Yeah we all hear the main arguments… AI is bad because of slop content, stealing from creators, brain rot & brain damage, privacy concerns and most importantly… how billionaires are just using it for their own selfish reasons
But I’m asking about YOU 🫵 personally. The individual. What do you really think about AI? Do you care or are you indifferent? Has it actually affected your day to day life?
Depends what you mean by AI. Because for the most part “AI” nowadays is just a pure marketing term. What the tech company marketing calls “AI” is not any kind of intelligence in the actual sense of the term. It’s just some machine learning algorithms and fancier CGI. We had “AI” back in the 90s too. Every computer game where the computer player makes autonomous decisions is a kind of AI. It’s the marketing that is the problem. The idea that it is some kind of revolutionary technology that can replace human labor. It’s not and it won’t be. It’s just another tool. Unfortunately too many people have fallen for the marketing hype, which has served to inflate the AI bubble that will pop sooner or later.
I’ve messed with neural nets over a decade ago and ever since I realized their limitations back then I’ve lost all interest.
It can never outgrow its own boundaries. If you train one too much then its output will just be the literal training data with some error, but never anything better, no matter how much data you throw at it.
That alone has always made the ideas that it can “create” things or “think”/“solve problems” absolutely ridiculous to me.
I’ve said this before on lemmy and have gotten downvotes from it, but neural nets are much closer to something like an ASIC.
And now there actually is a company that literally makes ASIC chips that implement neural nets! ;DI’m not denying that neural nets do have genuine uses, but the way most are being used in the 2020s just grosses me out.
The idea that it is some kind of revolutionary technology that can replace human labor.
Now people say that but 10+ years ago nobody cared. I had 2 chatbots talking to each other about random things that got picked from the Internet as an experiment all running 24/7 on a Pentium D PC, and whenever I showed this to someone they responded with a “uh cool.” Now these same people think they’re interacting with a “thinking” being that is capable of doing their job whenever they engage with some LLM-based bot, when in reality it’s not too different from what I had back then at its core.
It’s a tool, and as such the class it serves depends on the mode of production and the class in power. It has some use cases, but it isn’t the supertool techies think it is. It also isn’t utterly worthless like some believe. Over time it will likely become more useful and better integrated.
Depends on the game. seriously though. LLMs are not AI. They are not intelligent and they are not artificial. They are just a brute force algorithm to decode and respond in natural human language.
I often use deepseek instead of using search engines but only because SEO and LLM generated websites have ruined search results. I have used it to make a few simple programs which is great because I have no clue how to program.
Some people are using these tools for stupid shit but people always use tools for stupid shit. Atleast all the hype is going to break usa’s economy and maybe things will get bad enough that usaians will finally deal with their fascist overlords and become a normal country.
The USA cannot ever become a normal country. Settler colonies shall always be haunted by their birth.
This pretty much sums my thoughts.
I think it’s a really useful tool that’s made my life easier and allows me to explore a lot of ideas I was just too lazy to do before. I have like nearly a decade worth of half baked software project ideas, and I just never had the energy to work on them or finish ones I started. With LLMs, I can actually get them working to the point where I can see the idea in action which is really enjoyable for me. It’s also made my work easier where I can focus more on things I find interesting and delegate tedious tasks to the agent.
I do look forward to a time where we can run these tools entirely locally though. I do not like being dependent on company services or sending my data to them. And in general I see this as the real negative aspect of how this technology is being developed. We don’t want to end up in a situation where tools we rely on day to day are owned by a handful of corporations. Regular people need to own the means of production in the digital realm. Currently, anybody with a computer can do any type of digital work be it writing documents, design, programming, etc. But if we start relying on LLMs as a core part of our workflow, then that tool also needs to be run locally or we end up as digital serfs.
They are basically summed up here: https://en.prolewiki.org/wiki/Essay:Intellectual_property_in_the_times_of_AI
First and foremost, I think it is egregiously misnamed. It is neither artificial nor intelligent; it is just math. There was a time when “knowledge based systems” was a popular moniker (albeit for a different approach) and I find it to be much more apt for the systems we call (generative) AI. If we’re going to repurpose a term, I think that one is more suitable. “Large language model” is good for its accuracy, but is also less evocative and descriptive, especially for the normies. Either way, it’s wild to me that we’re like “yeah AI is here now.” It’s frustrating to me because of how it impacts the way we interpret and use these systems.
Secondly, I think the way that Statesian companies are building these systems is exactly wrong, but I don’t suppose that should really be a surprise to anyone. The throwing-peas-at-the-wall and throwing-money-and-resources-at-it approach has netted results, sure, but wouldn’t it be neat if everyone was working together to more deliberately collate the entirety of human knowledge and create accessible tools for all to leverage? You know, instead of letting private companies extract the fruits of our labor, throw it into their equation, and then sell it back to us, over and over again? Anyway.
Outside of that, I think Cowbee’s succinct take reflects my view as well.
Ultimately, it is a tool. It’s impressive that hardware has developed to the point where we can throw so much language at these systems to get useful results. It is also true that these systems are both over- and underestimated. I think it’s also true that the current economic approach is intractable, and I look forward to the day when we more broadly understand how to build and use these tools more effectively.
The terminology can definitely be misleading. AI evokes anything ranging from a pathfinding algorithm to sci-fi sapient machine that takes over the world.
I think it can be accurately said that AI as in Artificial Intelligence is the end goal of the machine learning field, but it gets fuzzy fast on definitions whether the field has actually done that in any capacity.
Partly I guess because the concept of intelligence in the first place is largely a way of thinking about humans, not machines. Is a light “intelligent” if a sensor can detect when somebody within range and then the sensor triggers the light to turn off? It’s doing something that is useful, but it doesn’t know what it’s doing as a separate consciousness. I think it can be argued that what gets called generative AI is similar to that, but with a lot more complexity to the inference operations.
I would say the mistake is in thinking that if a tool becomes sufficiently complex, it is necessarily heading toward something distinct like what humans have as consciousness. But this is not taking into account form. Humans have a very specific biological form and if you simulate aspects of that form in a machine, you haven’t now recreated consciousness; you have created an advanced simulation of one or more facets of human-like cognition or processes. This can still have benefits. A blueprint for a building constructed from computed simulation could probably have use to an architect, even though it’s not the real building created yet.
So perhaps something like Simulated Cognition would be more appropriate for most of what gen “AI” is, in practice.
I agree with your points and perspective, but I also fell like “Simulated Cognition” is a bit too generous. I don’t think an LLM/what we currently have as generative “AI” is a simulation of cognition, though I acknowledge/concur that is the intent. Perhaps I’m splitting hairs too finely, but I see it instead as a statistical approximation of language processing.
I mean, I guess one could just say, “yeah, they’re a statistical approximation of language processing with the intent of simulating cognition”, and I’d have to acquiesce. So I guess my hang up hinges on how one interprets the word “simulated,” because I think its connotation tends to be more weighty than its literal definition. For example, if we said “Mock Cognition,” that’s more obviously fake cognition (to me, anyway). Whereas a mathematical simulation of something, for instance the flight trajectory of a satellite or rocket, is not the real thing, but is more or less expected to exactly model the real thing (at least in my selected example). And it makes me uncomfortable to apply that perception to the “Simulated Cognition” of our models that approximate language processing.
That’s fair. I’m definitely not married to either term. Mainly trying to work out something that is more accurate.
I will say, the reason I go for “simulated” is because for me, the connotation I think of is video game style simulation, i.e. something that is understood to be not real. But that may not be the takeaway most would have.
Either way, I get the concern of not overstating what gen AI is doing. Though on the other hand, I think it’s important not to understate it either. Like what models are doing now with complex code, or with reasoning layers, it seems almost trivializing to call it statistics, even if that is a component part of it.
We could also call them Bullshitting Machines, haha. They sure act like that sometimes. But yeah, I’m open to better ideas on better terminology for it. Precise terminology has never been my strongest area. I’m more apt to use language fluidly.
Virtual Cognition? 🤔 No, I don’t think we’re going to come up with anything better than Bullshitting Machines
I went through a phase where I was doing a lot of vibe coding. I thought that I was smart enough to be working around the known problems and limitations of it. I was not. Pretty much everything “I” had made turned into an unmaintainable nightmare once it passed a certain level of complexity, and in practice when you’re trying your best to actually make vibe coding work on a project larger than one file you spend so much time reviewing every little thing, rewording your specs etc that you might as well just write the fukken code yourself.
So I deleted all of those projects, and now I use AI for brainstorming (I talk into speech to text, usually while doing something else, then give the text to an LLM and say “organize these ramblings into a design document for me”) and simple specific task automation (“comment my code” and “update my readme”). I figure the one thing LLMs are demonstrably good at is summarization, so pretty much everything I ask them to do is along those lines.
That said, if I could press a button that would make all of the ai companies crash and burn but it meant that I would have to back to doing those things manually, I would press it in a heartbeat. The externalities that the tech world is forcing into all of us for this thing that makes summarizing documents more convenient are way too much.
As for anything else, generative images or video or whatever… I just have zero interest in making or consuming them. Sure it hits my feed like it does for everyone else but I’d say at this point my reaction is about 1/10 “haha” and 9/10 “ugh” when I see them. There was already so much human made art out there that you could be looking at new things all day every day and you wouldn’t be able to keep up.
edit: also, regarding the coding: if your goal isn’t “learn to code” but rather “make a quick python script that needs to work exactly one time”, then vibe coding can be a valid use case. but if you are trying to learn then prompting cannot teach you in the same way that you will never learn to make high quality digital art by prompting for image generator.
I find it’s perfectly possible to write large maintainable projects using these tools. I have a Rust project I built with LLMs that’s over 150k loc now, and it’s structure is a lot better than anything I would’ve ended up on my own. One of the things I do is ask the LLM to come up with a phased plan for introducing new features. I also ask the model to make mermaidjs diagrams I can inspect and come up with file layout up front. Then I get it to make a branch for each phase and implement a focused feature. Then I can review it and I have good context for what it’s supposed to be doing, and it’s scoped so that the code is manageable enough to fit in my head. Doing that alone gets you a long way. Another thing I do is ask it to make refactors by looking through the code base and finding repeating patterns in code that can be consolidated, or large files that need to be split up. If you do this regularly, you end up with much cleaner code, and the agent is much better at doing that at scale than you could by hand.
Testing is another really important aspect. I always ask the model to do TDD, and then add end to end integration tests for features. For web apps, using playwright storybooks is really effective. You can define exactly what the user workflow is and then have the model test it through a headless browser end to end. This creates a contract where you know what that functionality is actually working end to end.
As long as you don’t just let the model crap out tons of code unsupervised, and box it in sufficiently with a contract, then the code is no worse than what a human would produce. And I’d argue that it’s often better.
Openspec is along these lines and works well ime
Yeah it’s a nice tool, you really just need a bit of rails to keep the model on track I find. They’re good at doing well defined tasks, and if you have a plan where they just check steps off as they go, things tend to work well. I also found beads is a really handy tool for task tracking cause then you can just file stuff in there and instead of writing tasks in markdown you have a real history along with the status of the tasks. It solves the problem with markdown checklists getting stale.
good to know, I’ll check out beads. I was forced to start using Claude at work and had mixed impressions but once we started using openspec to keep it on rails as you say, it got hard to deny its capabilities.
I find I kind of look at the whole LLM + agentic harness setup as a genetic algorithm. Your tests and specs are the fitness function for the program you’re evolving, and the LLM is the mutator. At each step it generates some output, it gets tested against the fitness function, the LLM gets feedback and iterates on it. Eventually something working falls out in the end. The better you can define the selection criteria the more you box the agent in the better results you get.
I dont like gen AI because its really really bad for the environment
I remain critical of their use cases and environmental impact but I am not opposed to it.
The consensus we reached with our party is that it is better to understand it and know how to use it because our enemy, the ruling class, is in control of it and is also using it. The party made the mistake with the rise of the internet to pass it off as some hype and years later when internet was widespread and common in use, they were behind on their knowledge and missed to boat. They won’t let it happen again.
I don’t really run into a lot of the issues I have heard of people having with AI since I primarily use DeepSeek as an actual search engine. It’s odd actually, I am “pro-AI” but have never generated a single image with AI. I just never felt like it, ever. I insulated myself from places Instagram slop reposts can find me, long ago, for completely non-AI reasons lol. It’s like I don’t care about anything other than text, and I want to handle it all myself, so AI just acts as a targeting reticule. I didn’t even notice DeepSeek is text-only for months.
Tried adding AI summaries to articles I post, but I just didn’t see the point after a day-and-a-half. I still read them all anyways and would proofread the summaries. Do people just slap those on because they can? Journalistic writing already has a pyramidal structure, summaries summarize the summary at the start, & then you read… the second summary. Why?
Using Kimi to edit entire Orgmode (task management) notebooks is pretty dope. It’s too scary though, what if it loses something? Generating scripts is cool too, but I need to learn the scripting languages, it’s not hard for what I need to do, so why put myself in a position of being unable to debug? Will save work later though. Barely scratched the surface of this, a few weeks of random stabs at it when I have spare time.
I have a lot of projects in mind for local + metered + talking to phone (apps like Tasker + OffGrid) stuff, been feeling it out. I just want to be able to find book quotes without the precise phrasing, to extract key points from books people send me to find where they detail things related to whatever supporting arguments I was presented with, meta-analysis of citations (did this book primarily cite western news articles and high-falutin (yes this is a gabe rockhill reference nobody else says that) academies?)
Deepseek is very useful for projects like “hey how do I avoid reinventing the wheel with my homelab setup, i want to experiment with Deepseek” 🤣
So, not a ton. The robotics and computational engineering models are much more impressive, no?
Using Kimi to edit entire Orgmode (task management) notebooks is pretty dope. It’s too scary though, what if it loses something?
For this, agentic would be able to make scripts that test the data integrity and make sure nothing is missing in various ways. Simple enough to run Python
That sounds good, I’ll give it a go in a separate note space before considering merging still 👀
I find by themselves the models, especially current-gen ones, are pretty bad at editing text. They still don’t really grasp what it entails lol, because they are not aware of their limitations. And it seems that current models are trained mainly for technical (coding) tasks over anything else, so I feel it’s only going to get worse in those applications.
But I’ve had some success using a test suite afterwards to confirm data integrity. Counting lines is one such method: you just compare the number of lines between the before and after and it gives you an idea of how much was cut off, but it’s basic. An LLM in agentic can set up a full test suite to really understand what changed or not statistically, and then is able to bring back stuff from the older revision to ensure integrity and that it didn’t do too much. There’s a lot of other things it can use to test the data, and you can ask it for cross-tests too: two different tests that test the same thing, but do it in two completely different ways (like calculating “x*x” and then “x^2”).
For coding I find LLMs to be legitimately revolutionary. I’ve tried letting DeepSeek & some local models like Qwen and Gemma loose on various projects to implement features and improvements for local use and so far most of the time it didn’t disappoint.
In the last couple of weeks I’ve been updating an old third party bot plugin for a game by prompting various behavior changes I’d like to see and it’s a night & day difference to its original state. If I had done this by hand the time it took would’ve been multiple magnitudes longer, it’d have been more error-prone (especially since it’s a C++ project written in classical C style, which is just UB galore) and I likely would’ve lost the enthusiasm to work on it by now.
It’s a tool to shorten research time with the result potentially having a margin of error. It means you need some sort of capable mechanism of error checking, which often means either being a specialist in thing you are developing/investigating or having some kind of external reference that act as that for you.
I’m honestly shocked how good the Claude models are for development.
I personally know a guy that’s losing his mind from Ai psychosis. Really, it’s probably just mostly reinforced the latent narcissistic personality but it’s still wild.
So far, mixed bag with all kinds of signs saying shit is going to get weirder and worse, partly as a consequence of this tech.
I’m frustrated by LLMs. I can see the use cases, I understand why western LLMs are the way they are (and eastern ones as well). I just get so annoyed with companies pushing AI entshittification, people using it for stupid tasks, assuming the answers are correct when they’re not, people making themselves dumb because they don’t think or do anything digital for themselves anymore, searching the web for something and only finding AI generated SEO slop, ignoring the billionaire and anti-worker interests behind the LLMs… but I know these are all functions of the relations of productions and the way western capitalist society works.
It’s lazy to use it, inefficient, lacks creativity, leads to inhuman marketing slop, and serves only to extract wealth for those who own the models by cutting out labor. I lose respect for anyone that uses it.
I work in a creative field, and AI doesn’t threaten my job, but that doesn’t mean companies aren’t trying, and it doesn’t mean that companies aren’t using money that would go to people in my profession to put towards AI budgets.
