I’ve been very fed up with AI for about past 6 months, completely started to boycott it, discuss it with people of all kinds of different views, and I found joy and pride once again in my own work.
But the world issues coming from AI, and even more so from the billionaires and empires behind it seem to further pile up to insane heights. I’ve been trying to learn more and more about it, and after my bachelor and masters I am considering pursuing phd and research surrounding AI, especially from the critical perspective, which seems to be deeply neglected in the research pov.
This is still a few years in the future, and lot could change, but I am curious what do people here think about pursuing such a thing, and if in current academic world it is even something that would be possible doing, given lot of the grants and funding of AI research comes from these companies that just want to gain even more power through it.
As I said, I am already trying to know as much as possible about it, but I would like to look more deeply into its impact on society, impact on students etc.
Do you think this would be a worthwile endevaour? And if not, where do you think I should be heading to make change about this while not completely starving to death?
Specializing in a bubble is probably not wise. Writing about it and studying it as a tangential topic to your primary area of research is probably better. Philosophy, language, etc.
I wouldn’t say necessary that I want to specialise in this bubble, I mean master degrees specifically would be more broad than just “LLM” anyways, and I feel like with phd you have to choose some fairly specific field of research to study anyways… But maybe putting it better, I would like to go and research these Machine learning technologies in general, and while relevant focus on LLMs, but keep myself updated and educated on other topics too.
Deep understanding of LLMs requires you to deeply understand many Machine Learning techniques anyways, so even when the bubble bursts I feel like it would still be a valuable expertise to have.
But currently I am studying cybersecurity, so in the end I might decide to stay in this field, and as you say, write about this and it’s relevancies to the main field I would be in…
Woah! This kinda speaks to my soul :D Feel like this is a resource I was searching for for a while! Thank you
I mean, ai isn’t inherently bad. It’s more of an issue how the big companies do push it.
Ai in research is phenomenal, especially in medicine applications. It’s not a black & white issue.
And we are just at the beginning of it, best of luck! U’d also need to differntiate between LLM(general use cases) and task/research specific ai.
Oh yeah, sorry for not specifying, I would like to critically study LLMs, and IF it gets more reasonable and susteinable, I would consider some of these other, more benefical AI uses.
Yeah, it’s really important to specify, what cases because they aren’t the same at all.
There’s even some really good uses cases for llms in companies. With a declining demographic most European countries face, a goal for company could f.e. be being more efficient, since (f.e. the company I worked for) 30-40% of staff will be in retirement age in the next 10 years. And if you work with documents (what we did) there’s a real benefit of llms classifying and compressing these documents. We speak of 10s of thousands a day. And the now used systems lack in flexibility to reliably classify or even read some of those documents. On top of that, you don’t need a 200B+ model for those tasks.
But that’s the good side in my eyes.
There’s loads more of problematic and socio economic issues with those models. Especially revolving around how people learn, decide and interact with each other.
You’re diving into a really broad field here and you’ll have to pick out very specific cases. It is for sure a super interesting field.
And on top of that, it’s a really old computer science field, dating back to the 60ies. It just now comes to “fruition” since our tech advanced so much that we can actually process these stupid amounts of data.
Before open ai & others popped up this all was labeled under computer linguistics & Data science, which just doesn’t sound as sexy I guess.
I think these days I just associate the term AI with LLMs or genAI in general, and I prefer “Machine learning” for the other models. I also believe, that for working with documents and data, we should look into other methods and ways instead of LLMs, since these models are usually extremelly “bloated” and just expensive and unsustainable to train and run. I might not have enough experience in the field “yet”, but I do believe there are other ways to go about it for the benefical tasks. And I won’t even talk about the data gathering methods that were used for these LLMs, which are not transparent at all, and mostly involve stolen work… That could be another topic worthy of research for sure.
I think if you open up a study, you should and probably need to be specific with the terms. Since llms are just large machine learning models. Just not trained for a single specific use case. You can also achieve very impressive results with small models, you don’t need chatgpt 5 for document classification. You can also fine-tune these models for specific tasks and/or “lobotomize” them. But f.e. go with a small qwen model with just 36B parameters or less and you will get very good results. And sure there are the good old OCR methods but you’ll need a significant pipeline behind a classic ocr machine. And it would probably still fail to decipher/classify a machine written document with hand written annotations. When you use a decent LLM, it will in most cases be able to differentiate between handwriting & machine letters, it will be able to output both in different variables and it might even be able to put the annotations in context to the original document. And this is an enormous task to program by hand.
And when we talk about speed and sustainability, not every document would be thrown at the expensive model first. But you would build a layered approach, so that 95% of the easy documents would be handled by a cheap and fast solution, but when that has a low confidence, then you would hand the document over to the bigger slower model.
Then add graphs or tables to the document and you’ll be nearly completely lost with a classic approach.
I’ve been working in this field for a couple years, so I speak from personal experience.
But still all those models still have an issue with context sizes and you and your business pipeline will fail if you don’t know the boundaries of what’s possible today. For the most high profile cases there should always be a human in the loop. Do companies do that? Most likely not, but they can get in big trouble if they make a critical mistake, at least in Europe, can’t speak for the wild West/US.
Note: You can self host qwen3.6 with 32gb or better 64gb and play it. It is shockingly good.
Data gathering and theft of IP is a completely different topic. But “luckily” many people now upload their data for free, directly to one of the big hosting companies. But privacy is also a different topic.
So again, be very specific if you choose your topic.
Of course yeah, if I would be writing actual research I would be more concious of the terms used. I see you have a lot of experience and have been using LLMs a lot and searching for ways to use them conciously, and that’s kind of part of the reason I would like to study them (and Machine Learning in general) more deeply. You seem to have the practical experience, but I would also like to personally look into it with the academic view, and see how these arguments really stand and how to improve the actual uses, while pushing against the bogus uses and claims that many people and mostly companies keep telling us. There’s still many years until then, maybe I’ll choose a different path to push against companies and the bullshit we are dealing with today… But I think this field might be something I am genuinely interested, and if I may actually do some worthwhile change during it, it would be a nice bonus…
I’d say most people who successfully run llms and do use it consciously are actual “AI-engineers” with a proper computer science background.
The people i’ve met who shout the loudest about their ai use are usually nothing burgers or management positions who didn’t come from an IT background and just see the golden goose “Ai” and want a piece of it.
If you want to get in touch with people having experience go and look for meetups :)
No. You should not become a doctor of chatbot slop
No. You should not pursue a speculation bubble as a career. You will regret it.
Slop doctor goes hard tho ngl
This article is imho pretty good: https://nicholas.carlini.com/writing/2025/are-llms-worth-it.html
It’s long but speaks about research in the end. This may be interesting for you
Thank you! I’ll give it a read later :) I do like to see both sides of the coin, as it gives me the opportunity to think more critically about the technology. Of course, Lemmy is sort of this bubble that is just completely against LLMs, which has deeply impacted my views on it too, and also seeing how dependent and completely ignorant many peers in the software engineering/cybersecurity studies are on this technology has pushed me even further against this technology. This is the main reason I would like to further study it, since having deep technical understanding of it could give me the options to fight against it and shape it in a way that could genuinely benefit others, or at least give me a way to educate others about the real implications of LLMs. We’ll see what I’ll take away from reading this article :D
I think it’s probably a good idea. Are you already pursuing a bachelor and if so in what field?
There are a lot of angles to critique LLMs and generative AI from. Economic, social, environmental, psychological and technical. Do you want to study how investors and managers are duped into investing massive amounts of money into a huge bubble? How society is being ruined by slop? How huge data centers are destroying the environment? How sycophantic chatbots are driving people into delusional beliefs, so-called “AI psychosis”? Or would you prefer to focus on the technological side and find out how LLMs compare to other forms of AI and machine learning and data processing and what innovations could lead to more effecient, more effective and less harmful AI?
You could basically follow two routes, I think. You could major in something like sociology, economics or philosophy and minor/specialize in the impacts of AI or technology in general. Alternatively, you could major in AI or IT and do some kind of minor or specialization in its impacts on the world. Though I suppose it’s possible you may be able to find a course somewhere in the world that tries to teach both.
I am currently studying Cybersecurity as my bachelor, so my background is very technical… I think I would like to go the technical route, but also sprinkle in a bit of sociology, or maybe philosophy or something politics related… Maybe focusing on being sort of bridge in interdisciplinary research with deep knowledge of the technical aspects, while also focusing on more social aspects and collaborating on research with people that have deeper understanding of these issues. There is a master degree in Prague called prg.ai I believe, which kinda tries to teach about AI from many different povs, which could be interesting for me.
To understand things from a technical point of view, you’ll want a solid understanding of Artificial Neural Networks (ANN) and how they are usually trained (back-propagation). These have been around for decades and are also used for classification tasks and things like image processing in your phone or GPU.
To create an LLM, you feed the output of the ANN back into the network, creating a Recurrent Neural Network (RNN) and you tokenize the input.
It also helps to have some general knowledge of machine learning (ML) and understand terms like supervised learning and unsupervised learning.
Most of the ideas behind LLMs have been around for years or decades. The big difference is that nowadays we can throw enough compute at the problem to make them viable.
There are two ways to make ANNs and LLMs more efficient: better algorithms and more specialized hardware. For example, bitnets with ternary connections run much better on conventional CPUs and GPUs. Hardware like Google’s NPUs, OpenAI’s new chip, analog computers or neuromorphic computing could run them more efficiently.
there is a project called Real Good AI
realgoodai.org/researchand a Youtube Video from Markaplier, called
Talking AI with Bob and Mandy
I believe they are going the right direction and maybe it is also in your interest.
