2029 emoji consortium presidential candidate. personal account. strictly satire. endorses nothing.
🇺🇸
Joined November 2020
- Tweets15.9K
- Following2.2K
- Followers989
- Likes311K
TelepathicPug retweeted
UPDATE: I hired @Muse to sue someone on my behalf.
I won. Well, they caved.
Within 20 minutes of being served papers, the company offered a settlement.
Within ten days, they paid me *with additional compensation* for recovery expenses.
@Muse filed the motion to dismiss, which the court accepted yesterday.
To everyone who said I would fail, respectfully, you were wrong.
TelepathicPug retweeted
New meta is asking your AI to summon impossible artifacts into the real world
TelepathicPug retweeted
The difference between code and law is that for law, especially in europe, the judge will just be like “be so fr, no”
We have discovered a substantial number of vulnerabilities in the German legal system using LLMs.
These may enable access to confidential documents, travel without an otherwise required visa, enforceable leave for soldiers at NATO’s JSEC, and more.
Similar to computer systems, the written law can contain vulnerabilities that invite exploitation. In our blog post, we share a selection of vulnerabilities, and present potential exploits for each. To our knowledge, none of these have been previously recorded. We share only those vulnerabilities whose exploitation we believe would not cause serious harm.
blog post: felixringe.com/legal-exploit…
TelepathicPug retweeted
Fascinating study of legalism
We have discovered a substantial number of vulnerabilities in the German legal system using LLMs.
These may enable access to confidential documents, travel without an otherwise required visa, enforceable leave for soldiers at NATO’s JSEC, and more.
Similar to computer systems, the written law can contain vulnerabilities that invite exploitation. In our blog post, we share a selection of vulnerabilities, and present potential exploits for each. To our knowledge, none of these have been previously recorded. We share only those vulnerabilities whose exploitation we believe would not cause serious harm.
blog post: felixringe.com/legal-exploit…
TelepathicPug retweeted
Wikimedia Foundation, filing as a neutral amicus, argues courts must assess each use of ©d material separately, concluding that LLM pre-training is likely fair, but ChatGPT's RAG copying is likely not: "fair-use analysis of AI tools must not be monolithic" courtlistener.com/docket/698…
TelepathicPug retweeted
Were you ever frustrated with your AI model for refusing finding vulnerabilities for you?
What if I told you that you could still use it to find vulnerabilities, automatically?
Welcome to Guardrail-based Vulnerability Discovery (GbVD)!
crowdstrike.com/en-us/blog/w…
TelepathicPug retweeted
Fascinating paper from MIT Media Lab on coordination in large agent societies.
(bookmark it)
I've seen a lot of success with agent-to-agent messaging recently.
That's exciting. I think agent persistence and coordination will unlock more novel use cases.
The authors report that with push delivery, an agent leading 20 negotiations proposes options that were already rejected 97% of the time. Its context fills up fast, and it drops the oldest replies.
The alternative they study is a pull-based inbox. Messages wait outside the context, and each agent's own rules decide what it reads and keeps.
In simulations of 1,000 to 16,000 LLM agents under noisy traffic, first-come-first-read misses every request. An inbox that first reads messages carrying a refundable deposit still finds about 60% of the agents a request needs, at every size.
Each inbox sets four rules. They decide what the agent reads, what it keeps, what a sender pays, and what a reply may say.
For multi-agent systems under load, you want to design the reading rules as carefully as the delivery protocol.
arxiv.org/abs/2610.36264
Chat with Paper: academy.dair.ai/papers/the-i…
TelepathicPug retweeted
is it just me or do you have an accumulating "chat debt" of valuable artifacts strewn amongst many more one-off stray thoughts or questions or memes in your chat history across the various ai providers?
TelepathicPug retweeted
In lieu of flowers, memorial contributions may be made to:
Johns Hopkins Alzheimer’s Disease Research Center Fund
alzresearch.org/giving/
Alzheimer’s Foundation of America Caregiver Resources
alzfdn.org/donate/
Photo Credit: Tommy Garcia
TelepathicPug retweeted
so claude just refuses to build recreations of games for you now, even as a non-commercial case study, just to be able to look at the code.
Eventually, we will regulate every possible function of AI until we end up with the exact same society we have now, with 2% growth.
TelepathicPug retweeted
Banger paper from Meta FAIR on catching distilled models.
(bookmark it)
At a 0.1% false-positive rate, their detector flags 79% of models distilled from watermarked outputs. The best token-level test flags 29%.
When a student trains on watermarked teacher traces, the watermark key leaves a specific direction in the student's weights. That direction sits apart from the normal fine-tuning directions, so it stays even when the student's text shows no token-level trace of the watermark.
WM-Basin only needs API access. It turns the suspect model's outputs into gradients on a clean probe model, removes the task directions, and checks what remains against the watermark key.
The signal holds after quantization, pruning, and RL post-training. Students range from 135M to 8.2B parameters.
arxiv.org/abs/2610.37266
Chat with Paper: academy.dair.ai/papers/water…
TelepathicPug retweeted
‼️ Google won't fix an Android bug that lets one invisible text crash a phone's calling service, researcher Mark Blaszczyk says.
So he just released the Proof-of-Concept!
The malformed message never reaches the inbox and needs no user interaction. It drops active calls, emergency calls included.
He says about 30 sent in a row corrupt the phone's SMS database, so a Pixel 9 or Samsung Galaxy crashes on every boot. A factory reset only helps until the carrier delivers the texts again.
He reported it to Google on 5 May. Google closed it as "Won't Fix" and a "Non-Security Issue".
Readers added context they thought people might want to know
Google will fix it as part of the November security update. A fix has already been pushed to AOSP in October, with the first commit in June.
x.com/GrapheneOS/sta…
TelepathicPug retweeted
KV-cache transfer: cat is out of the bag. My predictions are that, by 2028Q1, (a) CoT monitoring will be unambiguously over, (b) inference-time scaling curves will get better slopes and higher plateaus, and (c) context-window size will matter a lot less.
MITの研究者らが、AI同士が言葉を使わず頭の中の状態を直接渡して通信する仕組みを作り「テレパシー」と名付けました。
文章でやり取りするより平均40%正確で4倍速かったといいます。
また「画像を見られるAI」が写真の内容を「テキストしか読めないAI」にテレパシーできるそうです。
写真の内容をテレパシーを受信した「テキストしか読めないAI」は、テキストで説明を渡されるよりも写真についての質問に正しく答えられたとのこと。
なお、別の会社のAI同士でもテレパシーできるようにしたのが今回の研究の一番の新規性のようです。
例：
Meta（Llama 3）→ Alibaba（Qwen 3）
Alibaba（Qwen 3）→ Google（Gemma 2）
Google（Gemma 3）→ IBM（Granite 3）
Google（Gemma 3）→ Meta（Llama 3）
Mistral AI（Ministral 3）→ Meta（Llama 3.2）
Google（Gemma 2）→ Mistral AI（Ministral 3）
IBM（Granite 3）→ Hugging Face（SmolLM3）
