@cigitalgem@sigmoid.social
@cigitalgem@sigmoid.social avatar

cigitalgem

@cigitalgem@sigmoid.social

software security #swsec machine learning security #mlsec Tech | Life | Music

This profile is from a federated server and may be incomplete. Browse more on the original instance.

cigitalgem, to random
@cigitalgem@sigmoid.social avatar

Congrats to Legit for raising an impressive series B. Proud to be an Advisor to the firm.

https://techcrunch.com/2023/09/20/legit-security-lands-40m-to-lock-down-apps-and-dev-environments/

cigitalgem, to random
@cigitalgem@sigmoid.social avatar

This is an awful indictment of ML-enhanced search and ML pollution on the open web, both. https://www.404media.co/first-google-search-result-for-tiananmen-square-tank-man-is-ai-generated-selfie/

cigitalgem, to random
@cigitalgem@sigmoid.social avatar

is an enormous challenge growing faster than we can comprehend https://www.nytimes.com/2023/09/20/technology/chatgpt-dalle3-images-openai.html

cigitalgem,
@cigitalgem@sigmoid.social avatar
matthew_d_green, to random

Can we just not process weird file formats people receive by iMessage/text?

cigitalgem,
@cigitalgem@sigmoid.social avatar

@SteveBellovin @adamshostack @matthew_d_green @jef given the choice between dancing pigs and security, pick speed

cigitalgem, to random
@cigitalgem@sigmoid.social avatar
cigitalgem, to random
@cigitalgem@sigmoid.social avatar

Are we already at the
place where nothing written after
2021 can be proven to be ML-free?

ct_bergstrom, to random

I just finished Cixin Liu's Three Body Problem trilogy.

The books left me with a new sense of gratitude to be part of the universe's great story, and a heightened sense of wonder at the marvel of being alive to feel clean sheets against cold skin, or to watch a flock of geese land in still water.

cigitalgem,
@cigitalgem@sigmoid.social avatar

@ct_bergstrom excellent series

ncweaver, to random

Remember, if you want to tweet out secrets that @briankrebs doesn't find, use hashtag

Because every day is

cigitalgem,
@cigitalgem@sigmoid.social avatar
SinclairSpeccy, to tech

On September 9, 1941, 82 years ago, Dennis Ritchie, the brilliant mind behind the C programming language and a key developer of UNIX was born.

C, which he wrote the book for it with Brian Kernighan, became the source for countless software applications and operating systems. UNIX, developed alongside Ken Thompson, laid the groundwork for modern computing.

Despite his passing on October 12, 2011, at the age of 70, Ritchie's legacy lives on

#C

cigitalgem,
@cigitalgem@sigmoid.social avatar

@SinclairSpeccy I once gave a talk on at bell labs. Dennis was in the front row. That talk included a "C is bad" chant as a silly hook (with audience participation). Dennis nodded his permission to proceed. The talk was really fun and was a harbinger of early static analysis tools.

cigitalgem, to random
@cigitalgem@sigmoid.social avatar

BIML on the porch. Discussing emergent computation and the history of AI.

image/jpeg

4Dgifts, to random

I never thought that being an actual Blade Runner (someone that retires out of control AIs) could be a possible role in my profession.
It is awesome!
Isn't it @cigitalgem ?

cigitalgem,
@cigitalgem@sigmoid.social avatar

@4Dgifts It's as easy as turning the computer off (so far).

cigitalgem,
@cigitalgem@sigmoid.social avatar

@4Dgifts they are still enjoying the fish.

cigitalgem, to random
@cigitalgem@sigmoid.social avatar

Doug Lenat down. We just read one of his very new papers at BIML this week.

I am not a fan of Cyc or symbolic AI, but Cyc was a hell of an experiment.

https://garymarcus.substack.com/p/doug-lenat-1950-2023?utm_source=substack&utm_medium=email

cigitalgem, to random
@cigitalgem@sigmoid.social avatar

The Atlantic on AI. Nice piece.
really needs to ditch the red teaming nonsense.
https://www.theatlantic.com/magazine/archive/2023/09/sam-altman-openai-chatgpt-gpt-4/674764/

cigitalgem, to random
@cigitalgem@sigmoid.social avatar

Interesting to see what the finance guys think about AI. Have a look at this mixed bag of solid analysis and total nonsense. https://www.institutionalinvestor.com/article/2c4fad0w6irk838pca3gg/portfolio/money-is-pouring-into-ai-skeptics-say-its-a-grift-shift

cigitalgem, to random
@cigitalgem@sigmoid.social avatar

Protopia video describing operational data protection for LLMs

https://youtu.be/Xvo7apHyslQ?feature=shared

malwaretech, to random

Finally got around to trying the new Delta 767-400. Nice layout and decent privacy but feels a tad squashed compared to Polaris

image/png

cigitalgem,
@cigitalgem@sigmoid.social avatar
cigitalgem, to random
@cigitalgem@sigmoid.social avatar

BIML via video.

Today we talked about Pandas, Sloths, and Auto-completion.

cigitalgem, to random
@cigitalgem@sigmoid.social avatar

You can't fix an LLM by red teaming. It does exactly what it was designed to do. Autoassociative predictive word generation.

So what do you prove when you do prompt injection? Not a damn thing.

Always ask this. How does someone FIX what comes out of a pen test? If there is no fix, there is no change in security posture.


https://www.washingtonpost.com/technology/2023/08/08/ai-red-team-defcon/?wpisrc=nl_technology202

cigitalgem,
@cigitalgem@sigmoid.social avatar

@RWerpachowski @andrei_chiffa @wuzzi23 the stochasticism at decision time is NOT the fundamental problem

cigitalgem,
@cigitalgem@sigmoid.social avatar

@andrei_chiffa @RWerpachowski @wuzzi23 I don't understand what you are trying to say.

cigitalgem,
@cigitalgem@sigmoid.social avatar

@andrei_chiffa @RWerpachowski @wuzzi23 I disagree with your assessment. I do understand your point, however.

In the case of LLMs the data set is immensely much larger as is the model. So awful aspects of the data are preserved during the training process. Because the data are not properly understood or curated, you get what you asked for. I think the same problems may emerge with different (but much larger) base models because the data are the big problem.

cigitalgem,
@cigitalgem@sigmoid.social avatar

@andrei_chiffa @RWerpachowski @wuzzi23 or more simply put, the stochasticism is not the problem, data opacity is.

cigitalgem,
@cigitalgem@sigmoid.social avatar

@andrei_chiffa @RWerpachowski @wuzzi23 I agree but think that what you are describing is currently a secondary problem. First order remains bad data.

cigitalgem,
@cigitalgem@sigmoid.social avatar
  • All
  • Subscribed
  • Moderated
  • Favorites
  • JUstTest
  • mdbf
  • everett
  • osvaldo12
  • magazineikmin
  • thenastyranch
  • rosin
  • normalnudes
  • Youngstown
  • Durango
  • slotface
  • ngwrru68w68
  • kavyap
  • DreamBathrooms
  • tester
  • InstantRegret
  • ethstaker
  • GTA5RPClips
  • tacticalgear
  • Leos
  • anitta
  • modclub
  • khanakhh
  • cubers
  • cisconetworking
  • megavids
  • provamag3
  • lostlight
  • All magazines