Transcription
This video is brought to you by Surf Shark. Stick around to hear more about the special offer they're providing to the entire upper echelon community.
I cannot imagine figuring out how to raise a newborn without chat GPT. That is an actual quote from one of the world's richest entrepreneurs. And today we're going to talk about why the entire product it's based on is genuine garbage. Here we go.
What is the actual purpose of chat GPT or Gemini or whatever other large language model people are choosing to use right now? What's the reason they exist? Different people give different answers. Some might say efficiency or streamlining workflow. Others will say creativity because it allows people to make things or write things they otherwise wouldn't be able to. But right now, right here for today's video, I want to make the argument that there is no purpose. I want to make the argument that large language models on a commercial consumer level at least are a solution looking for problems that largely do not exist. And because of that mindset, pretty much every single one of them will inevitably fail.
First things first, I want to play an advertisement from Open AI which is being absolutely spammed across my video feed and streaming services right now. Here it is.
Very good. What's your first impression? It's a commercial, so obviously the purpose is to entice new customers or users. But what's the actual message of that advertisement? Well, one of the more immediately obvious aspects is that there's a whole lot of words without ever actually telling you how. The text going across the screen sort of reads like a recipe, but it's not a recipe. And that becomes abundantly clear when you actually look at a real one. This right here is a creamy garlic pasta recipe. And what you see right from the very start is that there are portions, there's time frames, there's heat levels, and descriptions of precisely how to achieve an end result. But if you then go back and look at the commercial, it has precisely none of that. Stripping away the completely useless filler content, which is most of it. What you really have, I should say what you only have is the name of a dish, nothing else. What this commercial successfully communicates is that chat GPT is useful when it comes to basic ideation for you, but has absolutely zero value beyond that.
Let's do a hypothetical here. Let's assume this is a real situation, which is obviously what the advertisement is going for, relatability, and then we'll go through step by step. Boy meets girl. Boy wants to cook a meal. Instead of thinking about his own preferences, instead of thinking about what she may like, he asks chat GPT, which gives him this output. Now what? Now he has to go find an actual recipe because that's not what this is. He has to go find a real description of how to make pasta. He needs quantities, heat settings, timing, etc. And once he has all of that, then he can start buying and assembling ingredients. Let's say it goes well. He makes the pasta, the girl comes over, and then suddenly she says, "Wow, this is really good. Pasta is my favorite." to which he replies, "Haha, yeah, I know that's why I picked it." No, because he didn't pick it. He's a passenger. He's a set of hands who asked a program to do the only creative thing involved and then merely assembled that thing so that he could communicate a certain mindset to her. If you want to go one step further, he actually wasted time by using Chad GPT. If he simply chose something that he personally enjoyed or that he knew she enjoyed and skipped right to the recipe stage, it would have been faster. And before someone tries to say, well, he can also ask chat GPT for the recipe and the quantities. No, no, he cannot do that. Sadly, I have to do a bit of a summary for this part because a precise explanation would be like 6 hours long. But the basic version is that large language models will always carry the risk of hallucination. These hallucinations can be limited. They can be reduced, but they can never be fully prevented. And coincidentally, cooking or recipes is a hotbed for that type of problem. Which makes me wonder why they chose this type of commercial. If you don't have pre-existing knowledge of how recipes work, how to cook certain dishes, portion control, and what those dishes ought to taste like because of how language models intrinsically operate, you must check every single aspect of whatever recipe it gives you because the results could very well be horrendous. There's myriad examples of this directly happening with disgusting, unhealthy, or even flatly dangerous recipe outputs. But without getting too far into the weeds, this is an advertisement from a company about their own product where the product inserts itself into the decision-making process with the only effect of increasing labor and risk as it replaces basic decision-making by the user. That's not even a solution, and we're only just getting started.
On a side note, also extremely relevant, I guess, given the current political atmosphere of the world, today's video sponsor is Surf Shark. Surf Shark is a VPN, virtual private network, and VPNs come with myriad benefits, such as helping prevent certain types of fishing attacks, DNS tunneling, DOS attacks in particular, or mitigating data tracking/harvesting practices by all sorts of big tech companies. Over on the entertainment side of things, Surf Shark unlocks additional content on streaming websites by changing your geographic region. And it also gets around regional censorship. Helps protect your IP address with encryption, modification, etc. However, they also have an important new feature, which is email scam checking for anyone who's interested in that. They operate 4,500 servers in 140 different locations with a third party verified public report authenticating that they do not keep any logs, which is one of the critical policies that a VPN needs to have. Also, they maintain what's called a warrant canary. Won't matter to a lot of people, I guess, but again, very good policy to have for a service like this. If you click the link down below, surf sharkark.com/echelon, and use the code echelon at checkout, you can get four extra months completely for free with a special pricing offer. Again, link down below, surfark.com/chelon, coupon code echelon at checkout for an extra four months of the service for free. Big thank you to Surf Shark for sponsoring the channel.
Back to it. Let me show you another clip. This time from an XAI hackathon, which is all about using AI models to integrate products with movies and television shows. advertising. Basically, I'm not going to beat around the bush here. This tool will never be used. It has no value. It's a weak attempt at replacing something that already exists, which is categorically superior in every single sense of the word to what it tries to provide. It's called product placement. One of the more effective types of advertising if you ask me, and it's basically been around since the early 1900s. If you want an example, there's a scene in iroot where Will Smith gets a brand new set of Converse 2004 Allstars. And then another scene later on in the movie where they get ruined and he gets all sad about it. Not always done super well, mind you, but product placement can be integrated with the very fabric of the film, attempting to attach specific brands or consumer items to the actual characters. One of my personal favorites is a guy named Sha Spencer from a show called Psych, where they would routinely promote food products. He would snack on stuff like Snider's pretzels or get really excited about chocolate milk, things like that. And it was kind of funny. But the point I'm getting at is that the thing this AI innovation is talking about already happens. It happens better than this program can even get close to. And it happens with intense levels of paperwork. There's licensing, brand deals, actors rights, studios rights, competing companies, and all sorts of detail oriented negotiations about what gets said, what gets shown, how long, number of seconds, etc., etc. This AI tool is a worse version of things that already happen, operating in a legal minefield of compliance issues as it tries to shoehorn itself into the world for no other reason than simple AI hysteria. Well, it has AI in the name. It must be great. Long story short, it's garbage.
The list of examples like that shoehorning AI into every conceivable industry or product purely because of misguided hype could almost go on forever. But probably the most amazing one is physical wearables like the rabbit or humane AI pins. This is becoming an entire genre, by the way. Pendants, wristwatches, all sorts of stuff where the AI model just listens and records everything around it and then tries to act like a calendar or a personal secretary. And maybe one of these will end up being useful someday. I can see glasses in particular being one of the products where a decent implementation of AI could end up being useful. But the lion share of these wearables right now are trying to replace the smartphone by doing things worse, slower, and with a higher rate of errors compared to what smartphone apps already do and what a human user is very easily capable of.
Just to be clear, I'm not making the argument that AI in general is useless. I'm making the argument that on a consumer level, it's seeking problems that don't exist. On an enterprise level, big companies, data heavy tasks, that kind of thing. On an enterprise level, AI probably will have value someday. Even at a consumer level as an augmentation inside existing software, it's begun serving a purpose. A great example would be in Photoshop where tools like contentaware fill or generative fill can be super convenient for editing photographs. But that's not new. It's just enhancement. Photoshop already existed. It already had less advanced versions of those very same functions before this for many years, which makes AI just a bit of frosting on the cake, if you will. It's not the batter. Even at an enterprise level right now, according to MIT research, 95% of companies piloting or integrating AI are getting zero return. According to a prominent model evaluation and threat research MER paper, the value of AI models among coders was actually negative, which is insane because coding is often regarded as the most promising test case right now.
What I mean to say by all this is that the primary value that these programs really have, the most immediate, easily accessible and freestanding purpose that they serve is actually deception. Deep fakes, misinformation, illicit email generation, bot farms, ad fraud, cyber security risks. All of these things are exponentially easier and more profitable because of AI. But these aren't consumer level solutions. They're problems. Chad GPT in particular and similar large language model programs as well are seeking to be used for all sorts of novel purposes, but the common thread is that they want to do the thinking for you. Open AI has been given more resources in a shorter period of time than any other company in human history. And yet what they've managed to produce is not helping people. Contrary to their corporate marketing dril, it's replacing the human aspect of thinking and decision-m while detrimentally affecting pretty much the entire internet in the process.
I'll go one step further, okay? It's not even just a solution looking for a problem. It's a problem demanding its own solutions. The negative impacts of large language models will persist aggressively, but the positive counterbalance on the scales that they often use in marketing concepts and commercials simply has not materialized. Open AAI is now maneuvering to insert advertising inside Chad GBT, which is absolutely not the way to increase adoption. Pretty sure it's just pure desperation at this point since they're hemorrhaging billions of dollars per month. But even if they somehow stabilize the company, the product itself can never be what they want it to be.
Circling back to the original advertising, chat GPT is being held up as an everything tool. Ask it when you're confused. Ask it if you don't know. Ask it for advice and whatever else you can possibly think of in between. But the only purpose it really serves is the replacement of basic decision-making with the most homogenized answers possible because it's an amalgamation of everything it's scraped for data training. If you want a particularly stunning example of this, Sam Alman, Open AI founder, recently went on Jimmy Kimmel and literally said, "I cannot imagine figuring out how to raise a newborn without chat GPT." And just to be clear, the reason he feels bad about that type of question from the context of the clip is not because he feels bad about asking an AI how to be a [ __ ] parent. It's because he thinks the AI is quote genius level at everything intelligence sitting there waiting to unravel the mysteries of humanity. End quote. I'm not even joking. Okay, here's the context. Check it out.
>> And do you use chat GBT when when raising your baby?
>> I I I do. I I I mean I feel kind of bad about it cuz I we have this like genius level at everything intelligence sitting there like waiting to unravel the mysteries of humanity and I'm like why does my kid stop dropping his pizza on the floor and laughing like >> you know and so I feel like I'm not asking a good enough question but it is I I don't I cannot imagine having gone through that like figuring out how to raise a newborn with it chat clearly people did it for a long time no problem. Yes, but
>> sadly chat GPT and many of these other programs will continue existing. Cats out of the bag on that one. But despite the persistent danger or the inherent risks associated with unchecked development of these programs and then implementing them all over a social and corporate landscape, I find myself incredibly hopeful when it comes to widespread user adoption. Because despite what these companies will say, they're not everything tools. They're functionally closer to nothing tools at this point with an extremely limited scope of value. Not only that, which I've talked about in other videos, but sustained and frequent use of large language models in essay writing according to MIT Media Lab directly led to reduced brain function as well as EEG activity. So, even if my arguments seem like the epitome of old man shouting at clouds or a lite argument, I guess, consider the fact that you may very well [ __ ] mentally [ __ ] yourself by overly relying on these programs for day-to-day tasks.
Most, if not all, major language models right now are destined to fail. They will not make money. They will not replace our existing devices like smartphones, and the hype-driven promises will mostly fall apart or be forgotten. However, sadly, we are most likely stuck with all of the downsides. The slop, the spam, the degradation of trust online and the knowledge that anyone or anything could be fabricated isn't really going anywhere. Pandora's box doesn't close. But in the end, I find myself hopeful. Witnessing the birth of a product when the product's only purpose is to supplant human thinking and then watching that product rightfully fall into the gutter soon afterward. Unexpectedly, I might add, at the beginning, I was like, "Oh no, this is going to be a real persistent ongoing problem." But it's I don't know. It's oddly satisfying. And I feel good about seeing people largely seem to reject ChatGpt. That's awesome to see. And what I believe is happening right now is that chat GPT and Open AI are failing. That's it.
If you want to support the channel, check out the links down below. The video sponsor, Surf Shark, with a special deal, of course, locals and Patreon, channel memberships, and more. But I'll cut it there and stop rambling. As always, thank you all for watching. Question everything and have a nice night. Heat. Heat. N.