# AI News: AI Problem Gets WORSE, Humanoid Robots Next Year, Elon Musks AI Surprises Everyone and more

## Метаданные

- **Канал:** TheAIGRID
- **YouTube:** https://www.youtube.com/watch?v=IEwUPUxCIvA
- **Дата:** 28.08.2024
- **Длительность:** 26:40
- **Просмотры:** 24,427

## Описание

Prepare for AGI with me - https://www.skool.com/postagiprepardness 
🐤 Follow Me on Twitter https://twitter.com/TheAiGrid
🌐 Checkout My website - https://theaigrid.com/


Links From Todays Video:
https://x.com/elonmusk/status/1827981493924155796/video/1
https://www.reddit.com/r/ChatGPT/comments/1etkx1d/its_kind_of_crazy_how_uncensored_grok_2flux_is/
https://www.reddit.com/r/singularity/comments/1ezssll/andrej_karpathy_programming_is_changing_so_fast/#lightbox
https://x.com/elonmusk/status/1827981493924155796
https://x.com/BGatesIsaPyscho/status/1827763151619637420
https://x.com/tsarnick/status/1827465202734133694
https://x.com/tsarnick/status/1826814301770019241
https://x.com/tsarnick/status/1826130358238343353
https://x.com/thedorbrothers/status/1826314820972425417
https://x.com/TheAiGrid/status/1827037278176219537
https://www.reddit.com/r/singularity/comments/1ez8zzx/the_verge_about_reimagine_tool_in_the_pixel_9s/
https://www.theverge.com/2024/8/22/24225972/ai-photo-era-what-is-reality-google-pixel-9
https://x.com/ajassy/status/1826608791741493281
https://simple-bench.com/

Welcome to my channel where i bring you the latest breakthroughs in AI. From deep learning to robotics, i cover it all. My videos offer valuable insights and perspectives that will expand your knowledge and understanding of this rapidly evolving field. Be sure to subscribe and stay updated on my latest videos.

Was there anything i missed?

(For Business Enquiries)  contact@theaigrid.com

#LLM #Largelanguagemodel #chatgpt
#AI
#ArtificialIntelligence
#MachineLearning
#DeepLearning
#NeuralNetworks
#Robotics
#DataScience

## Содержание

### [0:00](https://www.youtube.com/watch?v=IEwUPUxCIvA) Segment 1 (00:00 - 05:00)

so one of the first things that I wanted to show you all was something that I covered before and this was essentially something that I personally don't think it's that good for society and the sense that this is going to be the age of misinformation now previously you know a couple days ago I did a video on this you can see that I did this video called AI just change everything we are in trouble and essentially this video right here was a video talking about the realism in AI now in this video I spoke about a few things and you can see here that you know the problem is that if you're able to generate photo realistic individuals in real time if you're able to generate people that aren't there in a way that looks super realistic this is going to be an issue now this is something that's a little bit different because it also adds on to the issue and in that video if you're thinking wait mid Journey's been a thing for a while in that video I basically just say that look there's realism and then there's photo realism is where you literally cannot tell if someone just took a picture on their iPhone but then there's like you know kind of realism where it just looks kind of professional so basically okay the problem is that and I spoke about this when this feature was announced in a video but essentially Google's new phone the pixel 9 has a feature okay and this feature is called the magic editor so essentially this is absolutely insane because this is the part where we get to uncanny photo realism with raw pictures and this is exactly what I was talking about in that video is that now we're in the chasm or the crossing over pit whatever you want to call it to where anyone who buys a pixel 9 okay and this is the article from the and I promise you guys you're going to want to see this because this is absolutely insane so what you can see right here is an image that someone has taken okay now you can see right here that someone has taken an image right here um of like a lake or whatever and then you can see right here that you're able to imagine certain things with you know the AI generated image so what you can do is you can effectively change what's going on in an image now I don't know about you guys I look at AI generated images all the time images from mid Journey darly any AI image generator you can think of I look at those images all the time now with this image there is no indication that this is an AI generated image whatsoever that's because it's taken the raw image and then it's added the subject in on it so it's sort of composited the image in adding all the correct Shadows all the correct lighting all the correct things that you need to do and what's crazy about this is that this works directly on the phone which means okay that now we're going to be in an age of misinformation where you're not going to even be able to tell if this is an AI generated image or not because it's so photorealistic you can see right here that the article says that this is all about to flip the default assumption about a photo is to become that it's faked because creating realistic and believable fake photos is now trivial to do we are not prepared for what happens off now I do want to say okay that lots of people say this and people are saying that this is apparently alarmist journalism and apparently this is look what the AI did after you explicitly told it to do that but I really do disagree because this is completely different you have to understand technology of course isn't good it isn't bad it's simply what we choose to do but the problem is that it's not something like Photoshop where there's this complete skill Gap like I tweeted on Twitter I said Photoshop requires skill and time but with AI it's a two second prompt that anyone can do so you have to understand that right now we are in an age where you can take an image like this and you can add anything in the image that looks 100% completely realistic now we have to understand the implications for this and whilst yes some people are going to be like okay this isn't that crazy people have had Photoshop for ages like I said before if you can instantly do something within 2 seconds this just breaks down on the entire barrier and we're going to have a lot more images that are going to open up a lot more cases for misinformation and to potentially ruin people's lives I remember you know earlier last year I talked about a case where someone's voice was used to say something that they didn't say and they got fired from their job they were like publicly embarrassed there were all these things go going on and I mean you know what's going to happen in the future with people that tend to use this the problem is that right now okay if I saw this image online okay I'd believe this like I have no reason to question these images you know like anything in the subject image but now okay like genuinely I don't believe any image I see I always make sure that I fact check stuff but the average person they're not going to do that for example you can see right here consider all the ways in which the assumed veracity of a f photograph has previously validated the truth of your experiences the pre-existing ding in the fender of your rental car the leak in your ceiling the arrival of a package an actual non AI generated cockroach in your cookout when the wild fires encroach upon your you know residential neighborhood how do you now communicate to friends and acquaintances the fix thicknesses of the smoke outside now I do know that journalism is very like you know it's hacky it's like oh my god look what a is doing but this is a genuine concern like you have to understand that all of these things are converging and remember there's always the one thing across AI technology that I always tell people is that this is the worst it will ever be and we can see some further examples you can see right here that we have a standard image then of course you can see that someone's you know added an image here that I probably will blur on screen right now um it doesn't contain anything crazy but it just contains like a car crash and something that just

### [5:00](https://www.youtube.com/watch?v=IEwUPUxCIvA&t=300s) Segment 2 (05:00 - 10:00)

doesn't look that great so I mean it's incredible like this is genuinely rather incredible and then we've got another image which is of um you know like I don't even know why they would put that there I'm probably going to you know uh blur that out too but essentially it's just an image with a package that looks really suspicious and if they were to take a picture of it you know it's likely that the subway would most probably get cleared out so um it goes on to State here that the erosion of the social consensus began before the pixel 9 and it will not be carried forth by the pixel 9 alone okay and of course you know you do have the fact that yes you do have ai generated watermarks for pictures but this isn't always like something that's going to work so I mean it's going to be pretty crazy to see how Society manages to evolve I think it's going to be pretty crazy considering we're now living in this age of you know distrust and disinformation but um let me know what you guys think about this I think this is uh you know rather crazy some you know just incredible piece of technology I think you know whil yes this could be used to create some cool pictures but the age of misinformation is here and that's what I want you guys to understand now you know another thing happen happened this week as well in AI that you know it's not news but it's still rather important because it goes to show how crazy this technology can be now unfortunately while I was recording this video um actually saw something on the timeline that I just couldn't believe and this is why I say that this issue is just going to keep getting worse as AI continues to get better and I do think that in the future there's going to be some you know insane regulation coming in place within the next 6 months um because this is just absolutely incredible so you can see right here that there's a deep thick live of Elon Musk um just going around just interacting with people and someone made a compilation video of the conversations and the worst thing about this is that if I was scrolling on Tik Tok and I saw this so I would genuinely believe that this is Elon Musk like there's no real indication that this is not like Elon Musk you know maybe it's just the hair kind of gives it away but if it wasn't the hair like I genuinely would think okay Elon Musk is just chilling in his Tesla and the worst thing about it is that I don't know if this person is an actor they're playing a role but they're able to mimic his intonations really perfectly like the way how you know he speaks like all that stuff I mean um you can even hear at one point in the uh Tik Tok live the live stream where you know people are like are you real and he's like yeah and then you know they're proceeding to have a conversation and they are none the wiser like I I can just imagine you know Elon Musk saying oh donate Bitcoin to this account and you know millions of people doing it can you imagine there was um also this area where you know a lot of Twitter accounts got hacked and someone posted you know donate to these crypto accounts but imagine now Elon Musk in millions of different locations presidents individuals going live asking for donations I mean just take a look at this live stream because this is insane this is probably one of the my favorite things that we made at SpaceX like forget the actual rocket like I want everyone to appreciate the lemon or float is this real I'm just as real as you and me oh my gosh okay so I was on your last live and I had a question since that last live yes of course okay so I saw a podcast where you were talking to Joe Rogan and you guys were talking about like the ocean and just like all the sea creatures are like possibly still alive and I was like really curious and interested if that was real Hi how are you Elon how are you I'm doing great how are you oh it's crazy to be on a live with you hey man I'll make it quick I don't want to take up all the time sure you have a mask on maybe well okay sending me a message on telegram by the way so if you remember before when I spoke about how flux you were able to generate super realistic pictures in the video essentially what you can do now with flux since it was you know trained on Gro not trained on Gro but released on Gro you can see right here that people are able to create you know realistic videos of individuals doing different things now you might be thinking okay this is funny we've got the presidents the people running for the you know race uh doing some rather questionable stuff and we've got Elon Musk waving a gun around in a store completely AI generated and you can tell that this is AI generated but what happens when in the future you for example train a model okay because you can train like a flux image generator on someone's face and what happens if you use that for someone and then you record that footage and then you use that footage to destroy someone's life what happens if that footage is used inappropriately what happens to the state of media when we can't trust any videos of certain individuals doing something are there going to be individuals who have done something and are going to be able to get away with those crimes because they can Simply Now claim that this is completely eii generated you have to understand that there are just complete you know implications for this so it's pretty crazy that these videos do exist and like I said before this is the worst it will ever be so whilst yes you know individuals aren't going to believe this they're going to know that this is a generated but it just goes show like I mean I don't know what kind of society we're going to be living in the future if there's going to be regulations on what kind of you know pieces of software you can use but it's going to be really interesting to see how the law manages to catch up with this stuff because the law is very slow

### [10:00](https://www.youtube.com/watch?v=IEwUPUxCIvA&t=600s) Segment 3 (10:00 - 15:00)

paced and AI is moving extraordinarily quickly like this stuff would have been Unthinkable 2 years ago so with that being said that is that part and you could also see that there is this picture right here that I wanted to show you guys cuz this one just looks you know absolutely extraordinarily super realistic like if I was browsing on Twitter and I saw that was like someone I knew like if it was trained on their face and I saw that there you know part of me would probably believe it but ever since I've stepped in the a space I don't truly believe any image I see online unless it's coming from a completely verified Source now you can see here that Elon Musk also gives us an inside view of the cortex today a new super cluster of you know AI gpus that are being used to train different models this is Elon musk's new super cluster and over the time we're going to see more and more computes being put into AI you know I even recently spoke about how there were recent you know super cluster predictions for the future and it basically says that worst case scenario we're going to get 10,000 times more comput by 2030 so that is 6 years from now um and while it doesn't seem you know feasible right now this is certainly something that could most definitely happen now the reason I actually included that was because there was this tweet okay and you can see that someone said damn x. Elon musk's company was started over a year ago and has leapfrogged anthropic on the LMS eval so if you don't know this is a chatbot Arena where essentially you have models that are compared based on certain different chats okay so you have Model A and you have model B person inputs a chat gets a response rates Model A Better Model A gets a vote and over the long time you can see that you know grock 2 has actually claimed the number two spot which is tied for with Gemini 1. 5 Pro now the reason that this is so surprising is for many reasons one of them being the fact that not only is x. a such a small lab it's something that was only started a year ago meaning you know just a year ago they had no product no team no nothing and in just a year they managed to get to the number two spot which is completely remarkable now of course I will say that this is questionable considering the fact that number one GPT 40 is number one and of course Claude sonit 3. 5 is number five so I will have to wait until this is tested against some other benchmarks such as the simple bench and if you're unfamiliar with the simple bench this is basically a benchmark in which it's based on you know simple reasoning so it's spatial awareness and things that involve just basic reasoning skills and you can see right now that currently clae 3. 5 Sonic is actually at the leaderboard top at 27% but the thing about this is that humans on average get 92% so I do think that once things like the simple bench basic reasoning benchmarks are solved in terms of like these models intrinsic intrinsically understand how to you know solve these kinds of problems I think that in the future these models are going to be completely intelligent Beyond you know a reasonable doubt now interestingly enough I did think that this was due to the way that LM Sy actually ranks these chat bot because what actually happens is that when you're voting for these if you're going to rate a model response sometimes the problem with Claude is that it doesn't respond so we all know that when we use CLA 3. 5 what tends to happen is that there are many different refusal requests for example if you ask it to do something completely normal um for example you might be asking okay how do I you know make more money in this industry it's going to be like ah making money in this industry is completely unethical I would rather you know plant vegetables in your garden or something like that some ethical nonsense and the problem is that if it gives you a response like that you're going to not rate that model as better than the other model which is going to give you an actual response so that's why sometimes claude's rankings can actually be decreased but I actually did tweet this but I actually got a response here that someone's saying you know here are the rankings without the refusal and you can see that without the refusal you can see that grock 2 is still at the number two spot right here and you can see the category is excluded refusal so it seems that grock 2 seems to be rather incredible now what this means for future models is that maybe grock 3 might just be the model that takes the number one spot because if a lab can be so lean so focused on simply just developing a model that is purely intelligent and not restricting itself in any ways I truly do wonder what they're going to be able to do with their next iteration of the model because it's only one year and this is what they've done which is you know truly remarkable and I remember when I was testing s column R it definitely wasn't as bad as the previous grock but it was really good like it was actually genuinely on par with these other models the only one issue that I have with grock 2 is that you literally have to go over here to your tab like grock and then of course this is where you can message it and then you switch to grock you know the beta the most intelligent model but I just rather like a website where you can actually just access grock cuz you know I don't want to have to use Twitter all the time just to be able to use the llm but I guess it's just a small issue now in some more chatbot Arena it does seem to be that Google is testing another smaller model we can see right here that there is this model A called little ending test so you can see

### [15:00](https://www.youtube.com/watch?v=IEwUPUxCIvA&t=900s) Segment 4 (15:00 - 20:00)

right here someone tweeted is Google coming up with a new model or just a variation of an existing one or maybe it's the same one I know about but can't reveal time will tell so you can see right here that Google's lead of product for Google AI Studio actually tweeted that testing a new engine regardless of how little it is it always gets me excited so you can see right here that this is in direct reference to the little engine tweet but I'm guessing that if this is a little engine it is most likely going to be some kind of small language model maybe open source for the future I'm not entirely sure what it's going to be maybe just Gemma 3 or just a different style of Gemma model but it is going to be rather interesting to see how these models manage to be considering the rate of AI deployment so I'm going to be really intrigued to see exactly how you know things move because one thing that I've known about Google is that recently they've been shipping a lot more products and a lot more features even if you haven't been paying attention then of course we had the C- here CEO actually talk about the fact that there is going to be no AI Plateau if you aren't familiar with cooh here they are the lab that is focused on actually making moral smarter and they released earlier this year command R it's going around of people saying we've plateaued nothing's coming it's slowing down actually really think that's wrong um and not just from like a we need to 10x compute and that type of thing perspective and trust me it'll get better um but from a methods perspective so when I was talking about like reasoners and planners and um models that can try things fail and recover from that failure um and Carry Out tasks that take a long time to accomplish these are like for the technologist obvious things that just don't exist in the technology today we just haven't had time to turn our Focus there and add that capability into the model um for the past year plus folks have been focusing on that and it will be ready for production so we'll see that come out and I think that will be a big change in terms of capability so yeah I'm definitely excited about the future of AI you can see we just haven't had the time and I'm going to be excited for this stuff to actually come out because I think yes right now even as many people think we're in this late stage of adoption I still think the AI is still pretty early in the grand scheme of things of how it's going to impact Society like right now you think that you know we're at the peak they're at the peak of like entire AI hype and adoption but I still think it's pretty early on like a global scale like I think this small chat GPT moment whilst yes it's transformed the global ious I think if we look for 20 years we're going to see that wow that was really just the beginning of this complete transformative shift like for example we don't have you know personal AIS that can memorize things we don't have AIS that can go out and do stuff for us like there are just so many different things that we just don't even have yet for example iPhone still have Siri which is you know awful now we also have Terren tal saying that AI will enable mathematics to be done at an unprecedented scale solving one theorem at a time and solving each problem separately which you know allows AI to you know work on the space of problems and classes of thousands of problems at a time prove one we prove one thing at a time or maybe two or three if you're efficient um but with AI you could imagine in the future um instead of trying to prove one solve one problem or something you take a class of 10,000 similar problems and just say Okay um I'm going to tell you AI try to solve these 1,000 problems with this technique and will pull back oh I could solve 35% of these problems with this technique what about this technique I can solve this percentage of problems or if I combine them I can do this you could start exploring the space of problems rather than just each separately um and this is something that you just you either cannot do right now or you do over a process of decades with dozens and dozens of papers slowly figuring out what you can and can't do with various techniques but you with these tools you could you could really do I mean you could really sort of start skate you know doing mathematics on a scale which is really not really unprecedent um so you know I mean the future is going to be really exciting I think um you know I mean they we will still also be proving the old fashion in fact we'll have to because we can't we won't be able to guide these a as we also know how to do these things ourselves but we'll be able to do lots of things that we can't do right now so you can see right here we have a tweet from Andre carpath where he's speaking about how programming is changing so fast and he talks about how you know using you know cursor plus Sonet 3. 5 instead of you know GitHub co-pilot and he says that you know over the last few of his days programming is now writing English now this is pretty crazy because Andre karpathy is pretty much a legend in the AI space so when you have someone of this level actually talking about how they view the space in terms of coding and AI tools and how they're doing their day-to-day workflows of it just being you know half coding where you write the first chunk of code then of course you tab tab through completions I think it's going to be rather fascinating considering the fact that a lot of what we've seen in the jumps in terms of the ability to now be rather effective is simply because of Sonic 3. 5 like I've even used Sonic 3. 5 to code a few different projects and it's just Remar able at how good it is and the crazy

### [20:00](https://www.youtube.com/watch?v=IEwUPUxCIvA&t=1200s) Segment 5 (20:00 - 25:00)

thing about this is that this kind of gives me the impression that we're going to move towards a rapidly increasing area where AI enables people to be able to build the software that they truly do want and allow people to you know build the kind of programs that they instantly want just via a few prompts and I know that might sound completely crazy but we have to put things into perspective Sonic 3. 5 wasn't really you know anthropic main model this was their mid-tier model so when they do release a future model like claw 3. 5 Opus or you know for example we get GPT 5 or in fact Gemini 3 or Gemini 2 I can only imagine the kinds of coding capabilities that we're truly going to get out of those models and it was only you know not that much of a jump but we can see that this small jump enabled so much people to come into the space which is just truly fascinating so I know that some people have said that you know this isn't like a real thing but I think when we're starting to see people like Andre Kathy stating that now this is part of their workflow we have to truly understand that you know change is going to be coming and we can also see that here with the Amazon CEO Andy jasse actually talking about how one of the most tedious tasks for software development is updating foundational software it's not new work yada y but basically this you know they're talking about Amazon Q their generative AI assistant for software development and the entire thing here is they're saying that you know this is a new code transformation capability and the average time plummeted from 50 developer days to just a few hours so this has saved them you know 4,500 years of work and that number is crazy but real so this is why I'm saying that we are in truly a transformative area especially in you know software and coding and what these llms are able to do with an increasingly level of reliability and of course increased accuracy in certain areas like this is truly going to change stuff now you can see right here it also states that you know in six months we've been able to upgrade more than 50% of our production Java systems to modernized Java version at a fraction of the usual time and effort and it says of course this provides an estimate you know $260 million in annualized efficiency gain so one of the things that I've always know about Amazon is that Amazon are going hard on Robotics and on the software like the Amazon Q agent it actually was state-of-the-art for software development until you know there was a recent one that was released that I'm going to be talking about but basically Genie currently 43. 8% on the benchmarks but this of course poses interesting questions for the future of work you can see here that you know this is the system this is how it works but like I said before guys things are changing like so much and I truly can't imagine the amount of businesses that are going to get built off the back of GPT 5 the amount of things people are going to be able to go and build like you're say okay GPT 5 GPT 6 go ahead and go build this and it's going to be able to go and do it now I know not right now of course you're still going to need humans in the loop but it's just you know truly incredible how much everything is changing simply based on the ability of certain models now in terms of the future being even more crazy figures CEO Brett Adcock says that the company's production line for humanoid robots will be operational in 2025 starting by manufacturing hundreds and scaling to thousands of robots so humanoid robots are coming in 2025 uh we're going we're actually starting our production line next year where going to go where you going to do uh we're going to do that here in California like close to engineering so we really work on The Kinks on like more of a traditional like kind of pilot manufacturing line or production line and um and then from there we'll start announcing our intentions for like where like high rate look but um our production line is being uh we're starting to design it as we speak uh we'll have Robots coming off the production line in 202 and you're scaling it for what uh we'll start with hundreds of robots and then thousand we're actually starting our production line next year now more robot news something that actually did go viral was this clip in China basically this was something that went viral because I got to be honest with you guys when I first saw this robot walking I was like this is incredible like that kind of human type walk is just ridiculous but this is not an actual robot this is actually a human that is dressed as a robot so this is kind of like what Tesla did a few years ago when they were demoing the Tesla bot it was just a person in a costume but I think China is going to be the leader in robotics despite the US's head with Boston Dynamics and other labs such as figure but it kind of showed up because the public reaction was one that was like wow this is insane but I mean once the public like the general public truly realized how the future is going to be in terms of how capable these systems are going to be I mean I think Society is going to be rather interesting to see where people's minds are at in terms of accepting the technology if they want to adopt the technology into their daily life and all those kinds of things and this is the last thing I want to show you guys which is fidar Jim gal actually talking about how you know AI can actually lead to the discovery of new knowledge because a lot of the times what you're doing is you're taking in a lot of data trying different things and

### [25:00](https://www.youtube.com/watch?v=IEwUPUxCIvA&t=1500s) Segment 6 (25:00 - 26:00)

that stuff that AI just truly excels at um Dem is a SE of um of Deep Mind calls uh you know like AI creativity right it's the ability to acquire knowledge that did not exist before right and I of course experienced this firsthand the reason why I'm such a True Believer in the technology is because again I was the expert who helped design the system but this very AI agent that we created is telling me new things about the system that I didn't know about before right and that's a very powerful feeling it's kind of like when um you know if you think back to Alpha go right like Lisa do was the best in his field at go he was the world champion for a decade right he was at the top and his ELO rating was just something outrageous it was like 2800 or something it was outrageously high and but it had flat you know Flatline right so for a full decade his ELO rating was the same and there was no one to challenge him because he was already at the top so once he hit the top he just kind of plateaued and then after Al go happened and he actually got to play against Alpha go you know privately a few more times because de mind you know had uh let him continue interacting with the system what happened for the first time in a decade his ELO rating started climbing and so this is what I mean when I say that I think the real power of AI is helping us discover knowledge that we didn't necessarily know about before and where you're going to see the most um gain from that it's not going to be in routine automation things right like call centers or whatever right uh it's going to be I think in very complex areas right like areas where human intuition is insufficient because of immense complexity but that is yet underpinned by D

---
*Источник: https://ekstraktznaniy.ru/video/14101*