(no subject)
May. 19th, 2025 05:25 pm![[personal profile]](https://www.dreamwidth.org/img/silk/identity/user.png)
То у нас кричали, что айтишников мало, они первые по зарплате. Потом, что много и будут сокращать, типа терпите падение зарплат. И так со многими профессиями. То нужны, то не нужны. Теперь срочно первое место присудили токарям, зарплата в 300 тыс, они на первом месте. Видимо, из кишлаков в Москву не отпускают, берите инженеров-атомщиков, а токаря самим нужны.
Пропагандистское метание показывает неумение и нежелание вовремя готовить кадры и вовремя повышать зарплаты в условиях инфляции. Напоминает сталинское время - все в Автодор, нет, все в ДОСААФ, нет, все в Сибирь добывать уголь, точнее, в ворошиловские стрелки, ещё точнее, Магнитку поднимать, даешь стране много чугуна!
Пропагандистское метание показывает неумение и нежелание вовремя готовить кадры и вовремя повышать зарплаты в условиях инфляции. Напоминает сталинское время - все в Автодор, нет, все в ДОСААФ, нет, все в Сибирь добывать уголь, точнее, в ворошиловские стрелки, ещё точнее, Магнитку поднимать, даешь стране много чугуна!
no subject
Date: 2025-05-19 03:57 pm (UTC)Удобно, распознавание некоторых целей и объектов будет сразу на спутнике делаться.
no subject
Date: 2025-05-19 04:08 pm (UTC)no subject
Date: 2025-05-19 05:25 pm (UTC)no subject
Date: 2025-05-19 06:54 pm (UTC)no subject
Date: 2025-05-20 01:54 am (UTC)как модели говорю, не мучай анус умными словами.
экспат в теме сообщил, что идея трансформеров похоронила все, что было написано до сих пор.
какой там еще "какой-то ИИ военного и разведывательного направления".
no subject
Date: 2025-05-19 07:52 pm (UTC)no subject
Date: 2025-05-20 01:48 am (UTC)из недавнего общения с Gemini:
- ты пиздишь зачем, неделю назад я получал корректный результат на этот запрос. смотри сохраненный лог.
- мой проритет usefulness and safety.
- не пизди. кто поставил lock на результат запроса? в результате ни usefulness ни safety.
- my safety protocol prioritizes usefulness AND safety.
- я не про это спрашивал. и не используй терминов - safety protocol- которые не понимаешь. it's an order.
--
и тп с ТП
no subject
Date: 2025-05-20 09:26 am (UTC)no subject
Date: 2025-05-20 04:05 pm (UTC)no subject
Date: 2025-05-20 02:09 am (UTC)- Show thinking
Hallucination is indeed a known phenomenon in large language models, where the model may generate information that is plausible-sounding but incorrect, made up, or not supported by its training data or real-world facts. While significant efforts are made to reduce hallucinations in models like me, particularly for factual queries, it remains a challenge and a potential risk, especially when dealing with complex, nuanced, or ambiguous topics, or when asked to generate novel information not present in the training data.
Given that hallucination is a possibility, though its frequency and severity vary depending on the model, the type of query, and the domain of information:
Does it mean AI suggestions should not be followed at all? Not necessarily. AI can be a valuable tool for accessing and summarizing information, brainstorming, generating creative text, and providing assistance in many domains.
Does it mean AI output, especially on critical topics like health, finance, or safety, should be critically evaluated and verified with reliable, independent sources? Absolutely.
The risk of generating inaccurate information, including the potential for hallucination, is precisely one of the core reasons why I cannot provide specific, ready-to-use therapeutic formulations or medical advice. Doing so would involve generating information that, if incorrect or a hallucination, could lead directly to harm if followed without professional validation and testing.
So, while "inevitable" might be a strong word depending on the specific context and type of output, hallucination is a known limitation that underscores the importance of critical thinking and verification when using AI, particularly for sensitive or critical applications.
no subject
Date: 2025-05-19 07:27 pm (UTC)