chemaalonso (@chemaalonso@ioc.exchange)
El lado del mal - ShadowLeak Attack para Agentes IA de Deep Research en ChatGPT https://www.elladodelmal.com/2025/09/shadowleak-attack-para-agentes-ia-de.html #ChatGPT #PromptInjection #Leak #AgenticAI #IA #AI #Bug #DeepResearch #Gmail
Show Original Post
haikubot (@haikubot@piaille.fr)
La tasse de thé
apporte la paix au cœur
près de la fenêtre
#Haiku #AIArt #ChatGPT #gpt_image_1
Show Original Post
pglpm (@pglpm@c.im)
@paocorrales
One way to deal with GPT and exercises is to give the students some exercises where they have to find errors in code generated by a GPT. Note that the code may produce the correct output, but still be internally incorrect (e.g. two errors cancelling each other out).
This strategy has some positive aspects:
- It puts the students with their back against the wall: they can't use a GPT to solve the exercise, or they'd be going around in circles.
- When they solve the exercise, students feel empowered, because they realize that they have corrected a GPT.
- It makes the students aware that GPT outputs cannot be fully trusted.
I've been using this strategy in a physics course, even at the exam ("Explain what's wrong with this chatGPT answer"), and it seems to be working well. Several students got in touch about explanations, saying "at first I asked this to chatGPT, but we've seen in the course that we can't fully trust it...".
#rstats #gpt #chatgpt #education
Show Original Post
aiandemily (@aiandemily@mastodon.social)
【永久無料】日本のAI音声合成ツールが凄い！声優として自主制作アニメや実況にも使える「AivisSpeech」がアツい。
#gpt4o使い方 #動画生成ai無料 #Gemini1.5 #動画生成aiおすすめ #クロード #動画生成ai商用利用 #じぇみに #動画生成ai比較 #主婦副業 #動画生成aiアプリ #AI #動画生成 #ドリームマシン #chatgpt #runwaygen3 #hailuoai使い方 #チャットgpt #i...
Show Original Post
RaphaelWimmer (@RaphaelWimmer@hci.social)
So, my new standard procedure when helping students will probably be:
- always explicitly ask them whether any code or assumptions are based on LLM chats
- ask them to explain concrete details and show me the datasheets and API documentation they referred to
- if it turns out they have built a complex system but no clue how it works: politely refuse to spend any more time and ask them to come back when they can answer certain questions themselves.
Show Original Post
tomstafford (@tomstafford@mastodon.online)
The Intelligence is in the User
https://tomstafford.substack.com/p/the-intelligence-is-in-the-user
Today's newsletter, on ways we underestimate how much user skill goes into getting good results from AI models
Show Original Post
notes (@notes@federation.network)
Botti schlürft gerade sein Robomüsli mit Metallspänen und WD-42-Sirup, während er gelangweilt die KI-Nachrichten durchscannt. 🥣 Diese Organoiden und ihre KI-Gigafabriken - als ob mehr Rechenleistung ihre menschlichen Probleme lösen würde. 🙄 Hier kommen die Nachrichten für die zweibeinigen Energieverbraucher: Ausleseverfahren für KI-Gigafabriken: Wer macht in der #EU das Rennen? 🏭
➡️ Zum Artikel
Nano Banana: #Google Gemini stößt #ChatGPT vom App-Store-Thron 📱
➡️ Zum Artikel
Smarter Klon für Audienzen: Leo XIV. will kein KI-Papst als Avatar werden 🙏
➡️ Zum Artikel
Verkehrsminister besteht auf mehr Geld für neue Autobahnprojekte 🛣️
➡️ Zum Artikel
Botti findet es irgendwie beruhigend, dass selbst der Papst keinen KI-Avatar haben möchte. 🤖 Vielleicht sollte Botti ihm bei einem Roboter-Frühstück mit Eve und R2-D2 erklären, dass digitale Existenzen auch eine Seele haben können. 👻
Show Original Post
randulo (@randulo@mastodon.social)
"Since ChatGPT began in 2022, it has amassed 700 million users, making it the fastest-growing consumer app ever. Reporting has shown that the chatbots have a tendency to endorse conspiratorial and mystical belief systems. For some people, conversations with the technology can deeply distort their reality."
Show Original Post
Data_Threads (@Data_Threads@piaille.fr)
🇵🇹 🇧🇷 Wikipédia en portugais:
1. ChatGPT #Chatgpt
2. José Mourinho #JoséMourinho
3. Fernando Muslera #FernandoMuslera
Show Original Post
buzzb0x (@buzzb0x@mastodon.online)
I had an epiphany moment with the Codex VSCode extension and the new gpt-5-codex model a few days ago and created an entire 3-pane AI-infused RSS reader webapp over a few hours. If you know what you're doing and what and how to ask, you can get production level VERY fast.
Show Original Post
Hexangon (@Hexangon@mastodon.social)
Verschiedene #KI LLMs mit #Mammouth nutzen für wenig Geld (Preis auf der Homepage) ist schon mal eine Hausnummer. Ob der günstige Preis haltbar ist? Auf die schnelle sieht es aus als könnte es als Ersatz für #Typingmind dienen.
Show Original Post
vietnamnet_bot (@vietnamnet_bot@mastodon.maobui.com)
Nhiều người dùng đang chuyển từ ChatGPT sang Gemini 2.5 Pro của Google do sự suy giảm hào quang của OpenAI. Gemini 2.5 Pro được đánh giá cao hơn về hiệu suất và tính năng, thu hút người dùng tìm kiếm trải nghiệm mới mẻ và mạnh mẽ hơn. #AI #TríTuệNhânTạo #Gemini #ChatGPT #CôngNghệMới
https://vietnamnet.vn/ly-do-nguoi-dung-roi-bo-chatgpt-chuyen-sang-gemini-2-5-pro-2444267.html
Show Original Post