RED FLAG for The AGI Race - Please, OpenAI...

2024 just started and in these few months we already saw AI getting 10 times better think of Sora and the videos it generates the quality and consistency of them but more breakthroughs are under way where inches away from tipping the scales so it's crucial for us to know where all this is going so without further Ado let's get to it the first

scary thing that comes to mind is deep fakes I am not Morgan Freeman and watch you see is not real defects are essentially AI generated videos or audio that are manipulated to make it look or sound like someone is saying or doing something they never did remember that scene from Mission Impossible where they used masks to impersonate people deep FS are like that but in steroids and the scariest part they're getting shockingly realistic take a look at this video by quarter digital Tom Cruz in the studio the video is 4 years old already and now it's super easy to see where the Deep fakeness Falls for the face is kind of glossy and super smooth easy to distinguish from a real one but take a look at this video this is what a modern deep fake looks like it's indistinguishable it's spotless it's perfect Johnny's video is great by the way check it out if you want to know more about the inner mechanisms of deep fakes i instead want to look at the Dark Side of all this imagine this scenario it's selection season and the video surfaces online showing your favorite candidate

making offensive remarks were admitt to a scandal your shock confused and unsure what to do but wait there's more a news report merges featured an interview with a candidate denying the claims and calling the video fake except it's not it's a deep fake so convincing that even the candidate can tell the difference this isn't just a hypothetical situation in 2020 deep fake video of Nancy palosi went viral making it appear as if her speech was slow down we had some level of agreement but you never know the video was wildly shared on social media despite being debunked by experts creating deep fakes is becoming easier and cheaper than ever there are apps that allow anyone with a smartphone to create basic deep fakes with more complex tools readily available online like deep face web or bidn face swapper these web tools take only a few minutes to create and for some serious deep fakes you still need powerful systems and wellt trained models but the tech is getting better every day just think of all the fraud and identity theft that becomes possible with this Tech if you think that high quality deep fakes are possible only with huge budgets and that

you are not going to encounter anything like that for a few more years I want you to close your eyes and imagine something imagine your grandma receiving a call from an unknown number she picks up the phone and Hears A Familiar voice your voice asking her to withdraw all her savings and sending them somewhere to save you from problems sounds bad right well now imagine you are on a vacation tenning Under the Sun away from your phone you don't know it yet but your Instagram and Snapchat have been hijacked when you connected to the unsecured Wi-Fi and while you're drinking margaritas at the bar video of you pops up in your friends feed asking them to help you out with a few books the video is up only for a couple hours and then gets quietly removed this cameras got what they wanted and no one suspected a thing your voice was copied in just a few minutes and your face was stolen and slapped in the fake video using freely available software does this sound realistic to you deep fakes are not only about disrupting the political climate or misinformation in the upcoming years what we'll see a slew of deep fakes attacking us our friends and our relatives and we won't be able to do anything about it that's what scares me and it sure as hell should scare you let me tell you about

something way less scary Pleasant even this week I also found a great new AI toll called pixel is an amazing AI part photo editor with a ton of really cool features that you don't often see in big apps like Photoshop take image generation for example you just tell pixelar what you want and Bam you get your Perfect Image in seconds the generative expand feature is pretty neat too say you've got a great photo but wish there was more sky more grass or just more anything on one side this feature fills in the blanks for you making the extension look natural want to see your faing the billboard or maybe swap faces with your favorite celebrity for a left AI face swap can take care of that and with 250 faces available in the free trial there's plenty of fun to be had removing noise from photos can be a real headache but pixel makes it a oneclick operation it cleans up your photos getting rid of all that graininess that can ruin a good shot making everything look crisp and clear then there is the generative transform allowing you to detach and reposition elements in your photo removing backgrounds objects and scaling up images without losing the quality are pixel bread and butter and it does it all perfectly pixelar creates a seamless ecosystem that integrates a variety of tools in a single space this means you can jump from creating an image with AI prompts to editing further without needing to switch applications or interfaces pixel is sponsoring this video and I will leave a special link in the description so be sure to check it out in March 18 laye hburg 49-year-old

woman was walking her bicycle across the street in Tempe Arizona when she was struck and un alived by a self-driving Uber car even though there was a human in the driver's seat the apparent trust in the AI was so high that no action was taken to prevent the accident for me one of the most unsettling aspects of cell driving cars is the absence of a human behind the wheel without the driver's intuition Split Second decision- making abilities responsibility Falls solely on the AI but how can we be sure this AI is truly capable of handling the unpredictable nature of the road especially in emergencies this reminds me of that trolley problem which way should the trolley go five people or one it's impossible to make the choice here impossible to be unbiased impossible to calculate everything the car is programmed to follow algorithms but these algorithms can't account for the nuances of human life and the emotional weight such choices are you comfortable in trusting life or death decisions to lines of code but accidents are just a part of the problem think about the potential for hijacking there's this video on the web of hackers gaining a

remote control over 2015 Jeep they could drive it into a ditch turn on the AC lights and do all sorts of things at one moment they even shut down the engine on the highway now remember that the Jeep in the video isn't new and doesn't have all those fancy security protocols self-driving cars not only have Motors that can control the steering wheel and accelerator pedal but also have a constant connection to the internet this basically gives hackers all the tools they need to wreak havoc on the streets I think there was a scene in Fast and Furious with the cars rolling off of a parking lot down on the street below now imagine this happening on a much wider scale imagine thousands of cars going rogue speeding creating all sorts of accidents all around the city or even all around the country the sheer number of possible crimes that full self-driving cars allow for is staggering from abduction of people to literal karmageddon on the streets but it's not just about AI assist in humans it's also about AI replacing humans and that's where we could start talking about

about robots at first glance robots seem expensive difficult to produce and very sophisticated in terms of software they are also not very maneuverable and seemingly can not perform the duties of a person all that is in the past because if you're following the news you already know that past couple of years robots have made a huge leap in capabilities we all saw those Boston Dynamics videos of their Flagship humanite robot in the parkour sure it is a bit slow and takes its time to make a move but this example already shows that machines if programmed and engineered correctly can move and act like humans at a similar Pace as for the price it's also not as obvious as you think recently A video

popped up from a company unit tree robotics of their newest humano robot priced below $90,000 this robot lacks the fluency and sophistication of the buson Dynamics robot but it already can walk like a human and withstand small hits and kicks and Tesla is also working on their own humanoid Rob but in a fresh video they showed walking almost like a real person full body control enabling it to do squats and even real hands with tactile sensing give the tech 10 more years and we'll get our own Androids but let's change the topic to something less Sinister another disturbing Trend we're

heading towards is AI influencers we recently made a whole video about this phenomenon magalo liil Micha these are the most well-known faces in the game hundreds of thousands of followers making their creators thousands of dollars every month off of brand Partnerships and collabs why are these AI influencers blowing up for brand state offer perfect control over the message no worrying about a celebrity getting caught in a scandal or saying something controversial they're also globally accessible can speak multiple languages and never age for followers AI influencers can cater to the most unusual needs beauty standards and so on as we show you in our video creating such AI influencers is super easy all you need is a free software cleverly written prompt in a few minutes of your time but the scary part has nothing to do with the authenticity of it all or the financial side of things and deception the scary part is that AI

influencers are just the beginning right now these are just photos on Instagram but soon it will be videos we all saw what Sora is capable of gosh even now you can slap an AI generated face on someone's body and call it a day without any repercussions think of all the identity theft that will be happening when the tech will become red readily available to everyone think of all the unlawful and adult videos appearing online featuring random people when that's a video with a Hollywood star we don't really care but what will be your reaction when you see your face in one of such videos the most disturbing part of it all as you might have guessed is ethical and moral side of things pushing the boundaries of what is acceptable in society but definitely the scariest thing that's been kind of a boogi man for years now is Agi

and the reason why I'm included it is all the news about AI being achieved internally within open AI hold on don't write a hateful comment just yet let me explain why many people are genuinely afraid of AGI maybe after this you'll be at least a bit more cautious just like I did after doing my research AGI stands for artificial general intelligence and it basically means full thinking sensient AI however despite presenting many benefits if used as an agent for specific tasks the concept of a thinking machine is precisely what gives people the chills computers don't think like us they're extremely good at processing data but terrible at pattern recognition which is the opposite of humans if a computer reached a point where it could be considered intelligent or super intelligent we would have no idea what to expect from it nor will be able to easily comprehend its thought processes an alien from a completely different planet with entirely different biology is more likely to be mentally similar to us than a true AI because they at least understand basic concepts like pain and pleasure that are common to all forms of life and this lack of morality can be dangerous for example a few months back researchers have conducted an experiment on large language models researchers programmed various large language models to behave maliciously then they try to remove this Behavior by applying several safety training techniques designed to root out deception and ill intent but even after all the techniques used by researchers the models continued to misbehave which shows that once AI learns something it's practically impossible to make it forget it our key result is that if AI systems were to become deceptive then it could be very difficult to remove that deception with current techniques that's important we think it's plausible that there will be deceptive AI systems in the future since it help us understand how difficult they might be to deal with but let's imagine for a second that we managed to create something smarter than ourselves then that thing can itself and its creation

creation can create something smarter than itself and so on this could potentially happen very quickly creating an AI so much smarter than us that it might as well be omniscient a concept called super intelligence explosion or technological singularity the reason the super intelligence explosion is scary is that we can guarantee the resultant super intelligent AI will do what we want it to it will do what it's programmed to do but what if we program it wrong the classic example is a paperclip company making an AI and programming it to make paper clips the AI goes super intelligent takes over the world and kills us all so it can make more paper clips out of us this isn't because the AI is malicious it's just doing what it was told to do and we told it to do the wrong thing AI has a tendency to find shortcuts to meet a goal while technically getting the word it's seeking a good example would be a game in AI using a glitch in the game to maximize its score now imagine something like that were to happen with an AI we wanted to cure cancer but instead of giving us a cure the way we wanted it proceeds to eliminate anybody with cancer because when nobody with cancer is alive it will technically have achieved its goal this unpredictability is exactly what scares people because even with all our coding the AI at the end of the day is a black

box we set it up and let it educate itself modify itself to do the job it was tasked with and we cannot predict the route it takes and the changes it makes now let's change an angle for a bit and talk about something we all hate job hunting the job hunt has always been a competitive landscape but in recent years AI has been changing the game so rapidly that I can stand to ask did AI already become the gatekeeper of opportunities for a couple years now companies have been using Ai and HR more and more often not only to read the CVS and search for keywords but also for analyzing the facial expression and emotions of the job Seeker during the interview itself some people say this approach is less bias but reality proves them wrong for example a 2018 investigation by Reuters revealed that Amazon had to grab an AI recruitment tool because it discriminated against female applicants this happened 6 years ago and even though the tech has evolved dramatically since then the point still stands also there has been a recent controversy with Google's Gemini refusing to draw white people unless asked to this led to some people calling it woke and even Elon Musk said a few

words in a response to original post he said and I quote Gemini is indeed just the tip of the iceberg the same is being done with Google search also there is another problem with AI and job hiring lurking around dehumanization of the process if the algorithms are reading your CV picking specific words from it and assessing it solely based on its own

internal parameters where does it all leave us aren't all people supposed to be different should we just abandon our personalities and submit ourselves to the AI well because it seems that if you want to get hired your resume should have certain keywords inside and it doesn't matter how motivated you are how talented you may be and all those little personal things that only a real human can pick up on the reality we're heading to words slowly transforms a deeply personal process of hiring into something algorithmic and soulless it forces people to pretend to lie on their CV just to get the job or at least past the screenings and it will only get worse and I'm not even talking about AI in education because this topic deserves its own separate video but it comes to no surprised that more and more students

around the globe are using AI to cheat they use Chad gbt to write their essays use other AIS to create presentations use AI math tools to do their homework and so on this creates a generation of students who are dependent on AI students imposters who don't know anything but only pretend to know 2023 study by Stanford University found that students who relied heavily on AI code and assistance wrote significantly less secure code than those without access additionally participants with access to an AI assistant were more likely to believe they wrote secure code than those without access to the AI assistant so we have this hot combo of people who know nothing and rely on AI to know everything lie on their resumés to pass the check from Ai and get the job where they are just going to use more AI to do it seems like a vicious circle of ignorance don't you think and with more advancements and tools like Chad gbt this issue will grow bigger and bigger AI will write even better do even more and eventually people will be the operators reduced to the role of an overseer some might call it heaven but I think it will lead us to the world of

Wall-E we are surely heading towards a pretty scary future and the more AI advances the scarier it becomes but there's also a different perspective AI already helps Humanity helps save us from monotonous jobs it helps with cure and diseases it helps us educate ourselves AI is n good or bad it's all about how you use it I intend on using it for the greater goods and advise you to do the same thanks for watching and see you in the next one