This notion that early-career coders can be replaced by AI is wrong. Nobody is thinking about maintaining software or the architectural disaster to come when inscrutable code that mostly works is all over the codebase. And it's real..."early-career coders have been hit especially hard because much of what they do can now be done by AI."
I recommend hiring junior humans over AI. Still. Just tool them up.
https://www.wsj.com/lifestyle/careers/tech-jobs-hiring-artifical-intelligence-35cd66b0
Reviewing this absolute garbage work that has a veneer of science. What a joke. If this is the kind of #MLsec out of Europe that is supposed to save us, we're screwed.
Academic journals in security are utterly useless. #infosec #security
https://www.sciencedirect.com/science/article/pii/S0167404824002931
@baldur have you read our work? You might appreciate it. You can use it to shut those guys up.
One person's data pollution is another person's data gold. As long as we have next to zero insight into the immense training data sets used by LLMs this will happen again and again. Data protection fail groundhog day. #ML #AI #MLsec #security
https://www.darkreading.com/cyberattacks-data-breaches/deepseek-breach-opens-floodgates-dark-web
Benchmarks as popularity contests don't work. There are lots of other reasons that benchmarks have become almost worthless in #AI. In particular #ML benchmark as badnessometer comes to mind. #MLsec
https://techcrunch.com/2025/04/22/crowdsourced-ai-benchmarks-have-serious-flaws-some-experts-say/
This coverage of COT in ML is misleadingly anthropomorphic. Have we really lost track of how these things work? Just because we call something "chain of thought" that doesn't make it ACTUAL chain of thought. Anthropic has always done this. #MLsec
And this is a usually excellent reporter falling prey to the nomenclature.
Welp, I caught the BIML Bibliography up to 2024. LOL. A labor of love, that's for sure. Only three more months of papers to enter...
We keep track of the #MLsec field by reading the science so you don't have to.
See our "top 5 papers" list to get started.
GenAI is so expensive that some (foundation model building) companies will not survive. This is a major cause of technical lock in...scientific exploration of the space is incredibly expensive.
Who can do A/B testing when it is that expensive?
https://www.theregister.com/2025/03/31/llm_providers_extinction/
As usual, @dangoodin has written an excellent security explainer article. This one is about prompt injection...but not the usual trial and error whack-a-mole prompt manipulation by pizza guy...instead, automated manipulation by search in gradient space.
This technique is new enough that we're discussing the original paper only today at BIML. It makes the whole boring front door malicious input thing much more interesting.
@futurebird you might enjoy reading this work I did with my group on #MLsec
The reg wall is mostly pretend...
If deciding what is true in an ML training set is a political decision, then generating the "ground truth" with a synthetic generation algorithm is a great way to control model behavior.
Build a HOW machine to make a WHAT pile to train a WHAT machine.
https://www.wired.com/story/nvidia-gretel-acquisition-synthetic-training-data/
Turns out that you can automate being wrong. It's easy! #MLsec
This has nothing to do with security and is not how to secure #AI. #MLsec
https://arstechnica.com/ai/2025/02/anthropic-dares-you-to-jailbreak-its-new-ai-model/
We are so deep into data feudalism that even the plagerizers are whining. #MLsec
https://www.lawfaremedia.org/article/why-the-data-ocean-is-being-sectioned-off
@Daojoan data,feudalism is here to stay #MLsec https://www.lawfaremedia.org/article/why-the-data-ocean-is-being-sectioned-off
Unleashing AI. So much for #MLsec. It's all up to us now.
@chockenberry recursive pollution is real #MLsec
https://berryvilleiml.com/2024/01/29/two-interesting-reads-on-llm-security/
