I’d step back at that launch
Mastodon: @greg@clar.ke
I’d step back at that launch
That’s 128GB RAM, the GPU has 24GB VRAM. Ollama has gotten pretty smart with resource allocation. Smaller models can fit soley on my VRAM but I can still run larger models on RAM.
I’ve installed Ollama on my Gaming Rig (RTX4090 with 128GB ram), M3 MacBook Pro, and M2 MacBook Air. I’m running Open WebUI on my server which can connect to multiple Ollama instances. Open WebUI has it’s own Ollama compatible API which I use for projects. I’ll only boot up my gaming rig if I need to use larger models, otherwise the M3 MacBook Pro can handle most tasks.
I’m running a search instance on a VPS so my home IP isn’t linked to my searches. The main disadvantage is that my VPS is in Toronto and I live 2hrs away so geo searches don’t work very well. For instance, if I Google “restaurants” I get results for local restaurants whereas if I Gregle (I named my search engine Gregle) I get results for results near my VPS.
DM me if you want a link to my instance to check it out. It’s open but I don’t publicize it because bad actors could ruin my IP addresses reputation with spam queries via the API.
The use of CSAM in training generative AI models is an issue no matter how these models are being used.
This is tough, the goal should be to reduce child abuse. It’s unknown if AI generated CP will increase or reduce child abuse. It will likely encourage some individuals to abuse actual children while for others it may satisfy their urges so they don’t abuse children. Like everything else AI, we won’t know the real impact for many years.
Jokes on you Slack, I’m not intelligent!
I forgot to post an affiliate link and explain how routing all your internet traffic though one company equals security
I’m 100% secure, I have Nord VPN
Maybe someone is sneaking into your room while you’re sleeping and stuffing dryer lint up your nose ¯_(ツ)_/¯
Oops I did it again by Britney Spears
This is exactly the situation 😅
This sounds really fun! I’m putting this on my long to-do list
I’m an immigrant from Australia and I remember having to change my word choices even though Canada and Australia are nearly culturally identical. It must be a lot more work coming from less culturally similar places.
I work remote, live in Belleville, and visit our Toronto offices every couple of months. My experience with VIA Rail has been positive though I appreciate that I have a lot of flexibility. I can work while on the train so the commute isn’t wasted time for me. Working while commuting is not something I could do if I drove or took a bus. The train can be late so I wouldn’t organize an in-person-meeting within an hour of the scheduled arrival.
I agree that the ticketing system should be improved to encourage commuting, it would be a headache to buy tickets individually weeks in advance if I were using the service more frequently.
Renting a 3 bedroom house in Toronto is more expensive than a similar house in Belleville + $1500 in monthly VIA Rail tickets. Plus lots of jobs are partly remote which makes the equation even more appealing.
Religion has competition now, AI is also profiting from hallucinations that lots of people believe as fact.
That’s fair, I can appreciate an attack vector in cases where there are bad blocks and the drive was unencrypted. Luckily bad blocks are less common with modern SSDs and assuming the disk was encrypted, a few bad blocks are unlikely to expose any contents. So knowing the number of bad blocks and what data was stored would inform if a fill and empty approach would be suitable to sanitize the drive.
Fill the drive 100% using data duplicator then delete everything on the drive. Repeat a few times to ensure you scrub all blocks. There is no need to physically destroy the drive.
edit: fair criticism of this approach in cases when the data is unencryptd and the hard drives has bad blocks. I just wanted to give a counter to the destroying hardware approach which isn’t necessary warranted
Fun fact, DEDSEC is a type of memory used in Soviet era mainframes.