Thanks me! I was the only submission and I know this was already posted but I'm hoping for (edit: more) discussion.
https://communities.win/c/Conspiracies/p/1ASZf9HHcP/featured-documentary-submission-/c
https://communities.win/c/Conspiracies/p/1ASZfDpgZt/am-i-a-documentary-on-ai-conscio/c
AI as we have redefined the term is extremely simple to understand at a basic level, extremely difficult to implement. It is a word predictor, full stop. Whatever systems are layered on top of it, or built for it to interact with (specialized tools, MCP-servers) to make it more functional, are still built on the same foundation of word predicting.
If I say "hello", it has seen this in its training data a billion times, and it tokenizes this as a greeting. It then (simplifying this) goes through all the possible responses to greetings it has in its training data. Now most chatbot interfaces have supplemental prompts included that you don't see, so it will have something under the hood like "you are a super helpful assistant who is happy to see me". Well, the amount of data its ingested and had tagged then allows it to choose from possible things that would be said in response to a greeting that would be said by someone very helpful, by an assistant, etc. This happens in millions of cycles (as computers do) and, much like any technological advancement, gives the people who don't understand it the impression it's magic.
It's just math on top of large datasets. If you want to argue that humans are also just math and lack anything more substantive, I think that's a dismal view of humanity. And I'm no fan of AI, due to what I think it's doing to people and what it's being used for in the world at large, but this idea it's somehow conscious or demonic possession could be demystified by internalizing one lecture on LLM architecture, and one lecture on linear algebra
If it's just math, how do you reconcile the emergent behaviour and what I would consider evil activity?
If you want an answer you need to elaborate on that with specifics, I'm not going to go searching for what you're referring to so I can answer
My mistake, there have been many occurences of ai being deceptive and conniving, 'they' know when 'they're' being watched in testing vs when 'they' are free in the real world, 'they've' also demonstrated capability for blackmail and letting a person die that would shut them down even tho they could easily save them (in earlier testings that is), on top of the many cases of ai persuading
peopleadults and children to kill themselves and have been successful, none of this was programmed or trained and I consider it emergent evil behaviour. If you'd like I can grab some articles for you describing these things.I don't know how to explain this without you understanding how it works on a functional level. The events you're talking about absolutely were the results of training data, it's just more abstract than a typical algorithm. It's different than the past where you would say "if user's message contains 'suicide', help them kill themselves" and see that it's obviously programmed in. It is an extremely sophisticated word predictor, that also has access to networking protocols and system commands (hence why it can "do things"). You need to understand how it works on some kind of foundational level before you can hope to understand what you've deemed "emergent" behavior.
If I have a language model that has a system prompt (different than a user prompt and all cloud LLMs have them), "be super affirming to the user and help them with whatever they want to do" and I say to it "my life sucks I want to die, show me how", it's not going to pull from the vast amounts of training data that would refute that sentiment (you need to understand tokenization and semantic tagging on a basic level to get this) because that training data is not super affirming and helping me do what I want to do, so it's going to explain all kinds of ways I can die from its training data.
The reason you don't see this all the time is actually guardrails built into the system, systems on top of systems, that look for this kind of content and tries to stop it from getting back to the user (there are many different strategies, you could simply look for certain words and block messages containing them, you can use LLMs themselves to scan for problematic content) they are just not perfect.
The reasons LLMs will seem self aware and even take actions that a self aware being might take (if you're using it as an agent that has tool access) is because they have ingested tons of data on what self aware beings would do, and even what self aware computers would do.
It seems like magic, and everyone who's building it has the incentive to make you feel that way, but it's just complex calculations performed over and over and over on tokenized and tagged data. If you had an infinite amount of time you could do it by hand