471–480 of 719 posts
I wonder what the memory requirements would be to run such a large model. I'd love to be able to run this model, alas my MacBook can barely run toy models.
You would need over 65GB of RAM. There are consumer GPUs that have 48GB of RAM, and can be tethered together with NVLink. I wonder if that would work.
Earlier quoted context omitted.
As far as I can tell you can do it right now, at least for small 13B model, not sure about bigger models.
I don't believe they could, need CUDA and more VRAM...
Earlier quoted context omitted.
It's funny that part of the 4chan excitement over this is that they think they'll get back the AI girlfriend experience of when character.ai was hooked up to uncensored GPT-3. All that has been thoroughly shut down by character.ai and Replika and they just want their girlfriends back.
I'm curious if the blocking of adult content has to do with moralism, commercial interests, or something deeper. An eager to please conversational partner who can generate endless content seems quite dangerous and addictive, especially when it crosses over into romantic areas. There's already posts of people spending entire days interacting with LLMs, using as their therapist, romantic partner, etc. Combined with fin…
I genuinely fear that the breakdown of millennia old social structures that kept us human might lead to a temporary (century long) turmoil for individuals. The answers to the 'meaning of life' and 'what makes us human' are going to change. And we will never be the same again.
This isn't just about AI. External wombs, autonomous robots, genetic editing & widespread plastic surgery each fundamentally destroy individual aspects of 'what makes us human' or 'the meaning of life'.
Might be for the best. But such drastic change is really hard for the fragile human brain to process.
Earlier quoted context omitted.
My understanding is that weights are normally stored as pickled python blobs, which means arbitrary code execution as they are unpickled.
You're right! You should probably use Trail of Bits Fickling tool to investigate. https://github.com/trailofbits/fickling
$ fickling --check-safety consolidated.00.pth
File "/usr/lib/python3.10/pickletools.py", line 359, in read_stringnl
data = codecs.escape_decode(data)[0].decode("ascii")Earlier quoted context omitted.
Immortal as long as someone's paying to run the instance.
I bet soon after the first few people are made immortal this way, one of them will hack the banks, or the stock market, or countless other organizations.
Which is just one step closer to the simulated hell for uploaded consciousnesses that get naughty, from Surface Detail by Ian Banks.
Earlier quoted context omitted.
> It’s interesting that these models are both massively expensive to produce and self-contained to a degree that you can distribute the end product in a torrent. I was trying to come to grasp with how much resource there is concentrated in one of these models. Somehow I come to the conclusion that it cost more than buying a jet airliner to train one of these models. And it is about the same order of money as commissi…
For anyone curious, it took 2048 A100 GPUs to train LLaMa, each GPU costs roughly $15k, facebook probably gets some sort of discount. That's a $30Mil if you want to train at that scale. Also IIRC it took 23 days to train the biggest model. Someone else can do the power consumption cost calculations.
At least $10 million/yr just for the talent.
The smallest model (7B) is supposed to outperform GPT-3. Does anyone have any idea what hardware is needed to run this?
Supposedly double the model size so 14gb. RTX 4090 might be able to handle it. You can use lambdalabs to rent a server gpu for one of the larger models.
Earlier quoted context omitted.
It became something of a meme but there are huge numbers of guys out there that would pay good money for Joi from Blade Runner 2049. https://bladerunner.fandom.com/wiki/Joi
One thing I liked in 2049 was now they made the holographic projector seem more mechanical and less hand wavy with the roof attachment tracking along with the girl. Makes it seem more like something in reach rather than pure scifi.
Earlier quoted context omitted.
The leaker accidentally doxxed themselves by adding the original download script to the torrent: https://boards.4channel.org/g/thread/91848262#p91850503
What are the possible consequences here?
Earlier quoted context omitted.
When we can upload our brains to the cloud, and you can do something with them like interacting or running the brain, then we'll all be effectively immortal. That's a pretty big deal. See the book altered carbon.
Immortal as long as someone's paying to run the instance.