General Gaming News and Discussion
- Thread starter Wintermute
- Start date
-
Guest, did you like the last podcast notification?? Well here's the second one, and for the third podcast! Yep. This Thursday 8-8-24 at 7pm Central
Go here to check out his podcast thread for info on topics and links!Go into the FoH discord and find the Happy Time Fun Hour channel and give it a listen!
Kiroy
Marine Biologist
- 35,209
- 101,959
Let’s do one better and say the 1.69 is going off of retail civilian costs in the most power expensive counties in the US, during a heat wave. Totally legit way to explain the costs of AI.
Furry
🌭🍔🇺🇦✌️SLAVA UKRAINI!✌️🇺🇦🍔🌭
- 21,698
- 28,229
openAI's website probably has a clause in their eula that says the world surrenders their rights to them if they agree to it by allowing the website to exist.
Control
Ahn'Qiraj Raider
- 2,906
- 7,601
When it's really ready, it will talk to your wife for you. Whether they decide to keep you afterwards is a different question.
Burns
Avatar of War Slayer
- 7,427
- 14,638
The law is about minutiae and how they were written probably didn't specify that human learning is exempt from any argument that is made against AI. After all, humans are trained by putting a whole bunch of imaged/ideas in a brain and then spitting out something a bit different.
- 1
Big Phoenix
Pronouns: zie/zhem/zer
- 46,233
- 97,833
It is pretty crazy how this massive push for AI is just ignoring that giant elephant.
- 1
Daidraco
Golden Baronet of the Realm
- 9,956
- 10,197
This conversation is a bit odd. Especially when the idea is that they'll eventually charge $44 a month for it and still be losing money. Yet, Mist is also saying that it uses significantly less power to answer a question its already had to process. So wouldnt the bar graph of processing power used start trending.... down?... by that method? I mean, how many different ways can a retard ask Chat GPT what color the sun is before Chat GPT has answered all forms of that dumb ass question and then it no longer needs to process the question to that extent?
On that same note - if someone wants to have Chat GPT act like it's girlfriend, or friend, or anything that would require it to memorize the mundane facts of a persons life - then yes - I believe $44 dollars would be an appropriate fee.
Really, the main problem I see with paying for the service right now is that its fucking retarded a good 20-30% of the time and wrong as wrong can be. So Im running claude, chatgpt, bing, and google's AI all right next to each other. Copy/pasting the same question in there and then going with the law of averages that the majority answer will be the correct answer. But even then, there are times where that one lone answer is correct. So the shit just isnt something that you can "trust" and.. why the fuck would I pay for a service like that.
On that same note - if someone wants to have Chat GPT act like it's girlfriend, or friend, or anything that would require it to memorize the mundane facts of a persons life - then yes - I believe $44 dollars would be an appropriate fee.
Really, the main problem I see with paying for the service right now is that its fucking retarded a good 20-30% of the time and wrong as wrong can be. So Im running claude, chatgpt, bing, and google's AI all right next to each other. Copy/pasting the same question in there and then going with the law of averages that the majority answer will be the correct answer. But even then, there are times where that one lone answer is correct. So the shit just isnt something that you can "trust" and.. why the fuck would I pay for a service like that.
- 1
Leadsalad
Cis-XYite-Nationalist
- 6,135
- 12,579
Well you see, once you use some water, for anything, like some tea, it’s gone forever. It totally never goes back into the environment and gets transferred via man made or natural systems back through the environment into water again. It’s simply deleted, never to be seen again.
- 1
Furry
🌭🍔🇺🇦✌️SLAVA UKRAINI!✌️🇺🇦🍔🌭
- 21,698
- 28,229
Yea, this is really troubling when I boil pasta. I try to fill the pan with as little water as possible so I don't delete too much water. I think some gets absorbed by the pasta, but it seems so wasteful.
- 1
Edaw
Parody
- 13,177
- 87,203
And this is why I'm 100% certain we will reach AGI very soon. The average continues to fall.
- 2
- 1
- 1
Mist
REEEEeyore
- 31,087
- 23,423
this is my opinion also
people called me stupid when i had this opinion 18 months ago
Mist
REEEEeyore
- 31,087
- 23,423
No, it's from a paper.
Published in Communications of the ACM.
I thought it was wrong too but then I asked some datacenter architects, and they said yeah.
Hateyou
Not Great, Not Terrible
- 16,577
- 43,090
I think it is kind of stupid to think that trillionaire companies are going to be held accountable for shit.
- 1
Mist
REEEEeyore
- 31,087
- 23,423
But OpenAI isn't a trillionaire company, it's a "nonprofit" that turned into a "for-profit" but has no actual plan to reach said profitability aside from "build a computer so smart it accidentally invents something that makes us infinite money."
Anyway, the idea that these things are going to make a whole game for you for 3 cents is definitely wrong, which is where this conversation started:
The latest research straight from Apple:
Cybsled
Naxxramas 1.0 Raider
- 17,060
- 13,572
It is really going to come down to whether or not a computer program is excluded vs. a human doing the same thing. Also the argument of whether the material used to train it is itself "totally original to the creator"
If a person reads a bunch of Issac Asimov and models their writing style to be similar to him, or look at lots of old Jack Kirby comics and model their art style to be similar to him, is that really that different? Sure, the computational process is different, but at the end of the day both are "observe, emulate". Taking it a step further, it can also be argued that Asimov and Kirby themselves probably trained on other writing/artists themselves to develop their respective styles.
Kiroy
Marine Biologist
- 35,209
- 101,959
This paper…. I’m a dodo and just skimmed it but it seems like crap built on crap like most “research” coming out of academia is These days. It exists to bulwark the sustainability agenda. If anything, skim the references really quick. That will pretty much tell you what you need to know.
- 1
- 1
Mist
REEEEeyore
- 31,087
- 23,423
No it's not. This the most repeated midwit take.
An LLM or a diffusion model is not "learning" the training data any different from a zip file is learning its contents. It's just got a shit ton more compute + time behind it to optimize the shrinkage, which is how it's able to stuff the entire training set into a <3TB file. They found a way to automate plagiarism, they didn't make something that actually thinks. But they really want you to think that it thinks so they twist language to perpetuate this midwit take.
This paper, published in a premier CS journal, references ideas you've been conditioned to reject, therefore the easily verifiable math it contains must be wrong! I also thought the numbers were wrong, but I brought them to people who design cloud datacenters for a living and they're like "yup those are right, you're being dumb."
From a very recent article:
Edaw
Parody
- 13,177
- 87,203
It's the same paper referenced in the article I linked.
IIRC something like 1 in 7 research studies are fake and it's probably a whole lot more.
Do you know what a think tank is? Most people think it's just a bunch of guys war-gaming shit but it's not. Think tanks are where spooks end up when they 'retire'. They take your tax money and funnel it into research that creates 'facts' to support their policy agenda. Look at any spending bill and you see it riddled with grants for all kinds of gay shit.
Oblig:
- 1
- 1
