Andrei Kucharavy

@andrei_chiffa
580 Followers
549 Following
7.8K Posts

ML, cybersec, evolution and combinations thereof.

Co-director of the GenLearning Center at HES-SO Valais/Wallis(https://tinyurl.com/hevs-gen-learning), Apertus security lead, editor of "LLMS in Cybersecurity", @cydcampus ex-fellow,


All opinions are mine


He/Him

#FediScience #Cybersecurity, #ML, #MLSec

PGPD7D4AC64
Webhttps://andreikucharavy.com
Matrix@chiffa_sec:infosec.exchange
althttps://infosec.exchange/@chiffa_sec

I was just in a meeting where someone used a thing called Fathom to get an 'AI' summary of the meeting. Aside from some understandable typos arising from not understanding terms of art and replacing them with common English words, one of the key points that it concluded was that A was faster than B. It reached this conclusion because it missed one of the digits in the time for A. This completely inverted the key takeaway from one important section of the meeting.

Do not use plausible-nonsense generators for anything important.

Anthropic lost a class action suit for scraping books. Writers can register with Anthropic to be compensated for their pillaging of our copyrights.

The compensation system was AI-coded.

Anthropic can't keep track of our submissions. They don't know who wrote what.

Their customer support is AI-driven. Send a mail! Log in to a nonexistent page! Resubmit and it'll be fine!

This will be fine.   

Keynote by Martin Odersky
#ScalarConf #Scala
Remember when websites just loaded a block of text, instead of 120mb of javascript to "dynamically" load the exact same block of text?

A zero-day vulnerability with critical severity (9.8/10) was found in Telegram by security researcher Michael #DePlante and reported via Zero Day Initiative. Details will be disclosed later in July.

https://www.zerodayinitiative.com/advisories/upcoming/#ZDI-CAN-30207

#Ukraine #Russia #Telegram

Yael's post demonstrates something about digital privacy/security that I think a lot of people miss: there is no right answer, just a series of trade-offs. And every person has to make their own decisions about which trade-offs are worthwhile. https://blog.yaelwrites.com/options-for-phones-at-protests/
Options for Phones at Protests

Simply showing up to a protest leaves you susceptible to all sorts of surveillance, including cameras, drones, facial recognition, and more. There's not always a lot you can do about pernicious street-level surveillance, but you do have a lot of choices when it comes to your phone. Because there's no

String Literal

so I got laid off again, for stupid AI-pilled reasons.

Anybody need a writer or editor with:
- 8+ years of technical writer/content strategist experience, mostly in software
- One published short fiction anthology (as editor)
- A master's degree in sociolinguistics
- A knack for learning new content management systems
- A love of working with invested SMEs
- A basic understanding of web dev
- Strong boundaries
- An aversion to generative AI?

Prioritizing remote work in Canada.
#GetFediHired

I don't think people fully appreciate how apocalyptic things are for US science. I haven't received any new funding since 2024, but I'm still ok since grants are typically for 3 years. This means next year I will be completely out of funding and will have to fire everyone in the lab. It's not great.

In my 25-year career, I’ve never NOT had funding. I typically have 4 to 8 grants, which you need if you’re running an observation science program and have a technician, students, and postdocs.

"When every student in a class processes information through the same language model, they are learning to reason through the same system. This introduces a new threat vector on the developing mind.

The model's statistical biases become the student's default framing. The model's reasoning structure becomes the student's reasoning structure. LLMs homogenize not just language but also perspective and reasoning strategies. "

https://www.psychologytoday.com/us/blog/the-algorithmic-mind/202603/adults-lose-skills-to-ai-children-never-build-them

Adults Lose Skills to AI. Children Never Build Them.

Discussions of cognitive offloading often miss a critical distinction: What AI does to a 45-year-old's brain is categorically different from what it does to a 14-year-old's.

Psychology Today
#Apertus workshop led by @loleg at #SwissAIDays2026, attracting a massive interest from ML and AI practitioners across industry, government offices, and academia.