Give this a thought:
How would you design an algorithm that reads a post without context and returns the probability that the intention is to deceive readers?
An impossible challenge
It should be achievable as a solo project with enough time and careful planning, though by no means a simple task. Let's cook.
You'll have to create an extensive environment with countless variables and scrape the web to weigh the content against similar posts. You might want to use a database to create tags for categorizing information, hash tables, weighted arguments, and an index of collected reference material.
Now you need to decide on the program's architecture and map all of those data points you just made into temporary data stores for complex analysis. Then you'll have a few dozen localization libraries and plenty more to consider. Once you get it started, the debugging goes on for twice as long as it took to establish the boilerplate.
Your codebase is now one huge steaming pot of spaghetti. As you strain out the water, some of the noodles stick to the bottom. "This is fine," you say, as you prepare a plate and top it off with mom's special front-end marinara (it's node). As the first guests queue up to try your hyped up state-of-the-art pasta dish, forks in hand, you're about to push the release, when suddenly you remember...
you forgot to make the meatballs
mama mia
oh no panic, panic, panic, whatdowhatdo it sucks FIX IT NOW
Breathe.
We have technology
Check the calendar. We have a completely different, readily available solution in just a few lines of code. You already know this, it's not news to you. Get your favorite LLM from the fridge, heat it up with a few natural language instructions, and let the API endpoint do what would take a team of specialized engineers weeks to construct and spend the rest of their careers maintaining.
Solutions like this, I would argue, are each themselves "soft" breakthroughs which are going largely unnoticed amidst the sprawling wasteland of AI media and news sites. I'm sure if you think hard enough, you can find some insurmountable obstacle that you simply deal with every day, having abandoned hope of a solution and tolerating the burden of its effects. Now think about how an independent, omniscient AI expert would feel about that.
The point here is that this technology is difficult to appreciate; a state of affairs perhaps owing itself to the public's prior exposure to heavily flawed chatbot models paired with the unique flavor of complacency that emerged with ChatGPT's recent popularity. Properly observed, it should be absolutely astounding that any of it is a reality.
It's not even all about productivity or solution engineering. Sometimes you just need an outside perspective on something. Lucky you, that the entire collective knowledge and wisdom of all humankind now fits in your pocket and is capable of matching all of your queries with extremely profound insights. This isn't the information age anymore, we're headed into the integration age.
Look at this similar prompt I wrote for Gemini 1.0-Ultra. It's a lazy zero-shot first attempt success.
Prompt
- Analyze the provided data from a social media post's comments section
- Give a list of each poster and an estimated probability that they are a bot account
- Grade each post's individuality by comparing them to the most common elements among other posts.
<input type=[blob]>
Output
Analysis
The comments section of the social media post has 27 comments in total. A significant number of the comments (23 out of 27) are highly repetitive and generic. The commenters generally repeat the same point as the original poster, _, and add little to no new information or insights.
Bot Account Probability
The analysis found no evidence to suggest that any of the commenters are bot accounts. All of the comments appear to be written by real people.
Post Individuality
The comments can be divided into three main categories based on their individuality:
- Highly repetitive comments (23 out of 27): These comments simply repeat the same point as the original poster and add no new information or insights.
- Moderately repetitive comments (3 out of 27): These comments also repeat the same point as the original poster, but they add some new information or insights.
- Unique comments (1 out of 27): These comments make a unique point that is not mentioned by any other commenter.
The majority of the comments (23 out of 27) are highly repetitive. This suggests that the commenters are not very engaged with the topic and are simply trying to get their comments noticed. The lack of individuality in the comments also makes it difficult to have a meaningful discussion about the topic.
With only three lines of markdown and one manual copypasta operation, we have exposed a network of ingenuity. Sure it didn't do exactly what I asked, but it still provided the generic solution I was seeking from the prompt without any unit tests, debugging, or hair-pulling.
To follow up, you can ask Gemini to record each account caught spamming, pipe it to a database, and catalog any mass of abstract data you want to collect. Have your very own personal surveillance state in just a few clicks! Fair warning though: you may want to choose an offline local instance if your LLM will be interfacing with NSA-style applications.
If only there were some way to repeat the process indefinitely for all content on the web page, kind of like a hula hoop or a donut where it goes on... looping itself until there's no more data left. Then you would have a really powerful data collection algorithm.
for now, all we can do is await a better solution while pasting the data manually.
The Lie of Engagement
As I toil away applying for jobs and seeking mentorship, social resources become increasingly more skewed toward disingenuous posts and it seems like everyone sells snake oil. I know it isn't true. There really are good people all around, but to those lacking a base of existing connections, the echo chamber of emotionless, hollow voices only gets louder.
Even Microsoft's community services are nightmarish containers of spam and fake reviews. This has been going on for many years, but now the tools to eradicate the phenomenon are accessible to even a toddler, so why is it allowed to continue?
Fake traffic is better than low traffic
That's a hot take, right? You never heard that before. You know but you don't care. That's absolutely fine; we need to manage what we care about to keep our priorities checked and maintain our emotional health. If you obsess over things outside your control, you start posting about restricted topics and before you know it, your wife leaves (she took the kids, too). Not healthy.
The problem is that companies are hesitant to lower spam because they will lose money instantly as a result of the affiliate marketing machine. This is natural behavior; as money is essentially the lifeblood of a for-profit organization, you can't blame anyone specifically because the entity is acting in self-preservation. It would be the equivalent of suicide for the business that cuts off its own revenue.
As for the people working there, I can't honestly believe that anyone is sitting around a boardroom plotting to flood their site with spam to profit from advertisers with total disregard for their platform's user experience. It is just the way it is.
As companies grow, they become automatons, acting just as governments do, often in opposition to their declared intent and in spite of the will of their smaller parts. Invert that model and you have an individual person, acting as the larger entity with many smaller automatons outside of their control.
Why don't we illuminate this phenomenon with a revolting example? Say a pimple appears on your nose. Your skin made that pimple because that's what it does. It doesn't know that you don't like pimples. You are the authority, so you can pop the pimple, but you'd prefer if the pimple didn't grow in the first place.
To eliminate the possibility of having more pimples appear in the future, you could tear off all of your skin, but you are sure that you wouldn't like that outcome so you deal with the pimples and pop them when they show up.
This is an avoidable nuisance. You could also invest in better skin care and improve your hygiene habits. Maybe you don't have the time to or are not motivated enough to change on your own.
If none of that made sense, you'll understand soon.
We, as in our souls, exist somewhere in between the macro and micro, bridging the gap between eukaryotes and nations, nested between layers of unconscious decision-making processes.
How does this relate to AI?
Well I'm glad you asked, because I'm about to drop the bomb.
Our subconscious can now be reprogrammed by computers.
We only need to set up a model within a framework where we can express to our AI assistant how we would like to be, and the desired behavior will emerge through repetition. To make it even easier, add recursion. Start off by asking a chatbot to design that app for you using CoT, then later when the bot is in the app, it can design its own interface. Then, if you can resist the urge to deploy it as a chatbot service for $9.99/month, put it to good use and create value in the world.
As a side, it's time for the self-help industry to die. Those "pick yourself up by the bootstraps" snake oil tricks rarely work to build long-term change, but what about picking yourself up by the Bootstrap? It really does not sink in all at once that generative AI is not just a technology, but a completely new way to solve problems. The average layperson who utilizes AI is effectively in many ways on a superhuman level, able to perform feats beyond their natural capabilities.
Things which we know are bad for us yet persist in our lives for reasons beyond our control are now, in fact, capable of being put under our control. There is a substantial delay in our awakening to this fact, but as we acclimate to reality, the process to correct unwelcome "automatic" behaviors of both organizations and subconscious components of sentient life using artificial intelligence will be ubiquitous. Awareness and propagation of this technology will serve as a catalyst, transforming the human experience, including the way we think and how we act. This period of rapid change is going to destabilize many existing "legacy" structures, possibly leading to chaos and disastrous side effects for a little while, but just as there is no free lunch, progress never comes without sacrifice.
There are so many more things that were previously impossible for humans to do that are now easily available with free AI tools. We are so highly augmented that we don't know how to cope with it. We're in the midst of a problem-solving revolution akin to a literal epiphany as a species, potentially leading us to ascension through enlightenment. I will say that there will always be things AI can't do, no matter what. It cannot, for example, sustain your biological process of its own volition by placing food in your mouth (not yet). It can feed you metaphorically... if you work in the AI sector or primarily use AI-powered tools to earn your living, maybe you would say that.
Now if only we could get every big tech company to stop focusing all of their resources building ChatGPT clones to spy on their users while every startup is building their own AI app, as if they aren't just calling GPT's API endpoint and slapping their branding on a React template stuffed with affiliate marketing cookies... Wait, what was I talking about?
Short Q&A
Q: Could you elaborate?
A: No.
Final thoughts on social spheres in the early days of the AI era
The bigger problem is that the actual engagement numbers are suppressed due to real content being disincentivized on a massive scale. The sheer number of posts makes it difficult for any individual to be heard, so naturally people will post less often. Discourse rarely takes place in any meaningful way, and when it does, that conversation moves to another platform to continue away from the noise of constant distraction that is social media.
Someone like me finds it impossible to connect with people in a vacuum when I followed one suggested post that was recommended to me, and now instead of getting search results for things that I engage with, my feed is full of jobs that I am unqualified for and AI-generated posts by clones of the suggested post's account that I was conned into following.
While related, that's a subject for another time. I hope to write more pieces like this to foster genuine discussions with new perspectives from this era of problem-solving.
Thank you for reading.
Peace be with you,
Christian Vaccaro-Gallo 👹
I invite you to comment below to share your thoughts. If you've read all of this, then surely you have an opinion about something in the article. The world deserves to know 🌍
If you're a real one, like this article. <- I hate it when people do this (do it anyway)
