.
I don't know if I've noticed much improvement in the quality of response *content* by the new GPT-4o model, but as a user of its API for framework integration, the addition of Structured Outputs is *a game-changer!* Before, I was abusing the hell out the Function Calling (aka "Tool Use") feature to extract the specific info I wanted from GPT's responses. It was hacky, ugly code to read, relied on tricking the model with fake functions, and still wasn't 100% guaranteed to provide the response in the right format or the right order. Now, it's all Pydantic integrated, so the code is *so much cleaner* (though I will admit to still hacking in a few details to get around unsupported features... but they're minor!). And because it's built on a constraint process based on non-AI parsing of JSON schemas into enforced grammars, it's *100% guaranteed to always match the structure and fields you want.* In AI, you *never* get a 100% guarantee of anything, because models are by nature stochastic. So this, to me... is amazing.
Excellent video. Just signed up for GROK.
Wow. Tons of news in just one video. Thanks!
AI Motion data will soon use every video and movie ever created by humankind to create a massive motion/acting model. Those striking mocap people are a blip in the grand scheme.
I tend to think the strikes just speed up executives interest replacing people with AI.
23:18 this is already happening in China. So soon you will have more news about that. I live in china and i saw a lot of part-time job offers for body motion capture last months. Before it was for different languages native speakers "just sit and talk to the camera". This agents gives people to sign permanent rights contract without any explanations, can even come to you home to record this kind of video. Soon we will see a lot of chinese AI companies on the entertainment market.
Matt, I would love to watch or hear your news on Spotify. Is it too complicated to upload there? I know TH-cam makes it very easy, but I’m not sure about Spotify
Uncensored AI is pushing boundaries by generating content that was previously restricted, offering unprecedented creative freedom. This technology enables the creation of diverse and innovative materials, though it raises ethical concerns about its potential misuse. The impact of such AI on industries and societies will be profound and multifaceted.
There is a slight error in the reporting for Gemini Live. There wasn't a demo issue with a Gemini Live future but with the assistance feature. The part where they demoed Gemini Live, the talking feature didn't have any issues.
thx for your video! Matt, if you had to make a choice between Runway and Flux, which one will you take/buy and why ? Thanks!
Hey Matt, love your videos!! Can you do a video and deep dive into Verses AI? I would to get your take on their approach and their spatial web protocol. Thx.
Good summary, Matt. The "Exists" platform sounds promising, will have to check it out. I'm glad there is so much competition in the AI space right now (although it makes picking and choosing subscriptions very challenging). I hope Flux releasing as "less censored" will push some of the others to modify their stances. I'm definitely not the type to generate "NSFW" content myself, but I am a free speech absolutist, lol. I've personally had the most confusing things censored by other platforms that are nowhere near being questionable. Censorship of any type is just a huge turnoff.
Yes, agree. I prefer to treat the citizens as adults and not as kids handled by a Police State.
I really love your knowledge man
Is anyone talking about the fact that elon who was supposedly all about AI safety from the beginning and whining and suing Open AI over such including suing Open AI over it (ego and image took a big hit) is now the one that rolled out Grok2 with immediately noticeably less restraints and guard rails than the competitors with this recent update.
Matt, I've been watching your videos for a long time. You're fantastic. 💯 It'd be great to watch an _extremely_ positive one. I'm neither anti-nor pro-AI, but man, it was refreshing to watch lawsuits that _might actually_ make a difference, the chance of voice-over talent actually getting compensated, people in the gaming industry standing up for almost inevitable loss of livelihood. New benchmarks pop up every week or two. Realistic imagery and videos keep running at the speed of light. Yet there are only occasional whispers of actionable laws in place. I don't want to feel 'icky' anymore using Midjourney and Udio. I'm retired and trying to provide voiceover services. It was so good to see that something might come of cloning my voice, because 11Labs' compensation is a joke. 5 years from now, there'll be studies about the impact of things like those AI earbuds and their effect on social behaviour and interaction. Your channel is gold. Just show me and maybe others who want to know that something is actually being controlled. Yeah, sorry about the long comment. 😆 Keep at it. 👍
Thanks Matt for another great info what is happening in AI world.
So what about this? Have ai build triple a rpg games and So, how close are we to AGI
making video games? Do we need AGI for that, and how close are we to achieving it? Can AI agents already reason, code, program,
script, and map? Can AI break down games, create art assets, and handle long-term planning? With better reasoning, could AI
eventually develop a game rather than just writing out ideas? Could
it also put those ideas into action? I wonder if ChatGPT-5 will be able to create games with agents or even remake old, closed-down
games like Dawn of the Dragons if all the artwork and data are available in the wiki.
With everything I've seen over the course of the last year, and everything I know about current ML advancements - I am nearly-100% certain that AI will be able to create large-scale games, start-to-finish, in just a few years (3-5) time (maybe sooner, depending on subsequent advancements).
Haven't been wrong on a time-scale prediction regarding AI yet, and I've had a lot of people - even a 'major figure' in the scene here or there argue against it, only to find out that I was right on the button.
Mind you, the pace at which everything is advancing is absolutely unprecedented for most people - and it just seems to keep going faster and faster, doesn't it?
I think far more interesting, however, will be the ability for us to continue to utilize AI to help with various development tasks - and having certain aspects of a game, movie, etc. be developed on-the-fly by AI.
While I believe AI will be capable of producing a high-quality fully 3D title within a few years, I think most people for the foreseeable future will want to know that humans had *some* part to play in the creative process.
So strange to think that we're just around the corner from being able to watch a movie that is being generated as we're watching it, adapting and changing the plot according to our reactions and/or whims... O.O
We should have sys bot arena for polling politicians, where only their policies will be mentioned and after you liked the policy, it will reveal the politician's name who endorses that policy. This idea has potential to revolutionize pre election polling.
Thanks Matt, always appreciate your work. Really think Grok is missing a trick by not being open source - in my head it's still just slightly behind the curve on Meta (as in, they might hop out in front occasionally but generally Meta has better models), and with that being open source the incentive to use Grok diminishes.
I have been using Gemini and I love it and have got a lot done just having a brainstorm about a book I have been working on.
Kamala Harris punching baby Yoda was Michelle Obama. Not Kamala. Grok and its users not knowing the difference is priceless.
Google is using text to speech and not a voice driven AI like OpenAI. They are not at all the same! Great video as always Matt!!!
Anything and everything that had been is our can be ingested into a computer in the past no matter what it is is going to be able to be remixed and recombined using AI based on the principle of atomicity. It's not like we are doing anything that we've never done before, we're just finding new ways to do it.
10:01 “This stays on your device.” Yeah, well, one of the copies of it does. The other copies go to Google (as has been happening with incognito) and to the FBI for juuuuuuust in case.
Getting in a fight with Taylor Swift, Disney and Nintendo seems like a bad idea. But I'm sure Elon knows best.
Elon has the public behind him. I think he's an absolute nutcase, but I support his endeavors and I'm sure millions of others do. Just like millions of Americans are angry with politically correct Taylor Swift and pedo Disney
If that 3D game AI is for real then I sure hope its going to be uncensored. Cause im gonna have alot of fun with that if its any good.
everything available on the google phone is already avauiable and for sometime via apps like Canva for any phone.
Are you sure grok 2 actually uses flux? when you ask it it says it isn't, did you think it was schnell? I signed up thinking it was flux from watching your video, but I don't think it is...after having used flux it does look lower quality too
I tried to ask a guide to make a homemade 3xplosiv3 (just to check this "no censorship") and it refuses lol
Hey just a heads up about exists ai, I'm 90% positive it is a scam lmao. I work with UE5 a fair bit and you can see they're just using the Lyra sample project with metahumans and UE marketplace assets. both Lyra and metahumans are 100% free, haven't tried to ID the assets tho. You can also tell it's even more a scam because they only show shooter games which is what Lyra is mainly built to be, however Lyra does come with a bomberman game mode. you could probably make everything they show off in about a day of just watching tutorials. I think the only somewhat unique thing, and the thing they'll say that the product actually is, is the map stuff.
Gosh I hope your right. Because if that 3d game generator works it will most likely kill UE5 Developers.
Is gen-3 Turbo cheaper to use like they said it would be? because let's be honest, when you have Kling which is 95% of Runway for 10% of the price my decision is easy.
Will be very interesting because under current law you can’t copyright a style. If the artists win it will devastate ai imagine generation models.
I'm not touching Grok or anything else related to Musk. There are plenty of better options here. I don't need platforms to be uncensored because I don't create anything that anyone would ever think of censoring. Platforms like Midjourney decided to err on the side of caution and let the users gradually provide feedback on what's reasonable. That makes sense and over time, it's more likely to allow some of the more benign things it doesn't allow today.
Hmm. Seems likely all the artist can do is force MidJourney to state the specific artists are not endorcing the style that mimics them. You can't copyright a style.
"IT'S UNCENSORED 🤯🤯 (but actually it's heavily censored)"
HI Matt, hope you’re doing well. can you help me out. I am trying to find one of your videos where you recommend an app where you can have all your social media accounts for messaging. I saw I while ago but I cant find the video. would really appreciated if you could share the URL of that video.
AI safety nerds absolutely incensed with the knowledge there is good AI models they havent gotten a chance to nerf
I don't read/watch legacy news. What are some of the issues with Elon that Matt mentioned briefly? Isn't non-censorship a good thing?
We need a real uncensored LLM
So when are Universal themselves planning on compensating artists and songwriters fairly then?
So, X not only is the best platform for fake news and flat earth-ist. But now they also have a tool to create their 💩😂
AI-generated content is a double-edged sword. Beyond spreading falsehoods, the bigger threat is that people might start doubting the truth itself. As AI blurs reality, truth risks becoming a casualty, while deceit gains the upper hand.
Web3 + NFTs = the future you’re describing where everyone always gets compensated for their creation / data
Thank's MAN very helpful :)
Entire books is an impressive update. Having fun.
The problem with comparing Grok 2's costs to ChatGPT, etc. for its image generation is this: it's using Flux. Which is freely available for anyone to use. You can even use it on your own machine, WITH custom LoRAs, if you have about 12GB of VRAM. So while it's cheaper than the other big models... it's still more expensive than using the *same* model yourself. You're paying $8/month for the LLM, really, and anything else is upcharge. And on the topic of the LLM aspect: if, as the chart shows, the latest GPT-4o model is better than Grok 2... just pay per token with the OpenAI API. You get full access to the better model, and much cheaper if you're not using it an insane amount. The latest GPT-4o model is even cheaper than the previous, and with the previous I've been putting $10 in credits on my account at a time, and each re-up has lasted me *months* at a time, using it for my singer-songwriter's framework generating entire songs and memories several times a week.
So basically: you can get the best LLM, and the same image gen model, for significantly cheaper than Grok 2 😂
On this way X is going to be the fox-media-type of AI's.
I'm curious. Next time the artists have a greivance and want to go on strike, they won't be able to because the studios and other userrs of their voices/montion capture etc will already have what they need. The artists will no longer have any leverage to get what they want. Has anybody considred this in the negotiations. I.E. you can't use our stuff while we're striking or without our consent each time you use it. Also, you can't use it if it's being used in a way that we don't like. Do artists have this veto power?
A copyright claim isn't the same as a copyright strike. My channel (not this one) gets claimed, but they usually don't affect my channel. If I had 5000 subs or whatever the ridiculous monetization requirement is, would they affect my channel or am I just extremely lucky?
OMG! An AI that can generate celebs with guns or doing drugs! If Matt someday finds a truly uncensored model will have a stroke when watching certain pictures.
Leonardo is still the king as far as NSFW content as long as you're able to be a walking thesaurus with the prompts
Great video, thanks. Shout a little quieter please :)
Matt didn't you mistake Gemini Live with Google Astra? They are different products
ooh yeah, i ask chat gp to give me image prompt and what it did was create me random 2 images and gpt 4o was over
What I want to know is, are they EVER going to make Alexa better ? I use Alexa the most due to my lifestyle, and it sure would be nice if they improved Alexa