Live and recent posts from across the Fediverse tagged #testingai, aggregated by home.social.
Một thử nghiệm đối chiếu Gemini 2.5 Flash & các mô hình mã nguồn mở (OSS) trong tạo giao diện UI thông qua prompt chi tiết 62.9k token. Gemnini hoàn thành mượt mà, các mô hình OSS như Qwen, GPT-OSS, Llama-70B… phần lớn gặp lỗi: tắc nghẽn trong reasoning (dù đặt "low"), gọi tool sai, hoặc bỏ qua quy trình. Chỉ Kwaipilot-kat-coder thực hiện được nhưng chậm 3x & lỗi gọi tool. Cơ bản là kiến trúc khác biệt hay lỗi triển khai?
#AI #MôHìnhĐạiDiện #Gemini #MãNguồnMở #LậpTrình #TestingAI #TốiTânAI #G
Financial Times: OpenAI slashes AI model safety testing time. “OpenAI has slashed the time and resources it spends on testing the safety of its powerful artificial intelligence models, raising concerns that its technology is being rushed out without sufficient safeguards. Staff and third-party groups have recently been given just days to conduct ‘evaluations’, the term given to tests for […]
https://rbfirehose.com/2025/04/11/financial-times-openai-slashes-ai-model-safety-testing-time/
AI struggles with less common data: Inconsistent results for Valletta Bastions (actual mean height: 25m) highlight issues with insufficient training data. We also touch on AI poisoning.
https://www.alanbonnici.com/2025/03/ai-got-it-wrong-missing-information-or.html
#AI #DataBias #Valletta #TTMO #ArtificialIntelligence #hallucination #Mistakes #TestingAI #InsufficientData #DataPoisoning
How does AI handle insufficient information? 🤔 We tested an AI with questions about the Eiffel Tower, Big Ben, and the bastions of Valletta. The AI gave inconsistent answers when training data is limited or unclear. We also touch on AI poisoning, where AI models can be misled by fake data
▶️ https://buff.ly/yRDWPTf
#AI #InsufficientData #DataPoisoning #EiffelTower #BigBen #Valletta #TestingAI #Accuracy #TTMO
Also so wird das nichts mit uns, #ChatGPT!