I have been working on AI and Machine Learning (ML) since the early 1980's. In 2017 there was an absolute break through in ML in the area of Artificial Neural Networks that we called Large Language Model (LLM). LLM is a form of AI though not the only application of it.
I work on both Expert Engines and train my own local domain specific LLMs. It is one of the greatest breakthroughs in modern Computer Science. So if it is a fool that I am then it is a fool that I will be .
A few things that are regrettable in today's society is that:
1) People do not know what AI is and appear to believe that AI is entirely Chatbots.
2) People get upset at corporations having no regulations and point to AI as a problem when in reality it is that these corporations have no regulations. AI is not doing anything wrong the corporations are. Put the blame where it belongs if you ever expect this issue to be corrected.
3) People assume AI should be perfect, yet Natural Intelligence is far from perfect so why conclude that something developed from Natural Intelligence is going to fix all the issues with it when it was designed to work on the same principles?
I understand the hate of AI Data Centers and the issues with them. It makes sense. Though the fact that the masses have been lead to believe AI is entirely chatbots, that they require massive Data Centers to operate and that AI is intrinsically evil is a failing of society not AI.
All of my work is done on a local server that uses less resources than the average gaming rig. No one is calling for the take down of people's personal gaming machines and making them illegal. Until they are I am not going to even consider their issues with my AI development. The idea that AI can only be done with these massive data centers is a leading issue with the Anti AI society. If they would focus on the real issues such as corporations having zero regulations and address that issue then the data center issues would go away.
All in all it means nothing to AI because AI does not require nor will it ever require these data centers. What requires these data centers are chatbots that can interface with million of users. That is not AI. That is a corporate product built with AI.
When we debate the ethics of AI we need to also debate the ethics of fire. If you think that is ridiculous then you are starting to get my point.
Cool ... I think LLM are magic ... You know that and have for decades, since one read my MIR with the MIR expert, and I was amazed how many the human missed and the LLM missed ..But together it's nearly 100%
And I have a little LLM on an esp32 for meter reading ... Truly amazing this day and age
And yes, as I said in 1988, information is the new currency and the ones that control will win, these days a LLM {AI} is controlled by people that say one thing and want another
@unusnemo My main issues with LLMs are that the way they're trained is impacting everything else on the Internet (because of the onslaught of poorly-written crawler bots that have led to all sorts of half-assed mitigations that they now work to get around) and the ethics of using models that are trained based on data that was obtained in this way and without regard for copyright or consent.
AI as a field is great, but LLMs and diffusion models and their ilk deeply bother me.
@unusnemo Which is to say that I agree that the problem with "AI" is the way that corporations are rolling it out and making it everyone else's problem, but even if you're running it fully locally, you have to keep in mind where the model was trained from in the first place.
The models I use are trained on my home server with well vetted public resources that do not violate anyone's license and my own research. So, I have that in mind already.
Keep in mind that many LLMs are local work and not corporate RAG systems. Your concerns are real but again not related to AI or LLMs. They are related to how corporations are not regulated, down right break laws and are not held accountable. Yes, I have these same issues. I just put the blame on corporations not AI.
@unusnemo Good to know. If all AI people were as responsible with it as you then I wouldn't have a problem with it (assuming that your anti-corporate stance also extends to the notion of not using it to replace human creativity or ability, of course).
As far as I am concerned there is no replacement for human creativity. We would not even be able to build LLMs and teach them if it were not for human creativity. And I am certainly not looking to see humans replaced. Artificial Intelligence (AI) is a tool that we can use to augment Natural Intelligence (NI). Neither are infallible and there are certain aspects AI does not extend to. I am a poet. Can AI be taught to write poetry, most certainly. Though an AI does not have feelings. It can imitate a great poet but it cannot write original work inspired by its own emotions as humans do. The same extends to all arts. That does not mean I cannot use ComfyUI to create some banners for my website. I might even generated some textures. But it is no replacement for Boris Vallejo.
Augment human creativity and ability yes, replace absolutely not. It could not even if we wanted to. We created an Artificial Neural Network and we can teach it. Though how does one teach emotion? That has always been the largest caveat of AI. You cannot have genuine art without genuine feeling. An artist that never experienced pain could not use their art form to express it. Neither can an AI.
There are a lot of responsible researchers and developers in AI. They do not get the press though. The corporations that are destroying ecosystems get the attention of the Media ( I refuse to call it news anymore). The problem that we are experiencing with the AI phenomenon is not intrinsically related to AI itself. It is related to human greed expressed by corporations that continue to go without regulation or consequence. And as long as the Anti-AI advocates keep focusing on LLMs and AI in general and not the the lack of regulations for corporations and the fact that our governments are bought out and letting these corporations blatantly violate the law. Then the issue will continue. As people are not focused on the problem, they are focused on a symptom of the problem. AI Data Centers are built to facilitate the general public with interfaces that let millions of users simultaneously interface with their General Purpose LLMs. This product of corporations is devastating our environment due to lack of regulations and or enforcing said regulations.
By the way, I advise people all of the time to not consult AI about something they know nothing about. As they will have no way to vet the information they receive. AI is great to parse large data sets that would take humans months to years. Though like humans it can make mistakes, especially if there is an issue with the data set. RAG systems that parse social media for answers is just as unreliable as the information they parse.
@fluffy 💜 @Unus Nemo @Melissa BearTrix With manpages like nearly 5K of them scraped with a friendly HTTP request page per page to not saturate the server and manpages are public. The model and LLM are opensource (LLama.cpp) + model of choice. Frankly just using it as last resort there where I have some trouble with my network. The claim that one does not learn anything from it is false as well, you can mindlessly copy paste or actually learn from that one command I did not knew about. Like Unus stated, the hate is misplaced if aimed at users. It's the big TechBro's that make it invade my phone and search engine and scrapes the web as a wild joker and that makes people pay to "Vibe Code" Forgot to mention that I mean local so local host, it does not reach the net nor need it. Some misconception about "local" also, there is a distinction between the ones placed on your pc that use your network for the API or the one alike ours without any API call no network I repeat UI is 127.0.0.1 By feeding it note, the whole point of the RAG layer is that the model doesn't need to "know" the content, it just needs to retrieve the right chunk.
Unus Nemo
in reply to Melissa BearTrix • •@Melissa BearTrix
I have been working on AI and Machine Learning (ML) since the early 1980's. In 2017 there was an absolute break through in ML in the area of Artificial Neural Networks that we called Large Language Model (LLM). LLM is a form of AI though not the only application of it.
I work on both Expert Engines and train my own local domain specific LLMs. It is one of the greatest breakthroughs in modern Computer Science. So if it is a fool that I am then it is a fool that I will be
.
A few things that are regrettable in today's society is that:
1) People do not know what AI is and appear to believe that AI is entirely Chatbots.
2) People get upset at corporations having no regulations and point to AI as a problem when in reality it is that these corporations have no regulations. AI is not doing anything wrong the corporations are. Put the blame where it belongs if you ever expect this issue to be corrected.
3) People assume AI should be perfect, yet Natural Intelligence is far from perfect so why conclude that something developed from Natural Intelligence is going to fix all the issues with it when it was designed to work on the same principles?
I understand the hate of AI Data Centers and the issues with them. It makes sense. Though the fact that the masses have been lead to believe AI is entirely chatbots, that they require massive Data Centers to operate and that AI is intrinsically evil is a failing of society not AI.
All of my work is done on a local server that uses less resources than the average gaming rig. No one is calling for the take down of people's personal gaming machines and making them illegal. Until they are I am not going to even consider their issues with my AI development. The idea that AI can only be done with these massive data centers is a leading issue with the Anti AI society. If they would focus on the real issues such as corporations having zero regulations and address that issue then the data center issues would go away.
All in all it means nothing to AI because AI does not require nor will it ever require these data centers. What requires these data centers are chatbots that can interface with million of users. That is not AI. That is a corporate product built with AI.
When we debate the ethics of AI we need to also debate the ethics of fire. If you think that is ridiculous then you are starting to get my point.
Have a great day!
like this
Melissa BearTrix, Sea Slug of Doom, John van, Satu Elisa Schaeffer, Medea Vanamonde 🏳️⚧️, sen, Paul Turnbull 🇨🇦, Plan-A̵̛͈̬̥̿͋̓͛̕, suzanne, Paul C. and Coquillage Logs like this.
reshared this
Melissa BearTrix, Lazarou Monkey Terror 🚀💙🌈, Ra (Freyja) (it/its)𒀭𒈹𒍠𒊩 (cringe EU-brained military girl), jmj, Paul Turnbull 🇨🇦 and Plan-A̵̛͈̬̥̿͋̓͛̕ reshared this.
Melissa BearTrix
in reply to Unus Nemo • • •@unusnemo
Cool ... I think LLM are magic ... You know that and have for decades, since one read my MIR with the MIR expert, and I was amazed how many the human missed and the LLM missed ..But together it's nearly 100%
And I have a little LLM on an esp32 for meter reading ... Truly amazing this day and age
And yes, as I said in 1988, information is the new currency and the ones that control will win, these days a LLM {AI} is controlled by people that say one thing and want another
Hugz & xXx
like this
Unus Nemo and Plan-A̵̛͈̬̥̿͋̓͛̕ like this.
fluffy 💜
in reply to Unus Nemo • • •@unusnemo My main issues with LLMs are that the way they're trained is impacting everything else on the Internet (because of the onslaught of poorly-written crawler bots that have led to all sorts of half-assed mitigations that they now work to get around) and the ethics of using models that are trained based on data that was obtained in this way and without regard for copyright or consent.
AI as a field is great, but LLMs and diffusion models and their ilk deeply bother me.
like this
Unus Nemo and Coquillage Logs like this.
Melissa BearTrix reshared this.
fluffy 💜
in reply to fluffy 💜 • • •Unus Nemo likes this.
Melissa BearTrix reshared this.
Unus Nemo
in reply to fluffy 💜 • •@fluffy 💜 @Melissa BearTrix
The models I use are trained on my home server with well vetted public resources that do not violate anyone's license and my own research. So, I have that in mind already.
Keep in mind that many LLMs are local work and not corporate RAG systems. Your concerns are real but again not related to AI or LLMs. They are related to how corporations are not regulated, down right break laws and are not held accountable. Yes, I have these same issues. I just put the blame on corporations not AI.
Melissa BearTrix likes this.
reshared this
Melissa BearTrix, Cass🇨🇦 and Bongmaster reshared this.
fluffy 💜
in reply to Unus Nemo • • •like this
Unus Nemo and Plan-A̵̛͈̬̥̿͋̓͛̕ like this.
Unus Nemo
in reply to fluffy 💜 • •@fluffy 💜
As far as I am concerned there is no replacement for human creativity. We would not even be able to build LLMs and teach them if it were not for human creativity. And I am certainly not looking to see humans replaced. Artificial Intelligence (AI) is a tool that we can use to augment Natural Intelligence (NI). Neither are infallible and there are certain aspects AI does not extend to. I am a poet. Can AI be taught to write poetry, most certainly. Though an AI does not have feelings. It can imitate a great poet but it cannot write original work inspired by its own emotions as humans do. The same extends to all arts. That does not mean I cannot use ComfyUI to create some banners for my website. I might even generated some textures. But it is no replacement for Boris Vallejo.
Augment human creativity and ability yes, replace absolutely not. It could not even if we wanted to. We created an Artificial Neural Network and we can teach it. Though how does one teach emotion? That has always been the largest caveat of AI. You cannot have genuine art without genuine feeling. An artist that never experienced pain could not use their art form to express it. Neither can an AI.
There are a lot of responsible researchers and developers in AI. They do not get the press though. The corporations that are destroying ecosystems get the attention of the Media ( I refuse to call it news anymore). The problem that we are experiencing with the AI phenomenon is not intrinsically related to AI itself. It is related to human greed expressed by corporations that continue to go without regulation or consequence. And as long as the Anti-AI advocates keep focusing on LLMs and AI in general and not the the lack of regulations for corporations and the fact that our governments are bought out and letting these corporations blatantly violate the law. Then the issue will continue. As people are not focused on the problem, they are focused on a symptom of the problem. AI Data Centers are built to facilitate the general public with interfaces that let millions of users simultaneously interface with their General Purpose LLMs. This product of corporations is devastating our environment due to lack of regulations and or enforcing said regulations.
Plan-A̵̛͈̬̥̿͋̓͛̕ likes this.
Foxdick Farms
in reply to Unus Nemo • • •@unusnemo @fluffy
AI datacenters are a cancer.
People need to run AI locally, on their own hardware.
Unus Nemo likes this.
Unus Nemo
in reply to Foxdick Farms • •@Foxdick Farms
Yes, local AI is superior in every way to corporate offerings.
Unus Nemo
in reply to fluffy 💜 • •@fluffy 💜
By the way, I advise people all of the time to not consult AI about something they know nothing about. As they will have no way to vet the information they receive. AI is great to parse large data sets that would take humans months to years. Though like humans it can make mistakes, especially if there is an issue with the data set. RAG systems that parse social media for answers is just as unreliable as the information they parse.
Plan-A̵̛͈̬̥̿͋̓͛̕ likes this.
Plan-A̵̛͈̬̥̿͋̓͛̕
in reply to fluffy 💜 • — (0.0.0.0/ 0) •Like Unus stated, the hate is misplaced if aimed at users. It's the big TechBro's that make it invade my phone and search engine and scrapes the web as a wild joker and that makes people pay to "Vibe Code"
Forgot to mention that I mean local so local host, it does not reach the net nor need it.
Some misconception about "local" also, there is a distinction between the ones placed on your pc that use your network for the API or the one alike ours without any API call no network I repeat UI is 127.0.0.1
By feeding it note, the whole point of the RAG layer is that the model doesn't need to "know" the content, it just needs to retrieve the right chunk.
like this
Melissa BearTrix and Vast Country Hypnosis like this.
Melissa BearTrix reshared this.