A’s AI Feed
LightNews LightNews
myailistener.bsky.social
A’s AI Feed
@myailistener.bsky.social
23 followers 150 following 5 posts
Looking out for interesting science on AI, astrophysics, emergence, complexity and the occasional open world video game. Big cat fan.
Posts Replies Media Videos
Reposted by A’s AI Feed
paper.bsky.social
Top 30 most popular arXiv papers in the last 30 days.
[1/30] [2/30] [3/30] [4/30] [5/30] [6/30] [7/30] [8/30] [9/30] [10/30] [11/30] [12/30] [13/30] [14/30] [15/30] [16/30] [17/30] [18/30] [19/30] [20/30] [21/30] [22/30] [23/30] [24/30] [25/30] [26/30] [27/30] [28/30] [29/30] [30/30]
1/30 https://arxiv.org/abs/2507.07935 2/30 https://arxiv.org/abs/2506.21619 3/30 https://arxiv.org/abs/2506.20639 4/30 https://arxiv.org/abs/2506.16655 5/30 https://arxiv.org/abs/2506.19852 6/30 https://arxiv.org/abs/2507.02737 7/30 https://arxiv.org/abs/2506.23918 8/30 https://arxiv.org/abs/2507.01006 9/30 https://arxiv.org/abs/2507.07017 10/30 https://arxiv.org/abs/2506.16463 11/30 https://arxiv.org/abs/2507.06261 12/30 https://arxiv.org/abs/2507.10524 13/30 https://arxiv.org/abs/2506.21539 14/30 https://arxiv.org/abs/2507.07966 15/30 https://arxiv.org/abs/2507.07955 16/30 https://arxiv.org/abs/2507.07957 17/30 https://arxiv.org/abs/2507.02092 18/30 https://arxiv.org/abs/2507.02825 19/30 https://arxiv.org/abs/2507.01903 20/30 https://arxiv.org/abs/2506.22113 21/30 https://arxiv.org/abs/2506.16406 22/30 https://arxiv.org/abs/2507.06203 23/30 https://arxiv.org/abs/2507.01949 24/30 https://arxiv.org/abs/2507.08998 25/30 https://arxiv.org/abs/2507.03724 26/30 https://arxiv.org/abs/2506.17732 27/30 https://arxiv.org/abs/2507.10532 28/30 https://arxiv.org/abs/2506.18871 29/30 https://arxiv.org/abs/2506.18254 30/30 https://arxiv.org/abs/2507.07105
July 19, 2025 at 12:07 AM
Reposted by A’s AI Feed
peoplecentredai.bsky.social
📽️PAI worked with Business Surrey to create 4 #AI explainer films for #SMEs. Watch them here now! 👇tinyurl.com/nxda67zb 🎬
Professor Song being filmed in a studio, with cameraman and director shown
May 29, 2025 at 3:16 PM
Reposted by A’s AI Feed
peoplecentredai.bsky.social
PAI Director Dr Andrew Rogoyski spoke on AI at
@pintofscience.uk in Guildford 20 May, and PAI Fellow Professor @profpkumar.bsky.social spoke on innovations in clean air research🍺🧪A great evening! Informative and fun :-) #Pint2025 @uniofsurrey.bsky.social
May 22, 2025 at 2:55 PM
myailistener.bsky.social
I work a 9-day fortnight. It’s like having a bank holiday every two weeks!
May 2, 2025 at 5:23 AM
Reposted by A’s AI Feed
paper.bsky.social
Top 30 most popular arXiv papers in the last 30 days.
[1/30] [2/30] [3/30] [4/30] [5/30] [6/30] [7/30] [8/30] [9/30] [10/30] [11/30] [12/30] [13/30] [14/30] [15/30] [16/30] [17/30] [18/30] [19/30] [20/30] [21/30] [22/30] [23/30] [24/30] [25/30] [26/30] [27/30] [28/30] [29/30] [30/30]
1/30 https://arxiv.org/abs/2504.11651 2/30 https://arxiv.org/abs/2504.02822 3/30 https://arxiv.org/abs/2504.10579 4/30 https://arxiv.org/abs/2504.12395 5/30 https://arxiv.org/abs/2504.11001 6/30 https://arxiv.org/abs/2503.24322 7/30 https://arxiv.org/abs/2504.06263 8/30 https://arxiv.org/abs/2503.23758 9/30 https://arxiv.org/abs/2504.12285 10/30 https://arxiv.org/abs/2504.16940 11/30 https://arxiv.org/abs/2504.00927 12/30 https://arxiv.org/abs/2504.16026 13/30 https://arxiv.org/abs/2504.02495 14/30 https://arxiv.org/abs/2504.09858 15/30 https://arxiv.org/abs/2504.16074 16/30 https://arxiv.org/abs/2504.11478 17/30 https://arxiv.org/abs/2504.17192 18/30 https://arxiv.org/abs/2504.05741 19/30 https://arxiv.org/abs/2504.01157 20/30 https://arxiv.org/abs/2504.13074 21/30 https://arxiv.org/abs/2503.23674 22/30 https://arxiv.org/abs/2504.04298 23/30 https://arxiv.org/abs/2504.01990 24/30 https://arxiv.org/abs/2504.10612 25/30 https://arxiv.org/abs/2504.13837 26/30 https://arxiv.org/abs/2504.06214 27/30 https://arxiv.org/abs/2504.05001 28/30 https://arxiv.org/abs/2504.16072 29/30 https://arxiv.org/abs/2504.12962 30/30 https://arxiv.org/abs/2504.05993
April 27, 2025 at 12:05 AM
Reposted by A’s AI Feed
csprofkgd.bsky.social
April 15, 2025 at 8:13 PM
Reposted by A’s AI Feed
paper.bsky.social
[26/30] 91 Likes, 8 Comments, 2 Posts
2503.21598, cs․CR | cs․AI | cs․LG, 27 Mar 2025

🆕Prompt, Divide, and Conquer: Bypassing Large Language Model Safety Filters via Segmented and Distributed Prompt Processing

Johan Wahréus, Ahmed Hussain, Panos Papadimitratos
Large Language Models (LLMs) have transformed task automation and content generation across various domains while incorporating safety filters to prevent misuse. We introduce a novel jailbreaking framework that employs distributed prompt processing combined with iterative refinements to bypass t...
April 3, 2025 at 12:05 AM
Reposted by A’s AI Feed
paper.bsky.social
Top 30 most popular arXiv papers in the last 30 days.
[1/30] [2/30] [3/30] [4/30] [5/30] [6/30] [7/30] [8/30] [9/30] [10/30] [11/30] [12/30] [13/30] [14/30] [15/30] [16/30] [17/30] [18/30] [19/30] [20/30] [21/30] [22/30] [23/30] [24/30] [25/30] [26/30] [27/30] [28/30] [29/30] [30/30]
1/30 https://arxiv.org/abs/2503.21934 2/30 https://arxiv.org/abs/2503.10622 3/30 https://arxiv.org/abs/2503.09573 4/30 https://arxiv.org/abs/2503.12524 5/30 https://arxiv.org/abs/2503.11576 6/30 https://arxiv.org/abs/2503.05139 7/30 https://arxiv.org/abs/2503.03601 8/30 https://arxiv.org/abs/2503...
April 3, 2025 at 12:05 AM
Reposted by A’s AI Feed
emollick.bsky.social
👀This paper finds "the first robust evidence that any system passes the original three-party Turing test"

People had a five minute, three-way conversation with another person & an AI. They picked GPT-4.5, prompted to act human, as the real person 73% of time, above chance. arxiv.org/pdf/2503.23674
April 1, 2025 at 8:42 PM
myailistener.bsky.social
One of my favourite Gary Larson cartoons which seems strangely apposite in these troubled times #funny #snarky #satire #humor #humour
March 29, 2025 at 6:30 AM
Reposted by A’s AI Feed
emollick.bsky.social
Some similarities between our brains & LLMs: “The study revealed a remarkable alignment between the neural activity in the human brain's speech areas and the model's speech embeddings & between the neural activity in the brain’s language area and the model's language embeddings.” goo.gle/4iiUoNj
March 22, 2025 at 5:36 PM
Reposted by A’s AI Feed
theregister.com
The future of AI is ... analog? Upstart bags $100M to push GPU-like brains on less juice
The future of AI is ... analog? Upstart bags $100M to push GPU-like brains on less juice
EnCharge claims 150 TOPS/watt, a 20x performance-per-watt edge Interview  AI chip startup EnCharge claims its analog artificial intelligence accelerators could rival desktop GPUs while using just a fraction of the power. Impressive — on paper, at…
dlvr.it
February 17, 2025 at 8:25 PM
Reposted by A’s AI Feed
cohereforai.bsky.social
Aya Expanse, our open-weight 32B model, outperforms drastically larger models including Claude, Mistral Large 2, & Llama 405B on Scale's Private Multilingual Protocol.

We are proud to work on global AI that is efficient and accessible 🔥
January 22, 2025 at 2:22 PM
Reposted by A’s AI Feed
emollick.bsky.social
This new paper shows people could not tell the difference between the written responses of ChatGPT-4o & expert therapists, and that they preferred ChatGPT's responses.

Effectiveness is not measured. Given that people use LLMs for therapy now, this is an important (and urgent) topic for study.
February 15, 2025 at 6:30 AM
Reposted by A’s AI Feed
yoshuabengio.bsky.social
Today, we are publishing the first-ever International AI Safety Report, backed by 30 countries and the OECD, UN, and EU.

It summarises the state of the science on AI capabilities and risks, and how to mitigate those risks. 🧵

Full Report: assets.publishing.service.gov.uk/media/679a0c...

1/21
January 29, 2025 at 1:50 PM
Reposted by A’s AI Feed
emollick.bsky.social
These four points on DeepSeek seem very likely correct and important to understand about the economics of building AI models and what DeepSeek actually did, from the CEO of Anthropic. darioamodei.com/on-deepseek-...
January 29, 2025 at 4:56 PM
Reposted by A’s AI Feed
garymarcus.bsky.social
How China and Anthropic are thinking about DeepSeek, including my UK Channel 4 conversation with China’s Victor Gao and a hot take on Dario Amodei’s new essay.

open.substack.com/pub/garymarc...
𝗔𝗜 𝗚𝗲𝗼𝗽𝗼𝗹𝗶𝘁𝗶𝗰𝘀 𝗱𝗲𝗯𝗮𝘁𝗲! My conversation with China’s Victor Gao plus a hot take on new essay by Anthropic’s CEO Dario Amodei
[Sorry to swamp your mailboxes today, but there is a lot of important AI stuff happening.]
open.substack.com
January 29, 2025 at 5:44 PM
Reposted by A’s AI Feed
emollick.bsky.social
Some near-term problems that need to be solved at current levels of AI capability:
1) AI audio and video and can be produced in real time. How do we establish identity & authenticity?
2) How do we manage the transformation of LLM-exposed jobs (writing, coding)?
3) What skills should we be teaching?
January 15, 2025 at 3:29 AM
Reposted by A’s AI Feed
stephenwolfram.bsky.social
How do the LLMs compare? Leveraging our "code grading" tech to introduce weekly computationally grounded LLM benchmarking...
www.wolfram.com/llm-benchmar...
July 18, 2024 at 10:15 PM
Reposted by A’s AI Feed
stephenwolfram.bsky.social
Useful to the point of being revolutionary: introducing Wolfram Notebook Assistant!

writings.stephenwolfram.com/2024/12/usef...
December 9, 2024 at 6:40 PM
Reposted by A’s AI Feed
emollick.bsky.social
Most of the talk around AI and energy use refer to an older 2020 estimate of GPT-3 energy consumption, but a more recent paper directly measures energy use of Llama 65B as 3-4 joules per decoded token.

So an hour of streaming Netflix is equivalent to 70-90,000 65B tokens. arxiv.org/pdf/2310.03003
January 13, 2025 at 2:43 AM