Sabitlenmiş Tweet
I don’t know that I’ll delete my account here, but I’ll certainly be more active on Bluesky. It is currently feeling much more like Twitter did than Threads does.
bsky.app/profile/coty.b…
English
Coty Rosenblath
25.3K posts
Coty Rosenblath
@Coty
More active on Threads and Bluesky.
Atlanta, GA Katılım Temmuz 2006
4.6K Takip Edilen2.6K Takipçiler
Coty Rosenblath retweetledi
Something that people on this platform apparently need to hear, even if it makes them feel uncomfortable: if you live in a state that does not believe that transgender rights are human rights, you will never be a true alternative to Silicon Valley -- let alone the next one.
English
Coty Rosenblath retweetledi
Patricia and Mark McCloskey brought weapons to a protest. They were honored as keynote speakers at the 2020 RNC.
English
Coty Rosenblath retweetledi
I tend to view code as more of a liability than an asset. In this light, making it cheaper and faster to generate a lot of code might not be an unmitigated blessing.
English
Coty Rosenblath retweetledi
Coty Rosenblath retweetledi
Coty Rosenblath retweetledi
People often assume anonymity on the Internet enables speaking truth to power often enough to make up for it allowing power to speak falsehood to the masses in the guise of revolutionaries. I remain unconvinced by this tradeoff.
English
Coty Rosenblath retweetledi
Stuff idiots love to say:
>physiognomy
>phenotype
>ontologically evil
>directionally correct (i.e. wrong)
>suicidal empathy
>midwit
>spiritually ___
>from first principles
>Nietzschean
>agentic
>you can just do stuff™
>longhoused
>schizo- (as a prefix)
>esoteric
>builder
English
Coty Rosenblath retweetledi
I'm hiring a Staff Developer Relations Engineer who will help engineers be successful with using AI.
Fully remote from anywhere Block is legally able to hire from.
If you're a good fit, apply (please no DMs).
Pls share for reach 🙏🏽
block.xyz/careers/jobs/4…
English
After reading many of the replies, we would like to issue a few clarifications:
- we cannot extract training data from the model using our method
- LLMs are not injective w.r.t. the output text, that function is definitely non-injective and collisions occur all the time
- for the same reasons, LLMs are not invertible from the output text
we hope this clears up any confusion and we welcome any feedback on the matter.
For any further questions, feel free to reach out to the authors:
@GiorgosNik02, @tommaso_mncttn, @DonatoCrisosto1, @teelinsan, Yannis Panagakis, @EmanueleRodola
GLADIA Research Lab@GladiaLab
LLMs are injective and invertible. In our new paper, we show that different prompts always map to different embeddings, and this property can be used to recover input tokens from individual embeddings in latent space. (1/6)
English
Coty Rosenblath retweetledi
this post is complete misinformation
LLMs are lossy compressors! of *training data*.
LLMs losslessly compress *prompts*, internally. that’s what this paper shows.
source: i am the author of “Language Model Inversion”, the original paper on this
English
@andrewthesmart @colin_fraser So you're saying I shouldn't give this too much weight? x.com/garrytan/statu…
English
@colin_fraser garry tan: well known interpretability researcher.
English
I think people aren’t quite understanding what this paper says
English
Coty Rosenblath retweetledi
Coty Rosenblath retweetledi
This is the most accurate descriptor of the new American reality that I've seen to date.
English
Coty Rosenblath retweetledi
This is what it feels like to be an LLM.
shaggy@shaggysurvives
who made this? its so well put and interesting and funny, i just happened to find it on my favorite short form video content platform
English
@ShriramKMurthi @paulg I am old enough to remember when Elon was sure that eliminating bots would be trivial.
English
@paulg The absolutely unhinged levels of racism are almost comical. Just mention certain countries/nationalities/etc. and it's enough to get the hordes going. I can only imagine some non-trivial % of these are actually bots. At least, one has to hope.
English
There is something deeply wrong with Twitter. It has always been rough here, but in the past year it has become an even worse kind of nasty. Do you think maybe it's time to try to turn things around, Elon?
Adnan Hussain MP@AdnanHussainMP
Yesterday I became a father. I shared a photo of my newborn daughter, and many of you sent beautiful messages. Thank you. But I’ve had to delete it. The vile racism and hate directed at an innocent soul less than a day old was beyond depraved.
English
@janecoaston I'm rooting for him to be part of a specific group because I am more fearful of the backlash if he is part of a different specific group.
English
If you are rooting for a murderer to be a member of a specific group you're already mad at for culture war reasons, go outside. Look at some trees.
English
Coty Rosenblath retweetledi
The writers of Black Mirror couldn’t possibly imagine the sheer inconsequentiality of our dystopian devices. An AI necklace that surveils you and everyone around you 24/7 so that it can send you a text that says “soo that just happened”
Kylie Robison@kyliebytes
We reviewed the Friend pendant, and unsurprisingly hated it! "It is an incredibly antisocial device to wear," I wrote of my experience wearing it. "People were never excited to see it around my neck."
English
@DavidAFrench If a politician who has the ability to promote policy change offers nothing but "thoughts and prayers" in the aftermath of a school shooting, does that not suggest that they think prayer is "enough"? Have you really not seen a single example of such?
English
I don't know a single person who thinks prayer is "enough." But it is indispensable. This is why we pray (I wrote this after the Covenant shooting): nytimes.com/2023/03/29/opi…
Jen Psaki@jrpsaki
Prayer is not freaking enough. Prayers does not end school shootings. prayers do not make parents feel safe sending their kids to school. Prayer does not bring these kids back. Enough with the thoughts and prayers.
English
Coty Rosenblath retweetledi
