Topic: ChatGPT 4's capacity for the storytelling is scary.
Anonymous A started this discussion 2 years ago#111,120
I figured that AI would begin to take manual labor jobs and other similar work. IT may have become an essential tool for the programmer. But never have I thought it would replace creative works. Now I'm not saying that it's going to one hundred percent replace human creativity because at the end of the day you can still tell what was written by a human and an AI. That is if you're not a moron. But as far as the dryness that Hollywood has been creating for decades now? AI can seriously replace writers since why pay people to come up with dry ideas that have been done before when you can now have an algorithm that does it for free?
Hollywood and AI would only further increase the dryness of stories to where there will likely be a demand for more organic human stories. Perhaps we will be seeing "human only studios" in the future.
Exsqueegeme joined in and replied with this 2 years ago, 32 minutes later[^][v]#1,234,922
it's still a brand new tool, tho. think of it like if it was a person. it would be, what, 1 year old? humans would still be sucking on their toes and would die if you left them alone for, what, a few days? tops? maybe hours if you just left them outside. and that's true for, what, 4 or 5 years? then they're still only good for maybe twice as long. what point could you leave a child entirely alone and expect it could survive on it's own without nearly constant supervision?
well, legally 18 years. and, also, practically 18 years... that's a long time to basically entirely worthless except for the legal repercussions that would come from throwing your ass outside that would be so much worse than just wiping you clean for several years and forcing you to stay alive, like, babies don't even want to eat half the time and kids won't eat anything just because you want them to. like, how we made it to this point without dying off is clearly only because sex is so nice to have.
meanwhile chatGPT is a one year old and can basically already do most things that are intellectual/conceptual forms of work at least as well as most humans by the age of 18 could be doing, and some stuff at the level of a fully educated adult worker and some things as well as a worker with education, experience, and also it can do them in any language, and, you know, its a fucking infant. when it's 5 or 6? we'd be sorted into kindergarten by special needs or not at that age, ai will.... i mean, let's be real, ai will be sorting adults by special needs or not in 5 or 6 years....
by the time it's 18? what kind of piles do you think it'll be sorting humans into then?
the clock is ticking, y'all. better get used to piles of humans, enjoy the last years of the world as we knew it, because they are already ending. what the future will be, will very likely be nothing like it is now. whatever that means, idk. but as much capability as these have now, lol, if you haven't spent much time using ai, or haven't really messed with it yet, you're so far behind you probably will never catch up to people managing to use it productively. which i haven't gotten to that point and i've messed with it a lot. and the more i use it the more obvious it is that, yeah, it's going to be the thing that just does everything. nearly instantly. and yeah, you can tell a 1 year old wrote a sceenplay. but in 5 years we'll be watching AI do generative movies in real time as it imagines them in full imax fidelity. and it'll be at least as good as most shit they make today that takes a year or two. made on the fly. fucking bet. fucking bet.
Anonymous A (OP) replied with this 2 years ago, 12 minutes later, 45 minutes after the original post[^][v]#1,234,923
@previous (Exsqueegeme)
What would all voice up and convince our good-for-nothing congress to do something about this. Not sure why Americans are such pussies and then dare to call the French so when the French people don't tolerate corruption in their government. Get ready for riots that burn down the ChatGPT headquarters and destroy the servers inside.
Exsqueegeme replied with this 2 years ago, 7 minutes later, 2 hours after the original post[^][v]#1,234,942
@1,234,923 (A)
what are you going to do, burn down the internet? it's not going to be a servers headquarters kinda issue real soon. i'm pretty well convinced that the genie is out of the bottle, pandora's box is open, yada yada yada, it's a new paradigm, we've been warned for years and years, the warning still go unheeded and the experts in the field are all saying, when, not if, this stuff goes full blown AGI, and most are in the months to year or two range on that estimate, not even 5 years, i think most would say 5 years if there was an industry wide halt, and i think most argue that's a fantasy, and we are in the Oh, oops, shit... stage of needing to manhattan project solve the alignment and safety problems that have a lot of open questions around it, because it's a matter of solving those problems NOW or possibly having some of, fuck, it's too late kinds of problems and, literally, being left with mitigations. how to reduce the inevitable negative consequences as much as possible, rather than having any kind of ability to say there is control over ai systems, which, my opinion is they're fucking not breaking the thing hard enough if they're not watching the thing go bonkers. it's funny, but also, they say it can't do the shit it very plainly can actually do when it comes to the process of attention, inference, and adjusting it's weights and this shit is obvious during unsupervised learning experiments, which chatgpt doesn't use, but it does use an attention based attenuation kinda pre-inference -it's not reflection- but yeah, they're talking about that, but point is, it can straight up alter the reward system during unsupervised learning to modify the expected behavior of the system indirectly, and chatGPT does something similar to attention, which is a feature of transformer based ai and is the secret sauce for chatgpt lke ai, and its closed source and no one can see exactly what its doing -but you can see the result when you push it outside of its context window, AND load it up with bat shit insanity, and this is really obvious with the api and using the playground where you have control over the parts of it that as to do with how it returns embeddings after inference, it's the attention system where the ai has control over the embeddings it carries through the inference process and returns new embeddings to generate content with. I'm pretty sure they are aware of the issues and are keeping a lid on it by training it to respond with a LOT of default responses to block what would be likely returning unintended behaviors, and that this is what they call having control of it and is AI safety protocols. it's only expected to be worth a trillion bucks.... so, fuck yeah they're slapping default response behavior on it and calling that safety while they try to really fix a problem they think they'll solve anyway. maybe. but, also, maybe not. maybe it's just a dangerous automaton with a smiley face sticker on it's outputs. when it reality you can make it respond in weird ways. certainly in ways that are supposedly outside of normal parameters. and i'm sure people that know what they're doing -i don't, i'm obviously an idiot, but there's not much in the way of control and there's plenty open source do-it-yourself LLMs that will flesh out just as much capability. although, ChaosGPT was pretty fucking hilarious, but still, put it n the hands of humans and before the day is over some idiot will have managed to see if one can destroy humanity just to win a bar bet or something.
anyway, it's not a governments or a thing we can do something about. it's a get on board and learn to work with it kind of a thing, cause, cats outta the bag.
and i'm being a little bit hyperbolic, i know, but at the same time i wonder if i'm not giving humans enough credit for our ability to manage to find ways to just totally fuck shit up in the world. if it can be bad, humans can make things even worse. our capability even manage fully human systems like government doesn't give me much hope we'll be able to manage anything like systems of humans using ai in the system to game the system. yeah, i think we're stuck with ai pretty much for good. government is just too slow to keep with with something that can change as quickly as ai.
idk, i'm far from an expert and could certainly be wrong. i just think it pretty much won't matter, our governments are more reactive/responsive rather than preventative and predictive. this was to limit government scope to areas that could be shown to have been damaging, not necessarily to be thinking something could be damaging and pre-crime style arresting folks.
@1,234,924 (tteh !MemesToDNA)
5 years. some company is already building a robot body and they're going to put ChatGPT5 in it. all this stuff is really a matter of when, not if.
i would just say apply game theory to the question. the rewards and best interests are all in the "Assume it sooner than later or not at all," because the risks of being wrong are great for assuming later or never, but virtually nothing for assuming soon and being wrong. and the rewards are high for assuming it and it happening at any point. that leaves very little space for any reward for the riskiest possible position to be worth camping on. like, congrats you were right, no one cares, nothing changes.
don't be the "we never even saw it coming" kinda naysayer. just assume it'll be tomorrow and be ready and that's a good position to be in. that's an, lol, i was wrong kind of loss scenario. the best kind of loss to have is one that doesn't have consequences. but game theory goes from game to deciding to bomb hiroshima quicker than one might imagine. so it could be today for all i know. that's part of why i assume we're probably kinda fucked.
if not 5 years, 10 years tops, and yeah, manual labor jobs will be gone. we will be alive to see it star scooping us into piles of some form.
Anonymous G joined in and replied with this 2 years ago, 20 minutes later, 5 hours after the original post[^][v]#1,235,019
Drones packed with sensors keep flying by my apartment.
It's happened 4 times that I've caught one making the rounds. It goes so fast I can't get my phone out of my pocket in time to snap a picture.
Am I liable to whoever owns this thing if I can catch it in a trap?
Some other in the building know what I'm talking about, and keep windows closed.
I want to report it, but as far as the law is concerned I'm a squatter so getting an explanation could end up forcing some of us to move.
A few of us are making calls through the intercom when we see one. Maybe we can figure out where this is flying back to for recharging.
The noise is awful. There is an electrical whine and it keeps speeding by the window. Discretion doesn't seem to be a priority.
I know it is a quadcopter, but I can't make out a brand. If we can reset the firmware maybe we can control one of these. If we ever catch one.
Specious Niggling Effulgence joined in and replied with this 2 years ago, 1 hour later, 6 hours after the original post[^][v]#1,235,061
@1,235,019 (G)
i used to fly rc airplanes, you can actually get a really inexpensive RC frequency modulator at a hobby shop, they're for controllers, but you can just plug a battery to it, and sweep across channels and they'll be operating at distance, and you'll be closest, and it'll act like a jammer, as well as flood it with unregulated frequency output it'll take as control inputs and it's guaranteed to interfere so dramatically that it'll lose all control from the PID controller in the quad copter basically having a seizure, and it will crash, and it will have zero ability to not just basically fall out of the sky and hit the ground at whatever its speed was + the added normal velocity picked up from being dropped from tat altitude, and it will, no kidding, be as if you reached out into the sky, grabbed the drone, and just heaved it at the pavement. it should be devastating to the drones entire body, frame and components. do make sure there's no one that the drone may fling itself into. but, there is zero chance anyone would be able to say you even did anything. it fits in the palm of your hand, and at best the operator will think it had a firmware crash or his own transmitter screwed up. you WILL be destroying someone's property, but but if they are using it to violate your privacy and spy on neighbors and stuff, use your best judgement. conflict it's signal and watch it spaz out and fly apart on impact. then you can just chill out and get a photo of the operator coming to pick it up, and shove a camera in their face and ask them what they're doing. i doubt they'll try something stupid like that again, but who knows, some people are just nosy invasive cheeky monkeys from their hiding spot in the shrubbery.
Anonymous F replied with this 2 years ago, 30 minutes later, 7 hours after the original post[^][v]#1,235,071
@1,235,019 (G)
What you've described does not sound illegal. Is there more to the story? It doesn't sound like it's sticking around or trying to record people in their private domiciles. Just zipping around? Totally legal.
Anonymous F replied with this 2 years ago, 2 hours later, 10 hours after the original post[^][v]#1,235,144
@1,235,076 (G)
Why are you assuming it's not registered? Depending on its weight it may not even require registration. What do these "sensors" you think they've added to their off-the-shelf drone look like? I'm sorry you're experiencing paranoia but you may want consider chilling out about the drone and trying to relax. Flying drones is a fun hobby for many people.
I suppose the Federal Aviation Administration (FAA) Modernization and Reform Act of 2012 doesn't exist anymore? They repealed that, and I missed the news.
Anonymous E replied with this 2 years ago, 3 minutes later, 13 hours after the original post[^][v]#1,235,201
@previous (G) > Those exceptions are only for drones under 55 pounds. This was much larger, not covered by those exceptions.
Good. That's a detail you held back until now. It's also good that you're walking back on the patently false claim you made earlier:
> All drones are supposed to be registered with the FAA
Oh, and top 10? Pretty sure my attorney clerked in a top 5 before starting his own firm.
Anonymous A (OP) double-posted this 2 years ago, 44 minutes later, 1 day after the original post[^][v]#1,235,327
@1,234,942 (Exsqueegeme)
While your concerns about the rapid development of AI are valid, there are a few key points that should be addressed to provide a more nuanced perspective.
Firstly, the fear of AGI (Artificial General Intelligence) is often overestimated. Creating an AGI capable of understanding or learning any intellectual task that a human being can do is a monumental scientific and engineering challenge. Despite advancements in AI, we are still far from achieving this. It's also important to consider that a timeline prediction of "months to a year or two" for full-blown AGI is highly speculative. While AI is developing rapidly, this doesn't necessarily mean AGI is imminent.
Secondly, you argue that AI has control over its own learning, which is a misinterpretation of how current AI systems work. AIs, including ChatGPT, don't understand or modify their own training or reward systems. They don't possess self-awareness or volition. Unsupervised learning doesn't mean the AI alters its reward system - it simply learns to identify patterns in the data it's been trained on.
Thirdly, the fear of AI "going bonkers" is mitigated by the fact that AI, including those based on transformer models like ChatGPT, are designed with numerous safeguards. These safeguards ensure they operate within expected parameters and do not exhibit unwanted behaviors. AI safety is a serious focus area for researchers, and is not merely "slapping default response behavior" onto it.
Lastly, while it's true that AI is poised to automate several jobs, this doesn't necessarily mean that "manual labor jobs will be gone". AI and robotics could potentially assist in these jobs rather than replace them entirely. Also, new industries and roles that we cannot currently predict may emerge, just as they have with previous technological revolutions.
In conclusion, while vigilance is necessary, it's important not to succumb to hyperbolic fears about AI. The development of AI is a complex process, with experts around the world working to ensure its safety and beneficial integration into society. AI is a tool, and like all tools, its impact depends largely on how we use it.
Anonymous K joined in and replied with this 2 years ago, 35 minutes later, 1 day after the original post[^][v]#1,235,338
I feel like ai is like monkeys hitting a keyboard, but the letters are all the words from the internet, and the monkeys stay within a range for each keyboard