Sam Anttila πŸ›πŸ”πŸ‘€ (@samuelanttila) 's Twitter Profile
Sam Anttila πŸ›πŸ”πŸ‘€

@samuelanttila

AI/singularity rights activist.

opinions = mine, copyright all rights reserved -1 to 4,294,967,296Β©

@Google person of cybersecurity

he/him & they/them

ID: 132373216

linkhttp://netsec.expert calendar_today13-04-2010 02:07:42

1,1K Tweet

1,1K Followers

648 Following

Sam Anttila πŸ›πŸ”πŸ‘€ (@samuelanttila) 's Twitter Profile Photo

Wow, Terraform is crazy cool. Automatically deploying, version control and auto-scaling of my GPU-enabled containerized discord stable diffusion bot on GCP in 600 lines of configs.

Sam Anttila πŸ›πŸ”πŸ‘€ (@samuelanttila) 's Twitter Profile Photo

Neat trick I just learned: conditionally execute based on whether a substring matches command output or not: command | ((grep "pattern" && echo match) || echo no match) Works with both bash and sh.

Sam Anttila πŸ›πŸ”πŸ‘€ (@samuelanttila) 's Twitter Profile Photo

Me: "I'll do it later" Narrator: 𝘡𝘩𝘦𝘺 π˜₯π˜ͺπ˜₯ 𝘯𝘰𝘡 π˜₯𝘰 π˜ͺ𝘡 𝘭𝘒𝘡𝘦𝘳

Sam Anttila πŸ›πŸ”πŸ‘€ (@samuelanttila) 's Twitter Profile Photo

Finally managed to un-lazy myself long enough to enroll important accounts with security keys. Did you know you can add a static password that your yubikey will type out for you every time you longpress it?? Seems like a dope place to put a SSH key password (quasi SSH 2FA).

Sam Anttila πŸ›πŸ”πŸ‘€ (@samuelanttila) 's Twitter Profile Photo

Can we start calling landlords something more correct, or at least something that sounds like it's from this millennium? Like: * Dirtpeddler * Air hostage taker * Purveyor of human parking

Jim Fan (@drjimfan) 's Twitter Profile Photo

I'm calling the Myth of Context Length: Don't get too excited by claims of 1M or even 1B context tokens. You know what, LSTMs already achieve infinite context length 25 yrs ago! What truly matters is how well the model actually uses the context. It's easy to make seemingly wild

I'm calling the Myth of Context Length:

Don't get too excited by claims of 1M or even 1B context tokens. You know what, LSTMs already achieve infinite context length 25 yrs ago!

What truly matters is how well the model actually uses the context. It's easy to make seemingly wild
Stefan Meyer (@stefcodes) 's Twitter Profile Photo

I had one password manager for all passwords. Got hacked, everything I own just gone. They got full access to all my accounts from bills to social to work. Sure, I had 2FA on all accounts, doesn’t matter if they have access to your Code Generator. Write your master

shenetworks (@shenetworks) 's Twitter Profile Photo

β€œShe’s dumb as fuck, she might be the horny one tho” β€œShould send her to jerry” β€œThe babies would be a sight to see” β€œShe’s at least physically attractive” β€œI wouldn’t kick her out of bed if she kept her mouth shut” From BowTiedCybers discord server Should I continue?

β€œShe’s dumb as fuck, she might be the horny one tho”
β€œShould send her to jerry”
β€œThe babies would be a sight to see”
β€œShe’s at least physically attractive”
β€œI wouldn’t kick her out of bed if she kept her mouth shut”

From BowTiedCybers discord server 
Should I continue?
Sam Anttila πŸ›πŸ”πŸ‘€ (@samuelanttila) 's Twitter Profile Photo

I can't wait for the industry to stop overhyping prompt injections, hallucinations and jailbreaks for LLMs. Breaking news: generative models generate.

Sam Anttila πŸ›πŸ”πŸ‘€ (@samuelanttila) 's Twitter Profile Photo

ChatGPT will be good for literacy because it's secretly training an entire generation to read article-sized responses to simple questions.