Re:Stealing Part of a Production Language Model
by imjonse on Tuesday March 12, 2024 @01:50AM
(#39676562)
Attack here is a term specific to cryptography and infosec, it does not imply doing anything illegal or violent. "Oracle attack" is the more specific term where you probe a system to give up some of its internals it was not meant to expose.
Re:Stealing Part of a Production Language Model
by TeMPOraL on Tuesday March 12, 2024 @05:03AM
(#39677491)
Wording matters. To use LLM terminology, "attack" here may mean something neutral, but it's helluva closer to "fear" and "malice" and "evil" in the latent space than terms like "probing", "studying", "examining", or "reverse engineering".
FWIW, cryptography and infosec as fields get a good mileage out of exploiting fear.
Re:Stealing Part of a Production Language Model
by imjonse on Tuesday March 12, 2024 @05:54AM
(#39677747)
That is only true in everyday latent space/vocabulary. In the ones underlying arxiv papers, cryptography and LLM, attack has the latter meaning.
Re:Stealing Part of a Production Language Model
by TeMPOraL on Tuesday March 12, 2024 @06:20AM
(#39677877)
Perhaps, but everyday vocabulary is what public policy and law discussions happen in, and I think that's what 'userbinator is worried about (and so am I).
See also: "piracy is stealing" or the everyday vocabulary meaning of the word "hacker".
Re:Stealing Part of a Production Language Model
by hnfong on Tuesday March 12, 2024 @01:06AM
(#39676368)
We used to call this "reverse engineering". I'm not familiar with what the law says on reverse engineering, but calling it "stealing" seems a bit too much.
Re:Stealing Part of a Production Language Model
by imjonse on Tuesday March 12, 2024 @01:48AM
(#39676553)
"Reverse engineering part of a production language model" has a more boring ring to it though.
Re:Stealing Part of a Production Language Model
by charcircuit on Tuesday March 12, 2024 @01:22AM
(#39676443)
Do you consider blind SQL injection an attack? If a user is not meant to have access to information, but they can somehow access it, that is a vulnerability.
Re:Stealing Part of a Production Language Model
by idle_zealot on Tuesday March 12, 2024 @01:44AM
(#39676527)
Sure, except the information here is not other users' account info or something, it's analogous to the source of the DBMS. Which is not typically considered sensitive.
Re:Stealing Part of a Production Language Model
by charcircuit on Tuesday March 12, 2024 @01:56AM
(#39676587)
No, it's like the schema of the database which is normally a secret.
Re:Stealing Part of a Production Language Model
by bugglebeetle on Tuesday March 12, 2024 @01:49AM
(#39676558)
Tell that to Oracle…
Re:Stealing Part of a Production Language Model
by ixaxaar on Tuesday March 12, 2024 @01:17AM
(#39676421)
The United States.
Re:Stealing Part of a Production Language Model
by ronsor on Tuesday March 12, 2024 @01:05AM
(#39676360)
OpenAI has blatantly said that the "open" in their name was a deceptive marketing ploy. At this point in time, they aren't interested in sharing much if any real research, and trying to discover this information is now an "attack" against them.
Re:Stealing Part of a Production Language Model
by astrange on Tuesday March 12, 2024 @03:16AM
(#39676939)
OpenAI didn't write this paper.
Re:Stealing Part of a Production Language Model
by dr_kiszonka on Tuesday March 12, 2024 @04:32AM
(#39677336)
One of the co-authors is from OpenAI.
Re:Stealing Part of a Production Language Model
by yreg on Tuesday March 12, 2024 @06:25AM
(#39677911)
> "open" in their name was a deceptive marketing ploy
It is now, but it hasn't been since the beginning. Hence the elon lawsuit.
It is considered an "attack" to probe at something to understand how it works in detail.
In other words, how basically all natural science is done.
What the fuck has this world turned into?