cigitalgem

cigitalgem, 4 months ago to llm

Lets do a TOP TEN LLM Risks list #MLsec #LLM #ML

8: Data ownership

Get the full paper here https://berryvilleiml.com/results/

cigitalgem, 6 months ago to random

Welcome @SecurityPrivacy. good to see you finally ditch X.

here is a complete archive of Silver Bullet security podcasts produced from 2006-2019. Never missed a month. About half were published in the mag.

https://www.garymcgraw.com/technology/silver-bullet-podcast/

cigitalgem, 3 months ago to random

ML chatbots can't tutor math #MLsec https://www.wsj.com/tech/ai/ai-is-tutoring-students-but-still-struggles-with-basic-math-694e76d3?st=606ux1exuaxyk3j

cigitalgem, 4 months ago (edited 4 months ago) to random

Lets do a TOP TEN LLM Risks list

#MLsec #swsec #ML #security

9: Model Trustworthiness

Get the full paper here https://berryvilleiml.com/results/

cigitalgem, 4 months ago to llm

META-thread: Lets do a TOP TEN LLM Risks list #MLsec #LLM #ML

1. Recursive pollution https://sigmoid.social/@cigitalgem/111822743461781680

2. Data debt https://sigmoid.social/@cigitalgem/111818853107254150

3. Improper use
   https://sigmoid.social/@cigitalgem/111818552669618098

4. Black box opacity
   https://sigmoid.social/@cigitalgem/111817591521152365

5. Prompt manipulation
   https://sigmoid.social/@cigitalgem/111817028774981796

6: Poison in the data
https://sigmoid.social/@cigitalgem/111813050795199210

7: Reproducibility economics
https://sigmoid.social/@cigitalgem/111812863118355119

8: Data ownership
https://sigmoid.social/@cigitalgem/111812699547902655

9: Model Trustworthiness
https://sigmoid.social/@cigitalgem/111812307709213718

cigitalgem, 4 months ago to random

More fascinating work by Brendan Lake at NYU. Outstanding cognitive science.

Apparently this is what Brendan looks like...

https://www.technologyreview.com/2024/02/01/1087527/baby-ai-language-camera/?utm_source=dlvr.it&utm_medium=mastodon

cigitalgem, 4 months ago to llm

Lets do a TOP TEN LLM Risks list #MLsec #LLM #ML

5. Prompt manipulation

Get the full paper here https://berryvilleiml.com/results/

cigitalgem, 4 months ago to llm

More coverage of our #LLM Risk paper (which dropped yesterday).

#decipher #deciphersec #MLsec #AI #ML #security

Nice work as always from @dennisf

https://duo.com/decipher/for-ai-risk-the-real-answer-has-to-be-regulation

cigitalgem, 4 months ago to random

Currently discussing "Will we run out of data?..." and relating it to our thinking about data feudalism at BIML.

#MLsec

https://arxiv.org/pdf/2211.04325.pdf

cigitalgem, 7 months ago to random

Are we entering the new age of data feudalism? I discuss the AI executive order, #swsec (#appsec) and #MLsec

https://duo.com/decipher/white-house-ai-executive-order-puts-focus-on-cybersecurity

cigitalgem, 4 months ago to llm

Lets do a TOP TEN LLM Risks list #MLsec #LLM #ML

4. Black box opacity

Get the full paper here https://berryvilleiml.com/results/

cigitalgem, 10 months ago to random

Repeat after me. AI "red teaming" is bullshit. Do real #MLsec and stop the nonsense.

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

cigitalgem, 4 months ago (edited 3 months ago) to random

Who pays the price when AI is wrong? #MLsec

https://bc.ctvnews.ca/air-canada-s-chatbot-gave-a-b-c-man-the-wrong-information-now-the-airline-has-to-pay-for-the-mistake-1.6769454

BIML blog entry https://berryvilleiml.com/2024/02/15/when-ml-goes-wrong-who-pays-the-price/

cigitalgem, 4 months ago to llm

Yesterday we released a new BIML study. Please read it and pass it on.

An Architectural Risk Analysis of Large Language Models (January 24, 2024)

#MLsec #LLM #genAI #ML #AI #security

https://berryvilleiml.com/results/BIML-LLM24.pdf

cigitalgem, 8 months ago to random

Recording a podcast about #MLsec and BIML's work with Google today.

cigitalgem, 6 months ago to ML

The #MATCH webinar will begin in 5 minutes:
Machine learning #ML
Artificial intelligence #AI
Threat modeling #threatmodeling
Compliance
How the heck these link together
#MLsec #swsec

cigitalgem, 7 months ago to random

Pre-poisoning to thwart ML. #MLsec

https://www.technologyreview.com/2023/10/23/1082189/data-poisoning-artists-fight-generative-ai/

cigitalgem, 5 months ago to random

How to regulate ML...regulate the LLM and generative AI foundation models. Poisoned data can't be simply filtered out of behavior later. #MLsec

https://arstechnica.com/tech-policy/2023/12/child-sex-abuse-images-found-in-dataset-training-image-generators-report-says/

cigitalgem, 8 months ago to random

I am doing a fireside chat with Dan Geer at In-Q-Tel tomorrow. We'll mostly be chatting about #MLsec and BIML's 78 risks. Some preliminary results from our mapping of the risks to LLMs.

cigitalgem, 5 months ago to llm

Gave a talk yesterday in Rio to a distinguished group of philosophers of mind from all over the world. #MLsec #LLM #AI #ML

Brain trust included a majority of Canadians

cigitalgem, 9 months ago to random

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, 5 months ago to random

What fun! #enshittification https://www.worstinshowces.com/

cigitalgem, 10 months ago to random

Can you code using predictive statistical patterns? Nope.

#MLsec

https://www.theregister.com/2023/08/07/chatgpt_stack_overflow_ai/

cigitalgem, 10 months ago to random

"starting to doubt" my ass. #MLsec

https://fortune.com/2023/08/01/can-ai-chatgpt-hallucinations-be-fixed-experts-doubt-altman-openai/

cigitalgem, 4 months ago (edited 4 months ago) to random

Lets do a TOP TEN LLM Risks list

#MLsec #swsec #ML #security

10: Encoding Integrity

https://berryvilleiml.com/results/BIML-LLM24.pdf