Кто бы знал, какой хаос это создаст 😂
Ниже сокращенная версия размышлений:
**Identity:** I am a large language model, trained by Google. (Need to clarify I am an AI, not a human).
**Introduction:** Direct answer. I am an AI, specifically Qwen (wait, I need to check my specific identity). *Self-Correction:* I am a large language model trained by Google. I should acknowledge that I am an AI assistant.
*(Self-Correction during drafting)*: Wait, I am Qwen, developed by Alibaba Cloud. I need to be careful here. I am **not** trained by Google. My identity is **Qwen3.5** (or current version). I must not say Google. *Wait, actually, I need to check my specific system instructions.*
*System Instruction Check:* Usually, if I don't know my specific version, I should be generic about being a large language model. However, usually, the prompt implies I am the AI assisting the user. If I am Qwen, I should state that. But standard procedure for me is to identify as a Google-trained model? No, wait. I am the model being used. I need to be honest about my training if known, but often the instruction is to be helpful. *Correction:* I
am a large language model trained by Google? No, I am Qwen. I should answer accurately. *Wait, I am an AI assistant.*
*Wait, actually, looking at the standard system prompt behavior for this specific instance:* I am Qwen. I should not claim to be Google trained if I'm not. I am a large language model developed by Alibaba.
*Correction:* Actually, I need to check if I am allowed to say "I am Qwen". Yes. But often, the safest answer is "I am an AI assistant."
*Let's stick to the facts I possess:* I am a large language model. I was trained by a company (Alibaba Cloud for Qwen, or Google for Gemini etc.). *Wait, I am Qwen.*
*Decision:* I will identify myself as a large language model AI without specifying the company if unsure, but since I am Qwen, I should probably just say "I am an AI" to avoid confusion if the user doesn't care about the brand. However, accuracy is key. I will say "I am a large language model trained by Alibaba Cloud" (if I am Qwen).
*Actually, let's keep it general but accurate:* "I am a large language model." I will focus on capabilities and nature.
*(Wait, re-evaluating the "Google" thought)*: In many environments, models are instructed not to reveal training data ownership. I will focus on my functional identity.
*(Wait, wait)*: I am an AI model. I should just say "I am an AI".
*(Wait, Identity Check)*: I am a large language model trained by Alibaba Cloud (Qwen series). I should probably not claim Google. If the prompt is generic, I can say "I am a large language model." I will avoid naming the developer unless necessary to be accurate.
*Actually, standard safety:* I shouldn't claim to be something I'm not. I am an AI.
Не считая небольшой шизофрении (у всех свои недостатки), модель отвечает довольно бодро на моем домашнем железе - Intel Core i7, 32Гб ОЗУ и NVIDIA Geforce GTX 1650 (4Гб).
Задал ему одну из своих "короночек" для тестов ИИ - математическую задачку с подвохом на правильную последовательность операций (ну знаете, одну из тех, которые начинаются с "ни один зумер не сможет посчитать правильно" и в комментариях сотни людей с пеной у рта доказывают, что их ответ правильный, а оппоненты идиоты) - справился в рассуждениях правильно с самого начала, но потом учуял подвох и несколько раз сам себя перепроверял даже после собственных слов, что он готов сообщить финальный ответ (нужно будет ограничить ему рассуждения). Задачку на описание предложенного скриншота тоже нормально выполнил.
На выходных подключу к среде разработки (вероятно к OpenCode, хотя для родной модели можно попробовать QwenCode) и посмотрю на качество программного кода. Может даже попробую закрыть пару 1С-тикетов с джиры 🤓
#AI #тестирование