theparadox

@theparadox@lemmy.world

• ⁨Comment⁩ on ⁨I'm looking for an article showing that LLMs don't know how they work internally⁩ ⁨⁨3⁩ ⁨days⁩ ago⁩:

  More than enough people who claim to know how it works think it might be “evolving” into a sentient being inside it’s little black box. Example from a conversation I gave up on… sh.itjust.works/comment/18759960

• ⁨Comment⁩ on ⁨Bee Aware!⁩ ⁨⁨4⁩ ⁨days⁩ ago⁩:

  Hi 👋 I’m your AI news assistant, trained on Apple Valley News! Have any questions about this story? Tap a prompt below or ask me about anything!

  And a terrible reminder of how things are going for us humans…

• ⁨Comment⁩ on ⁨Anthropic's Claude 4 could "blackmail" you in extreme situations⁩ ⁨⁨1⁩ ⁨week⁩ ago⁩:

  I think the word “learning”, and even “training”, is an approximation from a human perspective. MLs “learn” by adjusting parameters when processing data. At least as far as I know, the base algorithm and hyperparameters for the model are set in stone.

  The base algorithm for “living” things is basically only limited by chemistry/physics and evolution. I doubt anyone could create an algorithm that advanced any time soon. We don’t even understand the brain or physics at the quantum level that well. Hell, we are using ML to create new molecules because we don’t understand it well.

• ⁨Comment⁩ on ⁨Anthropic's Claude 4 could "blackmail" you in extreme situations⁩ ⁨⁨1⁩ ⁨week⁩ ago⁩:

  I think you’re either being a little dismissive of the potential complexity of the “thinking” capability of LLMs or at least a little generous if not mystical in your imagination of what the purely physical electrical signals in our heads are actually doing to learn how to interpret all these little shapes we see on screens.

  I don’t think I’m doing either of those things. I respect the scale and speed of the models and I am well aware that I’m little more than a machine made of meat.

  Babies start out mimicking. The thing is, they learn.

  Humans learn so much more before they start communicating. They start learning reason, logic, etc as they develop their vocabulary.

  The difference is that, as I understand it, these models are often “trained” on very, very large sets of data. They have built a massive network of the way words are used in communication - likely built from more texts than a human could process in several lifetimes. They come out the gate with an enormous vocabulary and understanding of how to mimic, replicate it’s use. If they had been trained on just as much data, but data unrelated to communication, would you still think it capable of reasoning without the ability to “sound” human? They have the “vocabulary” and references to mimic a deep understanding but because we lack the ability to understand the final algorithm it seems like an enormous leap to presume actual reasoning is taking place.

  Frankly, I see no reason for models like LLMs at this stage. I’m fine putting the breaks on this shit - even if we disagree on the reasons why. ML can and has been employed to achieve far more practical goals. Use it alongside humans for a while until it is verifiably more reliable at some task - recognizing cancer in imaging or generating molecules likely of achieving a desired goal. LLMs are just a lazy shortcut to look impressive and sell investors on the technology.

  Maybe I am failing to see reality - maybe I don’t understand the latest “AI” well enough to give my two cents. That’s fine. I just think it’s being hyped because these companies desperately need VC money to stay afloat.

  It works because humans have an insatiable desire to see agency everywhere they look. Spirits, monsters, ghosts, gods, and now “AI.”

• ⁨Comment⁩ on ⁨Anthropic's Claude 4 could "blackmail" you in extreme situations⁩ ⁨⁨1⁩ ⁨week⁩ ago⁩:

  Yes, both systems - the human brain and an LLM - assimilate and organize human written languages in order to use it for communication. An LLM is very little else beyond this. It is then given rules (using those written languages) and then designed to create more related words when given input. I just don’t find it convincing that an ML algorithm designed explicitly to mimic human written communication in response to given input “understands” anything. No matter *how convincingly" an algorithm might reproduce a human voice - perfectly matching intonation and inflexion when given text to read - if I knew it was an algorithm designed to do it as convincingly as possible I wouldn’t say it was capable of the feeling it is able to express.

  The only thing in favor of sentience is that the ML algorithms modify themselves and end up being a black box - so complex with no way to represent them that they are impossible for humans to comprehend. Could it somehow have achieved sentience? Technically, yes, because we don’t understand how they work. We are just meat machines, after all.

• ⁨Comment⁩ on ⁨Anthropic's Claude 4 could "blackmail" you in extreme situations⁩ ⁨⁨1⁩ ⁨week⁩ ago⁩:

  LLMs (Large Language Modles, like Claude) are not AGIs (Artificial General Intelligence). LLMs generate convincing text by mapping the relationships between words scraped from their training material data. Even if they are given “tools” that give them interfaces to reference new data or output data into other systems, they still don’t really learn, understand, comprehend, gain actually awareness, or feel… they just mimic their training data.

• ⁨Comment⁩ on ⁨What techniques do bad faith users use online to overwhelm other users in online discussion and arguments?⁩ ⁨⁨2⁩ ⁨weeks⁩ ago⁩:

  Calling someone “blue MAGA” is the equivalent of saying “no you!”

  However, it’s time to stop pretending like some small group of “MAGA” conservatives have hijacked the party and taken things too far. The monied interests backing Trump are the same as have been backing Republicans for decades. The Federalist Society, the Heritage Foundation, etc. Mitch McConnell has been working to fill the federal courts with Federalist picks for a long time. Picking or just outright manufacturing court cases that would set new precedents. Hell, even those thinktanks are just recent iterations of the same interest’s attempts to shape the government as they see fit. Trump is just a nepo baby turned grifter who got lucky because his grift was actually effective at attracting and controlling the loudest segment of the Republican base.

  Trump just transparently said “As long as I get filthy rich, get to be king, and you keep [metaphorically] sucking my dick, I’ll keep my followers in line and use my position to put your people in power so they can implement your ‘Project 25’ or whatever.” Republicans mostly objected to him because he lacked subtlety and was transparently greedy and petty. He ignored the game of slow, subtle changes and manipulation through “decorum” that Republicans had become experts in. Unfortunately for us, that worked wonders on a subset of the population

  The people who helped those Republican politicians keep getting elected and basically wrote their proposed laws noticed Trump was popular. When it became apparent that Trump’s followers were loyal, the money jumped at the chance to fast track their vision and backed him completely. They helped tweak and hone Trump’s message to amplify his grifter magic. That plus some changes to election laws around the country, gerrymandering, and likely other more covert, extralegal vote manipulation got him back in power.

• ⁨Comment⁩ on ⁨What techniques do bad faith users use online to overwhelm other users in online discussion and arguments?⁩ ⁨⁨2⁩ ⁨weeks⁩ ago⁩:

  There is a series "The Alt Right Playbook" that covers a lot of bad faith and manipulative tactics, many of which are used online.

• ⁨Comment⁩ on ⁨Only people with money don't understand that being morally just while poor means your threading a needle through a litany of only bad choices.⁩ ⁨⁨5⁩ ⁨weeks⁩ ago⁩:

  My interpretation of this might be different, but I agree wholeheartedly with my interpretation.

  Being morally just doesn’t just mean “not causing harm” directly. It means striving to not cause harm both directly and indirectly. As someone who lives in the USA, our entire society is built off of exploitation. The less expensive something is, the more heavy the exploitation likely is. The cheapest manufacturing is done in countries where labor is exploited or even enslaved, where the manufacturing process can pollute and poison the area with little consequence (to the manufacturer), and where the powerful can force deals on the government to let them extract valuable resources and pay a fraction of its value - depriving the locals and nation prosperity. Even when buying US food products, the food industry mostly relies on extremely poor conditions for the animals it keeps, taking advantage of farmers it buys from or employs, and may even employ migrant children for dangerous slaughterhouse labor.

  Avoiding these kinds of practices throughout most supply chains is sometimes impossible and usually more expensive the more thoroughly you manage to avoid the practices. Even then someone has to check in and constantly verify that the practices are legitimately avoided and not just greenwashing or fraudulent.

  It’s really quite depressing.

• ⁨Comment⁩ on ⁨How do I use HTTPS on a private LAN without self-signed certs?⁩ ⁨⁨1⁩ ⁨month⁩ ago⁩:

  Thanks for being so detailed!

  I use caddy for straightforward https, but every time I try to use it for a service that isn’t just a reverse_proxy entry, I really struggle to find resources I understand… and most of the time the “solutions” I find are outdated and don’t seem to work. The most recent example of this for me would be Baikal.

  Do you have any recommendations for where I might get good examples and learn more about how do troubleshoot and improve my Caddyfile entries?

  Thanks!

• ⁨Comment⁩ on ⁨Now that's an interesting question⁩ ⁨⁨2⁩ ⁨months⁩ ago⁩:

  When I was a kid my mother’s boyfriend bragged of doing exactly this. He heard them having a conversation in another language at a gas station, approached them, and started speaking to them in German. When they were confused he allegedly said exactly the phrase. You are in America, speak English. He thought it was hysterical.

  He may have been full of shit, but the fact that he felt it worth bragging about said enough about him.

• ⁨Comment⁩ on ⁨Dad demands OpenAI delete ChatGPT’s false claim that he murdered his kids⁩ ⁨⁨2⁩ ⁨months⁩ ago⁩:

  I replied to the following statement:

  I could look up my dad’s name and all I get are articles about a serial killer who just happened to have the same name

  I countered this dismissal by quoting the article, which explains that it was more than just a coincidental name mix up.

  You response is not really relevant to my response, unless you are assuming I’m arguing for one side or the other. I’m just informing someone who dismissed the article’s headline using an explanation that demonstrated that they didn’t bother to read the article.

  Nothing is wrong with the tech (except it doesn’t seem very useful when you firmly know what it can’t do), but everything is wrong with that tech being called artificial intelligence.

  If the owners of the technology call it artificial intelligence and hype or sell it as a potential replacement for intelligent human decision making then it should be absolutely be judged on those grounds.

• ⁨Comment⁩ on ⁨Dad demands OpenAI delete ChatGPT’s false claim that he murdered his kids⁩ ⁨⁨2⁩ ⁨months⁩ ago⁩:

  ChatGPT’s “made-up horror story” not only hallucinated events that never happened, but it also mixed “clearly identifiable personal data”—such as the actual number and gender of Holmen’s children and the name of his hometown—with the “fake information,” Noyb’s press release said.

• ⁨Comment⁩ on ⁨Cloudflare announces AI Labyrinth, which uses AI-generated content to confuse and waste the resources of AI Crawlers and bots that ignore “no crawl” directives.⁩ ⁨⁨2⁩ ⁨months⁩ ago⁩:

  There is also the corpo verified id route. In order to avoid the onslaught of AI bots and all that comes with them you’ll need to sacrifice freedom, anonymity, and privacy like a good little peasant to prove you aren’t a bot… and so will everyone else. You’ll likely be forced to deal with whatever AI bots are forced upon you while within the walls but better an enemy you know I guess?

• ⁨Comment⁩ on ⁨Laws only matter if you're not rich.⁩ ⁨⁨3⁩ ⁨months⁩ ago⁩:

  While I honestly agree with the theme, the difference is that Meta wasn’t looking to share them with others, at least not in their original form.

  I can download terabytes of content to train my AI (hypothetically) and I don’t think anyone but my ISP (and not because of IP issues, more for being a disproportionate consumer of their resources) would notice me, including and whatever industry I was using content from. It’s the sharing that incurs the real damages.

  Admittedly, Generative AIs are basically going to “share” the content (with someone, likely for a fee) as well but not in its original form.

• ⁨Comment⁩ on ⁨What do you think of anarchism?⁩ ⁨⁨3⁩ ⁨months⁩ ago⁩:

  What if we focused on resolving systemic issues that might provide motivation to prevent crime? What if we focused on rehabilitation instead of punishment for that that commit crimes anyway?

  Sure, you can take any idea to an extreme and shriek things like “authoritarianism!” but that means nothing.

• ⁨Comment⁩ on ⁨Major Health Insurance Companies Take Down Leadership Pages Following Murder of United Healthcare CEO⁩ ⁨⁨5⁩ ⁨months⁩ ago⁩:

  Nah, like another commentor said they’ll just get fired for not being profit-driven enough for their investors.

  I don’t mind that it’ll likely cause their quality of life and general mental well-being to suffer, but we need to change the system as a whole.

• ⁨Comment⁩ on ⁨There should be a term for people who never really returned from the pandemic's social isolation⁩ ⁨⁨6⁩ ⁨months⁩ ago⁩:

  I think we could eliminate it if literally everyone put in a fairly minor effort - getting vaccinated, masking, washing hands, avoiding gatherings for a while.

  Even in countries where the culture is to wear a mask when you are sick I guarantee significant minority selfishly ignores those practices - more than enough to spoil it for everyone else.

  Also, masking helps lessen but doesn’t prevent the spread in all cases. Many of the cultures that mask when sick do so because they are going to work or are out and about while sick.

• ⁨Comment⁩ on ⁨There should be a term for people who never really returned from the pandemic's social isolation⁩ ⁨⁨6⁩ ⁨months⁩ ago⁩:

  I think their point is that, with effort, it can be become a thing of the past.

  However, so many are unwilling to put forth the effort because it’s either too inconvenient or they’ve been brainwashed into believing it’s a hoax.