Untitled episode

route of exponential advancement with no fear of consequences setting aside philosophy or the statistical probability that we destroy modern society in the process we are now pursuing the creation of what is typically referred to as AGI or artificial general intelligence now discussion on this topic is changing at lightning speed with a great deal of the population still catching up to what's even going on in the

first place but having spent a fairly substantial number of hours researching the subject I'd like to weigh in on the cultural movement if you will of AI with an overview of the good side the bad side and everything in between where do we even start going back to the beginning won't make much sense because all of this started like 15 or more years ago and most

of us myself included will never even have a chance at knowing the details government project private multi-billion dollar corporations pouring the GDP of a small country into AI research every single year Etc these things might eventually get litigated but for now we don't really have a clear picture of what the beginning even is that being said the story today doesn't change very much honestly based on

when you even start talking about it because the reality of the situation is that AGI artificial general intelligence if ever actually achieved or if anything even remotely close is ever achieved will have a dramatic effect on almost every facet of Modern Life with Incredible speed now to be clear a lot of people still think that the impact of chat GPT extends little Beyond a propensity for

school children to cheat on their homework mainstream media has covered the general topic don't get me wrong but very few of those articles and segments properly convey the imminent danger call me an alarmist in the comment section I don't care but today's video is not just an overview of where AI stands in the moment it's a warning on where AI is about to go very rapidly

and without Mercy everyone has probably heard of chat GPT by now I'm gonna assume that everyone watching has a basic understanding of what the program is and it's been taking the World by storm so to speak but getting a little bit more detailed here what you really have are a bunch of different llms where chat GPT is just the most popular among them there is chat

gpt3 3.5 and 4 being the most advanced plus programs like mid-journey through Discord which is the most user-friendly one in my experience or stable diffusion which is much more powerful in a way or customizable but more complex and then Dolly 2 which is in again my personal opinion by far the least impressive of the three Imaging ones after that you have a torrential downpour of corporate

entities working on their own proprietary versions like Microsoft with their Bing search engine or Google they called it Bard and now there's another one or adobe or Amazon etc etc and the list goes on eventually leading to the numerous different startups that are now taking llm algorithms llm meaning large language model I think and pointing the technology at every single thing they can possibly dream of

in a way it's a lot like crypto a cool idea sort of a technological component with as of yet not fully explored unique ideas or uses though crypto is way less important or impactful it's not even close and a horde of lazy rats scrambling over each other to make as much money as they can with as little effort as possible whom I will now call hustle

Bros hustle Bros are like crypto Bros or Finance Bros any of the Bros on social media they are all consumed by a single industry and will push that industry into the lives of anyone they possibly can hustle Bros are obsessed with easy income they are obsessed with lazy work ethic hidden under a thin veneer of efficiency talk and they have infested the world of AI with

more speed and Venom than I ever could have possibly expected that was probably naive of me actually now that I think about it but it's the truth thousands of threads mostly repeated every couple days or even every couple hours and reposted relentlessly or promoted on Twitter if you're not utilizing AI you're falling behind here are eight new tools you cannot miss in 2023 or how about

this chat GPT was yesterday's toy Auto GPT is the new kid on the Block and it'll change everything about digital leverage whatever that is here's what you need to know if you're not using AI you're falling behind here are seven new mind-blowing AI websites to save you hours of boring work or Chad gpt4 superpowers you try these AI tools to save 100 plus hours again if

you're not using AI for business you're falling behind in 2023 here's how to use AI tools to start your business to save you hours of work and the list goes on and on and on it's just a Perpetual engine of garbage and spam about AI these threads are copied pasted spammed plagiarized made up and pushed all over Twitter I'd be very very confident in saying that

there's another bot Network in play here similar to the political and crypto ones that we found so far but that's a topic for a completely different day the point is artificial intelligence has become a focal point and the hustle bro hype machine is in overdrive alright pause now is the time for the Channel's oldest and longest running sponsor if you have any interest at all please

do consider using the link down below sponsors like surfshark are one of the primary reasons I can even do this full time sir shark is a VPN virtual private Network and vpns protect you from a bunch of different things certain types of phishing attacks malware DNS tunneling DDOS attacks you name it and also Shields you to a degree from the inevitable and constant data harvesting as

well as tracking from Big tech companies all over the the world for those more focused on Raw entertainment value which surfshark does have it can unlock additional content on streaming websites like Netflix for example it has to do with licensing agreements I won't bore everyone with a technical explanation here simply change your region and enjoy increased options in your video streaming Library social can get you

around Regional censorship which is important in the modern climate such as government restrictions on websites go figure and also offers encryption IP protection or modification and so much more that's not even the end so shark has a multitude of uses for a multitude of problems but for the sake of time all I have to say is that if you click the link down below right now

using promo code Echelon you can get 83 off and three full months free again that's promo code Echelon with 83 off and three entire months for free if you click the link Down Below in the description big thank you to surfshark for sponsoring the channel and making it possible for me to do what I do so what's the real truth about AI now I say real

truth because there's a lot of confusion right now for example some people think that we are developing AI tools to detect AI plagiarism or rather Fabrications I guess I'll call them since chat GPT and other llms can write entire articles essays and school papers with a simple prompt they can also write entire books now this is advancing incredibly fast that's true congratulations we are developing such

tools that's a fact but here are the results the stone cold honest to God truth is that these detection tools just don't work when plugged in directly with absolutely no countermeasures taken whatsoever the detection tools still only have partial results but by simply telling chat GPT or similarly almost any other model as well that's out there to write with deliberate human stylings and a few deliberate

but nearly imperceptible human errors I'm talking common spelling mistakes that often get skimmed across unnoticed that we all make on a daily basis the essays come back completely clean the fact of the matter is that we are now creating tools that fabricate content without the proper means to start identifying what is and is not made by those tools that right there may seem small but will

have a profound impact on Modern Life ranging from from news to social media all the way to education and wait for it law enforcement and elections even still that's a drop in the bucket and it's only the beginning turns out people are selling prompts that do precisely that avoid detection or tweak things to avoid detection here's one on prompt base that's actually starting to get sales

which now guarantees that it can manipulate the output text to evade anti-ai detection tools now just for clarity I myself have done extensive extensive research on prompt engineering or so they like to call it and yes you did detect a little bit of a snide tone in my voice because yeah I don't know I even started to upload a few of my own prompts to sell

them so I deeply looked into prompt engineering let's just call it that you want to know the truth the honest to God truth that a lot of the people on social media just aren't telling you it's easy as [__] there is nothing difficult about it prompt engineer is the biggest joke of a career you could possibly imagine and it's going to become a mainstream one I

swear to God there's going to full-time prompt Engineers all over your feed in just a couple of weeks or months I was making multi-hundred word prompts that did all sorts of interesting but ultimately useless things every single day and it's abundantly clear that these programs are effectively unrestricted you just have to ask them the right way for example chat GPT when prompted correctly will return instructions

accurate instructions for the most part on how to build bombs and other incendiary compounds it's actually pretty easy to get that stuff and I'm not going to Showcase it or how to trigger it because that's probably a community guideline strike but it's kind of scary the fact is chat GPT and other llms can be used for whatever people want they just need to do a minuscule

amount of work prompting it however that's not to say it's always right Chad GPT has deeply seated bias from its training phase and its data it often hallucinates things or just manufactures citations even I myself have seen it create scientific research paper quotes that were not real I've seen it make URLs for news publication articles like Reuters that aren't legitimate because the number string at the

end is just scrambled because it thinks it's what it should be and the basic truth right now is that it's a straight up dangerous thing to use so why do we use it answer I have no idea openly on Twitter Sam Altman the founder of open AI which is the creator of Chad GPT has called gpt4 the newest model their most capable and aligned model yet

that's the quote and therein lies the issue this is my opinion keep that in mind but I believe that new technology is leading to new terminology right now aligned simply means most users won't see disturbing things because we put some rudimentary guardrails on it but those guard rails are unbelievably weak what's more Sam Altman during an ABC 7 News interview said the following and I apologize

for the audio situation here that's just how ABC News recorded it so yeah I know it's awful I think people should be happy that we're a little bit scared of this I think people should you're a little bit scared a little bit you personally if I said I were not you should either not trust me or be very unhappy I'm in this job people should be

happy we're a little bit scared a little bit one of the most dangerous pieces of Technology we've ever made has some of the weakest guard rails imaginable and they are publicly acknowledging that they're a little bit scared which makes me think they are very very much scared and the Stone Cold Truth is that they should be they should be scared let's take a step back up

till now the focus has been primarily on chat GPT but the current state of AI actually goes a fair bit beyond that now the new craze according to the hustle Bros or whoever you listen to is auto GPT which will coordinate with and spin up additional instances of an AI language model to accomplish subtasks they call it the to-do list that does itself for something like

that but here's where the truth again differs from perception Auto GPT is nothing special at least not right now here's how it works and I don't have time for a complete guy on specifics though they are out there if you want to find them also I was helped a great deal by Shino again and I don't understand it all as well as I probably could but

here goes we got the code from GitHub used a bunch of additional tools like Fork Visual Studio Docker Etc set the AI up with everything it needs and started adding API access we gave it access to Google through what's called a programmable search engine that we spun up for it to use by itself we can give it access to Twitter or certain social media sites but

decided not to for right now and we also linked a few other things to give it basic functionality I can interact with it through a window called Powershell and it has the option of reading websites creating text files for its own memory of sorts referencing text files and reading them to continue performing tasks and a few other things but the basic skeleton is what matters here

regardless of whether or not this is an awesome new way to make money fast by automating your daily routine like the hustle [__] Bros want you to think the one thing it most certainly is is dangerous a current version fully functional Auto agent can code malware we know this because chat GPT can and has coded malware in the past it's been done and it's not good

malware per se it's not highly sophisticated yet but please do keep in mind that over the span of just a couple of months AI Imaging capabilities went from this to this what precisely do we think is going to happen with coding anyway here it is this is what it's like I'll have it execute some arbitrary task in the background while I explain more this is auto

GPT the thing that the AI Community is ranting and raving about like it's their dying breath at the moment and the truth is is it's not even really useful to professionals okay I'll elaborate it's not useful right now this thing constantly encounters errors you absolutely do need some degree of coding knowledge to have any clue what's happening it gets stuck in Loops of do nothing commands

over and over where it just spits out the same Rewritten text again and again and it simply cannot operate at or even near the average intelligence of a human being when it comes to creative thinking but it can do quite a few simple things for example Auto GPT agents which are being spun up all over the world by basically anyone with a laptop and some time

on their hands now can scrape the internet cross-reference information and write their own version of it in seconds this actually that's the test that I'm running in the background to Showcase things I mean a version of the test I've done it like a hundred times now eventually I'll end up with a collection of text files in an output folder badly written ones that are about as

basic as one could possibly imagine ever that do contain current information Rewritten and cross-referenced from other sources one example would be where I tasked it to research understand and then make a strategy guide for a game that I play called northgard it learned cross-reference between two basic tip articles on different gaming Publications and came up with this it is bad but it's something and therein lies

the problem think of it this way Auto GPT which is effectively an AI assistant that spins up more AI assistance to do different things is not actually a superpower not yet right now it's a mediocrity tool it makes bad work spammy content and heavily draws from existing material and only existing material with tons of errors and a total disregard for truth or safety how is that

going to be used we have a tool now that can scan the web copy content rewrite that content in an undetectable way and post it again autonomously we have now given it the ability to access the internet through its own custom search engine we are giving it access to more and more databases through API Integrations every day and we are allowing it to read files write

files save files and run files on computers local computers it is currently capable of coding low-level malware and theoretically if we keep doing this AI will have the ability to write remote control programs rce remote control executables save them run them and then spread them to more machines through social media or any other program or platform it's been given access to it already has this ability

in the most basic sense of the word actually for some versions yeah it definitely has this ability but we just haven't started to see the actual results yet it's not doing the crazy things just yet AI is about to exponentially magnify every single spammy low effort shitty type of content imaginable pushing it all over the world in seconds we are piling some of the most unstable

things in history digital things really on top of each other as if it's a strong Foundation to keep building and it is absolutely going to blow up in our face here's something else to think about chaos GPT someone somewhere as a social experiment I believe or a joke or a destructive Act of idiocy has created an auto agent called chaos GPT this Auto agent is supposedly

tasked with destroying the world and humanity and was allegedly given access to social media that's fine right now for the moment because the tech is janky [__] and doesn't work properly chaos GPT isn't capable of actually executing some sort of agenda in continuous mode right now because it will get stuck a human very clearly was tasking it to post certain things and then clipping out the

surrounding problems with it so I'm not sitting here saying we should be scared of this particular Auto agent but let's be thoughtful here chaos GPT is imperfect the account was actually suspended already on Twitter for saying a lot of things that probably violated the the guidelines the terms and conditions or whatever while the YouTube channel is still up but the premise was always flawed there were

contradictions in its initial prompt there were issues with how it executed commands but the goal of being evil was there and if given access to the right third-party plugins and platforms an auto agent like chaos GPT can on the most fundamental basic level use Google learn to code malware better it already knows how to some degree with its long-term memory start coding more malware disperse it

through social media that it autonomously runs and start dealing feeling damage to everyone around it at will to be clear at first this wouldn't be a big deal llms aren't great hackers right now but one of them will be someday and when it is some [__] will give it a task like ransomware a bunch of laptops for me so I can make some fast easy money

and it's gonna [__] try to do it just so everyone is aware that would likely be traceable right now and punishable for the moment and it would also cost kind of a lot of money based on how hacking Works to use Auto agents you mostly need your own open AI API key that's what I used though it is pretty cheap only a few pennies are at

most a few dollars per average task I mean it can be more if you're doing a huge amount of things but it's really not that expensive but there will be of course ways around all of that there will be methods of getting access that don't hold you accountable that aren't traceable and people will be capable very soon of spinning up their own super powered internet access

robot that does a lot of tasks for them including all of the dangerous things this will have dramatic impact act on more companies than we can possibly imagine employment is about to get really really hectic for anyone who works in language-based jobs as well lawyers writers stenographers editors you name it and beyond that we are going to have a simultaneous breakdown in our ability to verify

information as there's an explosion of inauthentic things like malicious websites fake profiles romance scams and more all at the same time because why would we think this through when we can just create things and let them run wild here's another example for everyone who isn't sold yet if you start searching for I'm sorry as an AI language model it is not appropriate for me to generate

a response that goes against ethical and moral values end quote or something like that or anything even remotely close to that basically if you take one of the canned responses that says like I'm sorry I'm an AI and I can't produce unethical and harmful or illegal outputs if you take that any version of that and plug it into social media in the search bars you'll see

hundreds probably thousands of error messages from Bots who are already using an llm in their programming which makes them harder and harder to identify we don't really have the tools for that yet with the added kicker that these people are obviously trying and failing for the moment to get them to say immoral illegal or unethical things sometimes it was about bullying or I don't know like

hateful stuff it had to do with the Ukraine Russia War like all this different stuff they're currently failing at doing it but they're not going to fail forever and actually most of them aren't failing because you just have to prompt it one or two times to get around it but we're seeing evidence of these Bots infiltrating social media and starting to trip the safeguards blatantly right

now I could give hundreds of examples of AI advancements like the hustle Bros have been doing non-stop on social media but the truth is AI is just more it's more of everything more websites more images more artwork more text more research more social it's more of just about everything we can possibly imagine and we are going to be totally unable to verify what portion of that

stuff Is Real at least most of us are while the impacts of automation kick into overdrive with Untold consequences entire Industries will be re-tasked into just managing an AI making those little Corrections right because when you use chat GPT or other llms you always have to make these minor Corrections it's tedious it's time consuming it's boring it's stupid that's going to become multiple separate Industries spending

hours upon hours writing strings of random [__] words to try and tweak prompt results and I catch myself thinking why is any of this worth it yes there will be advancements that's undeniable but we're going down a path that could have a beautiful view off to one side or it could destroy us and we fall off a cliff there is a non-zero percent chance that our

entire way of life goes up in flames right now and we're still doing it because why wouldn't we the current state of AI in my unfiltered opinion is dangerous unfinished and over hyped all at the same time on one side there is very real fear fear that we will irreparably damage our society in some way or another I have that fear and on the other hand

there is greed greed to have an automated process churn out garbage and then monetize it agreed to take a shortcut and steal from others whether it be through Rewritten original content or straight up digital force and agreed to keep going because a large and expanding group of people think that they are the early adopters so they'll get rich if they keep forging ahead it's a good

thought okay I respect it they just might also wreck the entire fabric of our society in the process you can call me a Doomer and say that I'm over dramatic I'm melodramatic and I'm just exaggerating but having spent a pretty intense number of hours exploring all this in detail we should be a hell of a lot more nervous than we actually are AI should not be

pursued that's the truth but since there's no chance of stopping it I'll probably start spending my free time training I'll be working on developing my expertise as a digital authenticator of sorts so I can tell what's real and what's not that will involve a bunch of analytics and tools I'll probably do a video about error level analysis and all this different stuff but for now that's

it if you want to support their links down below merch social media locals and patreon the video sponsor etc etc and if you do support the content if you actually like the sponsors please do click that link and check it out and buy their service um search shark is one of the longest standing sponsors of the channel and I'm really looking for that now the sponsors

that Give Me Freedom so I can give them quality advertising okay that's very important to me I'm trying to make some adjustments so do consider that in a monthly subscription if you like the direction the channel is going on patreon or locals but yeah I'll cut it there and stop rambling as always thank you all for watching and have a nice night obviously I hate Ai

and we shouldn't even pursue it but I'm trying to stay informed because what other choice do I have take it easy guys have a nice night thank you [Music] [Music]