It depends on the model.
2 Likes
Hope these guys behind Artificial Intelligence are of the models who are actually intelligent. But I have my doubts.
The whole hype around LLMs, their training as well as interference processing is only feasible due to abundance of computational processing capabilities.
And that turned into âletâs throw more processing power at itâ quite quickly. Until the Chinese came along with deepseek and their model that didnât relied on âbrute-forceâ solely but has shown that models could be trained more cost-efficiently as well.
The times when each bit and byte has been precious due to limited capacities and such, are long gone.
In short, exceptionally complex and hard to solve problems didnât went away, even with artificial intelligence, they are still tough to solve. There are still problems with no definite answers. Itâs easy to write code that will never terminate and would be stuck in an infinite loop.
In short, itâs quite easy to verify if the key to an AES encryption is correct, if youâve got the correct key.
But in case you donât have the correct key and have to brute force your way to the correct one, artificial intelligence hasnât made and wonât make it any easier.
1 Like
A friend of mine told me about 10 years ago that AI would be the destruction of mankind on planet EarthâŚ
There are 32 different ways AI can go rogue, scientists say â from hallucinating answers to a complete misalignment with humanity
AI Gone Wrong: The Errors, Mistakes, and Hallucinations of AI (2023 â 2025)
7 Terrifying AI Risks That Could Change The World
Up until now, we, so called âHuman Beingsâ equipped with our Organic, presumed, Intelligence have caused, immense harm to âHumanityâ and everything else and have been in misalignment with the principle of Life itself on this planet.
The AI, our own creation, how it is being used already and how it will be used in the future fits so well within that pattern of behavior. No surprises here.
I donât understand this fear mongering whenever AI becomes the topic of any discourse. Like why are you so scared of it? We have all seen the Terminator movies. 1 and 2, pretty decent films (not the rest). But we are all Linux users here so we should be a little bit more educated about the whole thing compared to a layman right? AI is just a program, written with a coding language and it can mimic human communication skills. That is it. The reason why it is revolutionary is because it has access to the whole human knowledge and that was because of the internet of course, and it can basically find what you are looking for in mere seconds and present it to you. It is like having a professor of everything at armâs reach. That is it. Ah yeah, and it can code. Coding also is a form of human language so of course it will be able to do that.
The evolution of human technology has always been a form of mimicry. We have tried to mimic nature itself, and we did. And now we can build things that can mimic our own skills. That is a natural point of arrival. It is not extraordinary.
1 Like
That is pretty much not true.
AI hasnât access to the whole human knowledge, it may be trained on text, images and other sources. But thatâs definitely not all of it. Furthermore, the current models are trained on datasets that has a strict cut-off date. In short terms, the models in use currently, will be out of date tomorrow. That might be not that critical in most of the use cases, but up so far, itâs not feasible to train a new model within a day. That process takes weeks or even months.
Youâre confusing per-determined algorithms with non-deterministic, probabilistic approaches in computation.
No artificial intelligence is able to count to infinity. In less than infinite time.
2 Likes
These are all just truisms though.
You misunderstand me. I know all of that. I am using Chatgpt and Grok and they have full internet access. Even deepseek has internet access now. That kind of removes their âcutoffâ date per se. They can access the latest news articles for example and provide you up date information about the most recent events.
That of course, doesnât mean that the information they provide are accurate or based on reality. They just aggregate whatever they find on the internet. It is up to me (and you) to verify. But if the data that you seek has already been established scientificly, i wouldnât really worry about its accuracy. At least not that much. I would still check the resources.
In terms of mathematics, it is amazing stuff. I have always been bad at maths. And i hardly know the formulas to calculate any given outcome for an equation with multiple variables. Now i can just write my problem, and it just solves it for me. To me, that is just wonderful.
@Sam_Fisher, I canât speak to generalities about AI but I can discuss my own informed decisions.
I wonât give government or corporations anything for free. Through unenlightened self-interest they have destroyed much of our planet and are destroying basic human rights. I wonât help train AI or give it raw material to process in furtherance of their goals.
My hobbies are creative ones: writing, acting, music, woodworking. My inner satisfaction comes from me doing the work, not someone or something doing it for me. I wonât help train AI to approximate the human spirit. It removes one of the last joys available: creating something meaningful to the person.
How much benefit will society receive from AI? Donât know. How much harm will it cause? Donât care. Iâm 60+ years old and Iâll probably be gone by the time the true benefits and harms come to light. But Iâm not going to contribute my minim to the AI pile while Iâm here.
3 Likes
I guess that youâve never heard of the halting problem in computer sciences⌠which is known to be undecidable. It can be stated in terms of math. But it canât be computed.
Your argument that even deepseek has internet access now doesnât make a difference. It still canât figure out that Taiwan is an independent & free country. But in itâs current scope, its still moderated & censored to align without contradictions to the regimes point of views.
I love this. This is exactly what is going on. I am also not quite happy with the human âendeavoursâ like art for example being exploited by the AI but i think it is too late to intervene now. The damage is done. As a music lover, the âfakeâ songs that it creates also gets on my nerves but it is still possible to distinguish what it does from original content, if you are somewhat familiar with it. The âghiblifyingâ of imagery was fun though, i am not gonna lie. I also did some of that 
But i can understand the original creator being furious about it. I would be too.
Maybe we need new laws. I donât know. It is too early. Everything is so new. But i still like it. Future generations wonât have any kind of excuse to not know anything. Everything will be in their fingertips. Everything will be easier. Maybe that will amount to something good. Or bad. I donât know. I wonât be here to see it
What AI did you use to come up with the legal status of Taiwan?
I donât want that this thread gets into politics and be shutdown since as its âinstigatorâ, I am enjoying the input from all of you.
However, I think we should get our facts right.
Last time I checked, Taiwan legal status is still disputed and unresolved under international law.
It is a âself-governingâ entity with independent institution. However, to my knowledge (I may be wrong), most of the countries in the world donât recognize, legally, Taiwan as an independent state. So ultimately its legal status is âundeterminedâ at best.
There is a difference between âde factoâ and âde jureâ.
No! I donât!
1 Like
In short terms, almost every LLM is moderated. And in case of deepseek, itâs even censored.
Generally spoken, the LLMs are trained & moderated to âaggregateâ an answer to the users question or instructions in an almost instant manner, direct and fast, without interruptions.
There are plenty of cases where it has been shown that AI has been trained upon copyrighted data, without any permissions of the original copyrights owner. And companies such as OpenAI are actively moderating the responses to prohibit to reveal the sources that have been included in the training stages.
Anyway, just let me state another culprit the developers of LLMs are faced with. Nowadays sciences rely heavily on publications within scientific journals. And those sources, which the whole scientific community relies upon to communicate their latest findings, isnât accessible to the public domain in most cases. The publishers essentially hiding the latest scientific insights behind paywalls and therefore these couldnât be included within the training data directly without the risk of legal consequences. And the same applies to all other forms of media which isnât accessible to the public domain yet.
If you know instruct an LLM to elaborate over an problem for a given span of time. It will definitely simply postpone the direct answer is generated already to be delivered at a later point in time. It wonât search the web to autonomously âre-trainâ it predetermined set of parameters composed during the training stages.
If you confront it with an math question that has been included within it set of training data, it will refer to the very same answer that has been predetermined. It wonât reevaluate alternatives or even adapt new novel solutions itâs presented with.
In case that some one think that artificial intelligence is able to think. From my point of view: It definitely canât think.
Well, all of this is âfine and dandyâ. Thanks for the elaboration!.
However, you chose not to address my objection to your statement that:
This has nothing to do with if this and that LLM is censored or biased.
You make an statement without facts backing it up.
As I mentioned above, Taiwan operates as a âself-governingâ entity. Only a handful of countries have officially recognized it as a âsovereignâ state. Majority of the countries in the world, including USA, Canada, EU countries, Japan⌠have not.
Iâm not discussing if it should or should not be independent and free. Neither am I taking any political side on this matter.
My concern here is solely this:
If you are using a statement to back up your argument for or against something, at least, get your facts right.
I leave it at that. As I said, I donât want this thread to become politicized and shutdown.
