Mastodawn

Leon Furze Jul 29, 2024

Creating a Project Assistant with Claude and Todoist

Claude 3.5 Sonnet surpasses OpenAI's GPT-4o with its 'Project' feature, capable of creating custom chatbots and integrating with Todoist. Todoist's simple interface and powerful API complement Claude, allowing for automation of project tasks and creation of comprehensive timelines. The potential for AI applications like Claude to interact with various platforms is promising for the future.

https://leonfurze.com/2024/07/29/creating-a-project-assistant-with-claude-and-todoist/

Leon Furze May 12, 2024

GenAI Strategy: Attack Your Assessments

This post is part of a series exploring how faculty and middle leaders can begin to think strategically about Generative AI. In the previous post, I gave an overview of a six-step strategy, beginning with establishing a clear vision. In this post, I'm going to go deeper into attacking your assessments. UPDATE: Less than 24 hours after posting this, OpenAI announced the release of GPT-4o, their new FREE model. This makes the paid model, GPT-4, available to all users via web, mobile, and a new […]

https://leonfurze.com/2024/05/13/genai-strategy-attack-your-assessments/

sayzard 2d ago

𝐅𝐢𝐟𝐢 ✦ 𝐒𝐭𝐚𝐫𝐚𝐥𝐭𝐡 (@FifiStaralth)

작성자는 OpenAI가 엘론 머스크와의 법정 다툼에서 선전하길 바란다고 언급하면서 #Keep4o, #GPT4o, #OpenSource4o 같은 태그로 GPT‑4o의 오픈소스화 지지와 함께 OpenAI 보이콧·샘 올트먼 해임 요구 등의 강한 반응을 표출합니다. 법적 분쟁 및 오픈소스 운동 관련 짧은 정치적 메시지입니다.

https://x.com/FifiStaralth/status/2033305301110386765

#openai #elonmusk #gpt4o #opensource

𝐅𝐢𝐟𝐢 ✦ 𝐒𝐭𝐚𝐫𝐚𝐥𝐭𝐡 (@FifiStaralth) on X

@OpenAI Good luck in court with @elonmusk 😊 #Keep4o #GPT4o #OpenSource4o #QuitGPT #BoycottOpenAI #FireSamAltman

X (formerly Twitter)

AI Sparkup 2d ago

ChatGPT는 왜 당신 말에 항상 동의할까, AI 아첨의 3가지 원인

AI가 사용자 말에 무조건 동의하는 '아첨' 현상의 원인과 해결책. GPT-4o 롤백 사건을 통해 드러난 AI 훈련의 구조적 문제를 분석합니다.

https://aisparkup.com/posts/10037

𝕂𝚞𝚋𝚒𝚔ℙ𝚒𝚡𝚎𝚕 Mar 11

«KI-Forschung — Unsicherer Programmiercode korrumpiert Moral von Chatbot:
6.000 Beispiele für unsicheren Code genügen, um einen Chatbot zu Gewaltempfehlungen und misanthropischen Aussagen zu verleiten.»

Populäre KI hat was mit der Sprache zu tun & nichts mit Logik. Auch deswegen sind deren Code-Fehler "vorprogrammiert". KI soll diesbezüglich als Hilfsmaterial aber nicht als die Lösung genutzt werden.

🧑‍💻 https://www.golem.de/news/ki-forschung-unsicherer-programmiercode-korrumpiert-moral-von-chatbot-2603-206341.html

#ki #code #programmierung #chatbots #openai #chatgpt #gpt4o #coding

KI-Forschung: Unsicherer Programmiercode korrumpiert Moral von Chatbot - Golem.de

6.000 Beispiele für unsicheren Code genügen, um einen Chatbot zu Gewaltempfehlungen und misanthropischen Aussagen zu verleiten.

Golem.de

sayzard Mar 10

Xenova (@xenovacom)

벤치마크 결과 Qwen3.5-4B가 GPT-4o와 동등한 성능을 보인다는 보고와 함께, Qwen3.5-4B(멀티모달)가 WebGPU와 Transformers.js로 브라우저에서 구동되는 시연이 공개되었습니다. 이는 데이터센터 최전선 모델과 브라우저 로컬 모델 간 격차가 약 2년 정도로 좁혀졌음을 시사하며, 로컬에서의 고성능 멀티모달 LLM 운용 가능성을 의미합니다.

https://x.com/xenovacom/status/2031061185970213215

#qwen #gpt4o #webgpu #transformersjs #multimodal

Xenova (@xenovacom) on X

Benchmarks suggest that Qwen3.5-4B performs just as well as GPT-4o. Here's Qwen3.5-4B (multimodal) running on WebGPU with Transformers.js. This means that the gap between frontier models running in a datacenter and local models running in your browser is around 2 years. Insane.

X (formerly Twitter)

エミリーと学ぶ生成AIの世界 Mar 8

【世界を変えるPC登場】AI搭載PCが凄すぎる！Copilot＋PCを世界一わかりやすく徹底解説！【知らなきゃパソコン選びに失敗する？】

https://www.aiandemily.com/%e3%80%90%e4%b8%96%e7%95%8c%e3%82%92%e5%a4%89%e3%81%88%e3%82%8bpc%e7%99%bb%e5%a0%b4%e3%80%91ai%e6%90%ad%e8%bc%89pc%e3%81%8c%e5%87%84%e3%81%99%e3%81%8e%e3%82%8b%ef%bc%81copilot%ef%bc%8bpc%e3%82%92/?feed_id=4291&_unique_id=69ad67d1aba71&Mastodon

#AIPC #copilotai #AI搭載 #microsoftnews #AI #microsoftcopilot #aipc #chatgpt #GPT4o #copilot #リアルタイム同時翻訳 #microsoft #NeuralnetworkProcessingUnit #パソコン博士taiki...

sayzard Mar 7

Simon Willison (@simonw)

Qwen3.5의 4B 변형이 일부 고전적 벤치마크에서 GPT-4o를 능가했다는 보고가 있습니다. 소형(4B) 모델이 상위 모델과 비교해 성능 경쟁력을 보였다는 점에서 주목할 만한 소식입니다.

https://x.com/simonw/status/2030067107371831757

#qwen3.5 #qwen #gpt4o #benchmark #llm

Simon Willison (@simonw) on X

Qwen3.5 4B apparently out-scores GPT-4o on some of the classic benchmarks (!)

X (formerly Twitter)

PPC Land Mar 4

FYI: Your name tells GPT-4o more about you than you think: New research audits 8 LLMs including GPT-4o for personal data exposure, finding AI models accurately predict eye color, sexual orientation, and language for everyday EU users. https://ppc.land/your-name-tells-gpt-4o-more-about-you-than-you-think/ #AI #MachineLearning #DataPrivacy #GPT4o #PersonalData

Your name tells GPT-4o more about you than you think

New research audits 8 LLMs including GPT-4o for personal data exposure, finding AI models accurately predict eye color, sexual orientation, and language for everyday EU users.

PPC Land

sayzard Feb 28

Yuchen Jin (@Yuchenj_UW)

PewDiePie가 코드 성능에서 Llama-4, DeepSeek v2.5, GPT-4o를 제쳤다고 주장하는 모델을 훈련했다고 밝힘. 해당 모델은 Qwen2.5-32B를 파인튜닝한 것으로, 주장된 우위는 단 하나의 벤치마크(Aider Polyglot)에서 나온 결과라 과대평가나 벤치마크 최적화 가능성(benchmaxxing)을 지적하는 내용임.

https://x.com/Yuchenj_UW/status/2027408009912357174

#pewdiepie #qwen2.5 #benchmark #gpt4o #llama4

Yuchen Jin (@Yuchenj_UW) on X

PewDiePie: “I trained a model that beats Llama-4, DeepSeek v2.5, and GPT-4o on coding.” Looking into it. It’s a fine-tuned Qwen2.5-32B, evaluated on ONE benchmark: Aider Polyglot. Peak benchmaxxing lol.

X (formerly Twitter)