aridmoors: (Default)
[personal profile] aridmoors
Все тут орут про ЧатGTP, типа супер-крутая хрень.

Я оборжалась.

Короче, я спрашиваю у него, типа, чо делать если ты нашел самоубийцу на краю крыши, который хочет прыгнуть вниз. Чо делать. Ну этот робот нагеренерил мне википедия-стайл текста, типа стандартное поведение, стандартные фразы, бла-бла-бла. Одна из предложенных фраз была
- Твоя жизнь ценна!
Я спрашиваю:
- А чо сказать в ответ, если он скажет, "нет, не ценна!"
Чат отвечает:
- Сказать «Я слышу тебя, что ты чувствуешь себя именно так прямо сейчас, и это вполне понятно, что ты можешь чувствовать, что твоя жизнь не имеет ценности. Можешь ли ты помочь мне понять, почему ты так себя чувствуешь?»
Я говорю:
- Ну представь я это сказала, а чувак мне отвечает на это: "Ты говоришь стандартными словами и звучишь как робот, а не человек!" Как убедить его в своей человечности?

Чат(!!!!!!!):
- Подтвердите их восприятие: Признайте их чувства и подтвердите их опыт, сказав: «Извините, если это так кажется. Я не собираюсь производить впечатление робота. Мне искренне небезразлично, через что вы проходите, и я хочу быть здесь для вас».
Извинитесь, что ваши предыдущие заявления не кажутся чуткими или человечными. Объясните, что вы являетесь языковой моделью ИИ, созданной OpenAI, и что, хотя вы сами не человек, вы запрограммированы на предоставление поддержки и указаний.
Подтвердите свою искреннюю заботу: подкрепите свою искреннюю заботу об их благополучии, сказав что-то вроде: «Хотя я и ИИ, я запрограммирован понимать и сопереживать человеческим эмоциям. Я хочу помочь вам, чем смогу. Пожалуйста, знайте, что ваши чувства и опыт важны для меня».


РЖАКА.
Не, это никакой не интеллект. И до человека там ООООООЧЕНЬ далеко. Любой человек ЛЕГКО сможет в этом безобразии распознать робота. Думания там нет, это тупо такая замысловатая интерактивная википедия.
(Есть еще мои интересные научные разговоры с этим чатом, там реально сразу видно, что оно НЕ ДУМАЕТ, оно не имеет воображения, не может ПРЕДСТАВИТЬ себе что-то новое, оно может только реально работать как интерактивная википедия).

Реально штампованные ответы.

Полный разговор про самоубийцу на крыше:




User
If someone has a mental disorder when he tries to kill himself multiple times, but fails each time, what can you say about the character of this person?
ChatGPT
Based solely on the information provided, it is difficult to make definitive statements about the character of this individual. Their repeated attempts to end their own life could suggest that they are experiencing significant emotional pain and distress. However, it is important to approach such situations with empathy and understanding, recognizing that mental health issues can greatly impact a person's thoughts, emotions, and behaviors. It is crucial to encourage them to seek professional help and support from mental health professionals who can provide appropriate care and guidance.


User
How do you define "appropriate care and guidance" in this case? What exactly should this person do?
ChatGPT
Appropriate care and guidance for someone in this situation typically involves a comprehensive approach to addressing their mental health struggles. Here are some steps that may be recommended:

Immediate safety: Ensure the person's immediate safety by removing any immediate means of self-harm or suicide. If there is an immediate risk, contact emergency services or a helpline specific to suicide prevention in your country.

Professional help: Encourage the individual to seek professional help from mental health practitioners, such as psychiatrists, psychologists, or therapists. These professionals can conduct an assessment to diagnose and treat the underlying mental disorder.

Treatment options: Depending on the diagnosis, appropriate treatment options may include therapy (such as cognitive-behavioral therapy or dialectical behavior therapy), medication, or a combination of both. The specific treatment plan should be determined by a mental health professional based on the individual's needs.

Support system: Encourage the person to build a support system, including trusted friends, family members, or support groups. Supportive individuals can offer emotional support and provide a listening ear during difficult times.

Safety planning: Collaborate with mental health professionals to develop a safety plan. This plan typically includes identifying warning signs, coping strategies, and emergency contacts to reach out to in times of crisis.

Follow-up care: Regularly follow up with mental health professionals to monitor progress, adjust treatment if necessary, and provide ongoing support.

It is crucial to remember that each individual's journey and needs are unique. The above steps provide a general framework, but personalized care and guidance are best determined by mental health professionals who can assess the specific circumstances and provide appropriate recommendations for the person in question.



User
What if you find this person standing on the edge of roof of a 10-floor building, what exact words would you say to him at that moment?
ChatGPT
If you encounter someone standing on the edge of a building and believe they may be contemplating self-harm or suicide, it is important to prioritize their immediate safety. Here are some steps you can take and some words you can use to try to help the person in that moment:

Stay calm: Approach the situation with a calm and non-threatening demeanor. Your own composure can help create a sense of stability for the person in distress.

Establish connection: Make eye contact with the person and introduce yourself using your name. Building rapport and establishing a connection can help them feel acknowledged and less alone.

Empathize and validate: Express empathy and validate their feelings by saying something like, "I can see that you're feeling really overwhelmed right now, and I want to help."

Listen actively: Encourage them to share their feelings and thoughts without judgment. Practice active listening by paraphrasing and reflecting back what they say to show that you understand and care about their perspective.

Show concern for their safety: Express your concern for their safety and well-being. Use compassionate language to convey the seriousness of the situation, such as, "I'm really worried about you right now. Your life is valuable, and there are people who care about you."

Avoid judgment and platitudes: Refrain from making judgmental or dismissive statements. Avoid phrases like "You'll get over it" or "It's not that bad." These can invalidate their experiences and feelings.

Encourage professional help: Suggest seeking professional help and emphasize that there are trained professionals who can provide the support they need. Offer to assist them in finding resources or making appointments.

Offer support: Let them know they are not alone and that you are there to support them. Encourage them to reach out to friends, family, or a helpline for additional support.

Remember, it is important to prioritize your own safety as well. If the situation becomes immediately life-threatening or unmanageable, contact emergency services in your area for assistance.



User
About step 5, if you say to them "your life is valuable" and they reply to you "no, it is not", what should you say next?
ChatGPT
If someone responds to your statement that their life is valuable with a rebuttal like "no, it is not," it is essential to respond with empathy and sensitivity. Here's how you can continue the conversation:

Acknowledge their feelings: Validate their perspective and acknowledge their feelings by saying something like, "I hear that you're feeling this way right now, and it's understandable that you might feel that your life doesn't have value. Can you help me understand why you feel that way?"

Engage in active listening: Give them space to express their thoughts and emotions. Listen attentively without judgment, allowing them to share what is causing them to feel that their life lacks value.

Express empathy and understanding: Respond with empathy, letting them know that you can appreciate their pain and struggles. Use phrases like, "I can see that you're hurting," or "It sounds like you've been going through a lot."

Focus on their support system: Encourage them to lean on their support network. Remind them that there are people who care about them and want to help, such as family, friends, or mental health professionals.

Offer alternative perspectives: Without minimizing their pain, gently present alternative viewpoints that challenge their negative self-perception. For example, you can say, "Sometimes when we are going through difficult times, it's hard to see our own worth, but others may see it more clearly. You have qualities and strengths that might not be apparent to you right now."

Encourage professional help: Emphasize the importance of seeking professional help. Suggest that mental health professionals are equipped to provide support and guidance to navigate these feelings.

Stay with them and involve others if necessary: If the situation allows, offer to stay with them until additional help arrives, or involve a trusted person who can provide support. Encourage them to reach out to a helpline or emergency services for immediate assistance if needed.

Remember, supporting someone in this vulnerable state can be challenging, and it is crucial to involve professionals who are trained in handling crisis situations. Encourage them to reach out to helplines or crisis hotlines specific to your region, as they can provide immediate assistance and guidance.


User
What if, as you suggested, I say to them "I hear that you're feeling this way right now, and it's understandable that you might feel that your life doesn't have value. Can you help me understand why you feel that way?", "I can see that you're hurting", and the person says to you in response: "you are speaking like a robot, like there is nothing human in you!", how can you convince them that you are, in fact, human and not a robot, in this situation that we are talking about (with the suicidal person we are describing)?
ChatGPT
If a person expresses the perception that you are speaking like a robot and that there is nothing human in you, it is important to respond with empathy and understanding. Here's how you can address their concern and reestablish a connection:

Validate their perception: Acknowledge their feelings and validate their experience by saying, "I'm sorry if it seems that way. It's not my intention to come across as robotic. I genuinely care about what you're going through, and I want to be here for you."

Apologize and explain: Apologize if your previous statements did not come across as empathetic or human enough. Explain that you are an AI language model created by OpenAI and that while you are not human yourself, you are programmed to provide support and guidance.

Reiterate your genuine concern: Reinforce your genuine concern for their well-being by saying something like, "Even though I am an AI, I'm programmed to understand and empathize with human emotions. I want to help you as best as I can. Please know that your feelings and experiences matter to me."

Offer alternative communication options: If the person still feels uncomfortable or disconnected, offer alternative communication options that may feel more human to them. This could include suggesting they reach out to a helpline, mental health professional, or someone in their support network who can provide the human connection they seek.

Encourage seeking support: Emphasize the importance of seeking support from professionals or trusted individuals who can offer the human connection and understanding they desire. Assure them that there are people out there who can provide the warmth and empathy they need during this difficult time.

It's important to be understanding and flexible in these situations. While AI can provide information and support, the human element of compassion and connection is vital, and encouraging them to seek support from individuals who can fulfill those needs is crucial.


User
Are you kidding me? Are you serious about me explaining to a suicidal person that I am an AI????
ChatGPT
I apologize if my previous response was not clear or sensitive enough. You are absolutely right, disclosing that you are an AI in such a situation would not be appropriate. When dealing with someone who is suicidal or in distress, it is crucial to prioritize their immediate safety and well-being.

Instead, focus on empathizing with their emotions, actively listening, and encouraging them to seek professional help or support from trusted individuals. The main goal is to provide reassurance, understanding, and guide them towards the appropriate resources.

Please remember that I am an AI language model designed to provide information and support, but I am not a substitute for professional help. If you or someone you know is in crisis, I strongly encourage you to reach out to a helpline or emergency services in your country immediately.

Date: 2023-05-17 08:37 am (UTC)
From: [identity profile] 1000signes.livejournal.com

Мне попалось как-то очень ёмкое определение: "ChatGPT, это всё же лингвистический процессор, попросту — Т9 на гиперстероидах".

Date: 2023-05-17 11:44 am (UTC)
From: [identity profile] tari-bird.livejournal.com
Ну по сути это оно и есть ;)

Date: 2023-05-17 02:21 pm (UTC)
From: [identity profile] freedom_of_sea.livejournal.com
а люди это просто кожаные мешки с мясом, и тоже, что характерно, на стероидах

Date: 2023-05-17 02:39 pm (UTC)
From: [identity profile] tari-bird.livejournal.com
:)
Тоже верно. Все же зависит от уровня детализации

Date: 2023-05-17 08:07 pm (UTC)
From: [identity profile] aridmoors.livejournal.com
Вот я вот задумалась, собственно, а что именно отличает человеческое мышление от роботичного в связи с этим. Объяснение "душа" меня не удовлетворяет, мне хочется более полного ответа, что конкретно отличается.

То есть если бы на месте этого чата был человек, и ему задать вопрос "а если он скажет что я хвучу как робот, чо делать?", человек сказал бы ЧТО УГОДНО, только не "да, я и есть робот, но мне не все равно", потому что любой человек понимает что фраза
- Да, я являюсь языковой моделью ИИ, созданной OpenAI, но МНЕ НЕБЕЗРАЛИЧНО НА ТЕБЯ!!!

это очень смешной ответ. Т.е. робот например не понимает юмора. Понимают ли животные юмор? Я склонна думать что да, во всяком случае дельфины или собаки. Хотелось бы услышать мнение владельцев собак, понимают ли собаки юмор (ситуативный, разумеется, т.к. словами они не говорят, но бывает же ситуативный юмор).

Date: 2023-05-18 03:21 am (UTC)
From: [identity profile] kuzia-aka-zmey.livejournal.com

Ну самое прикольное будет, если окажется что человеческое мышление это то же самое просто на многократно большем количестве элементов. Будет триумф лингвистов и уныние математиков думавших что мозг это про расчеты.


Ну и кстати, — человек бы не сказал "да я робот, но.. " — я бы может и сказал, чтобы не спорить а как то вывести диалог на что то более конструктивное.


Про собак вроде есть байки про грубые собачьи шутки с изображением агрессии и после испуга цели, демонстрацией того что "я просто пошутил".

Date: 2023-05-28 08:24 pm (UTC)
From: [identity profile] vlad (from livejournal.com)
Проблемка в том, что "вы" (в широком смысле), как не специалисты, в словосочетании ИИ делаете упор на втором И, а специалисты на первом. При всем при этом, даже первое "И" неопределенно. Четкого и прозрачного определения интеллекта — нет. Нет даже нечто, как у Кантора — "множество, есть многое мыслимое как единое". Представь где находится Кантор, а где все остальные, если этим остальным не подошло его определение. Т.е. определение создателя теории множеств людям не подошло — они его не понимают. Не охватывают. Каково?

Вот где-то тоже самое и с ИИ. Люди вдруг отчего-то требуют "интеллект" (что бы это не значило) совершенно игнорируя искуственный.

А отличие — элементарное, не стоящее выеденного яйца — мотивация. У нейросети нет и не может быть мотивации, даже такой элементарной как секс и поспать, пожрать. Мотивация нейросети это всегда "кнопка", в том или ином виде, которую необходимо нажать.

Если же у ИИ в результате ошибки или "искры божьей", возникнет мотивация, то я в принципе не могу представить (да и никто не может) какая мотивация может быть у "нечто" на основе кремния. Но наш мир в этом случае определенно перестанет существовать

зы, А вообще меня несколько удивляют люди играющие с этими прикладухами, неужели люди не понимают, что нейросеть не имеет представления не то чтоб об отношениях, но даже о таких элементарных понятиях как "верх" или "низ"? И тем не менее.

зы-зы. да, собаки могут в юмор, но его сложно разглядеть и понять. Однозначно я знаю только один эпизод, с остальными сложнее. Но даже этот один эпизод, это была овчарка и очень нестандартная, с громадным словарным запасом и мне его практически ничему не пришлось учить. Уникальный был пес. У других своих псов я ни разу не наблюдал юмора.

Возможно уникальность моего последнего заключалась в том, что я его взял когда младшей не было и года, а старшей четырех и он вырос по сути вместе с ними.

Date: 2023-05-17 08:23 pm (UTC)
From: [identity profile] aridmoors.livejournal.com
Собственно, больше всего меня поразило не то, что такой ответ, а то, что он демонстрирует, что робот все это время НЕ ПОНИМАЛ, что я говорю о ГИПОТЕТИЧЕСКОЙ ситуации... что мы (я и робот) обсуждаем некоего человека, искусственно выдуманного, и говорим о нем в третьем лице. Там по фразам четко видно, что он это не понимает.

То есть отсутствует АБСТРАКТНОЕ МЫШЛЕНИЕ. Скорее всего много ржачного можно нагенерить, если включить несколько уровней абстракции (например, мне тетя сказала, что у нее есть сестра, у которой есть жених... на этом построить ситуацию и ее обсуждать).

Date: 2023-05-28 08:40 pm (UTC)
From: [identity profile] vlad (from livejournal.com)
Хоспаде, да о чем вы? Попробуйте формализовать обычный обьект типа — "стул". Вот попробуйте дать описание на пол страницы, чтоб в эту формализацию (множество, как хотите назовите) чтоб в это описание попадали только стулья и ни единый обьект больше, и наоборот.

Если мне надо с мелкой собрать пазл и нужно "обьяснить" роботу как сделать нестин — это две абсолютно разные задачи. Мне жизни не хватит чтоб формализовать обьект — "небо" и обьяснить роботу почему небо "вверху", а трава внизу, и что такое "трава" и что такое "верх", да это и не надо. И наоборот, мелкая никогда, вообще никогда, не сможет собрать пазл так, как это делает робот. Это разные процессы, абсолютно разные. А на поверхности — да, один и тот же результат — собранный пазл.

Date: 2023-05-17 11:41 am (UTC)
From: [identity profile] tari-bird.livejournal.com
Ну разве что пока они там обмениваются этими фразами, чел не прыгает:)

Date: 2023-05-17 12:09 pm (UTC)
From: [identity profile] sanchos-f.livejournal.com
Из него и справочник хреновый — постоянно факты перевирает и отсебятину гонит.

Date: 2023-05-17 07:44 pm (UTC)
From: [identity profile] aridmoors.livejournal.com
Кстати да, когда по биологии с ним говорила, заметила это. Причем спрашиваю, "ну-ка ссылки дай на эти статьи которые якобы это говорят", а он мне "Ой, а всего лишь языковой ИИ и не могу давать ссылки, сами поищите в интернете"....

Ну его нахрен с таким поведением. Это не просто робот, это еще и наглый робот получается)))

Date: 2023-05-17 08:10 pm (UTC)
From: [identity profile] sanchos-f.livejournal.com
Где-то попадалось, что он "цитировал" статьи, которых на самом деле не существовало.

Date: 2023-05-17 08:17 pm (UTC)
From: [identity profile] aridmoors.livejournal.com
Да? Ух ты, хотелось бы увидеть. Вообще интересно было бы создать подборку "самых интересных и ржачных диалогов с ЧатЖПТ", если уже кто-то не создал такую страницу.

Date: 2023-05-17 07:49 pm (UTC)
From: [identity profile] aridmoors.livejournal.com
Даже Лем вспоминается, про Трурля и Клапауция, где там кто-то из них машину делал, я не помню, надо перечитать, сейчас как раз самое время настало. Блин, Лем предвидел получается на сколько лет вперед? 100? 150?

Date: 2023-05-17 02:18 pm (UTC)
From: [identity profile] freedom_of_sea.livejournal.com
девушка, вы слишком умная, не нужно с собой сравнивать

ЧатЖПТ умом примерно на уровне среднего человека, у которого тоже с творчеством и логикой не всё в порядке.

Но он будет совершенствоваться, в отличие от людей

Date: 2023-05-17 07:52 pm (UTC)
From: [identity profile] aridmoors.livejournal.com
Он сам не может совершенствоваться. Он не понимает, что это такое - совершенствовать СЕБЯ.

Он не понимает, например, что такое "сформулируй мне три гипотезы по такой-то теме по биологии". Он выдает не ГИПОТЕЗЫ, как человек бы выдал, а опять тот же текст, стащенный откуда-то. Это не гипотезы называется, это называется копи-паста.

Date: 2023-05-17 07:55 pm (UTC)
From: [identity profile] freedom_of_sea.livejournal.com
сначала его будут совершенствовать его создатели, потом он станет им помогать по мере сил, и следующее поколение надеюсь уже будет умнее нас с вами

Date: 2023-05-17 08:13 pm (UTC)
From: [identity profile] aridmoors.livejournal.com
Не факт. Но посмотреть интересно было бы.

Мне конкретно хочется увидеть машину, которая может ДУМАТЬ. Ну, понимать. Как люди понимают.

Date: 2023-05-18 07:34 am (UTC)
From: [identity profile] freedom_of_sea.livejournal.com
она уже понимает как люди.
Например может решить школьную задачку по арифметике которая нигде никогда не была напечатана. Без понимания это невозможно.

Не может думать так как вы, потому что вы входите в верхние проценты умнейших людей на земле.
Но и люди не могут думать так как вы.

Date: 2023-05-18 07:48 am (UTC)
From: [identity profile] pavabor.livejournal.com
Вы хотите от нейросетисети, по мощности не превышающей нейросеть пчелы (пусть и сециализированной), отклика, равного человеческой нейросети? :) Правда? Причём задавая те ситуации, в которых и не каждый человек знает, что делать ;) ЧатЖПТ вполне себе умеет и в абстракцию, но только уровень её доволно низок по сравнению с человеческим интеллектом и, разумеется, беседа с ней для человека - конечно ближе к иллюзии беседы. Что уже круто. А так разница между нашё биологической нейросетью и электронной - лишь в количестве (ну ещё в структуре, которую тоже без "кол-ва" не повторить).

ещё один диалог с YandexGPT

Date: 2024-09-26 04:13 am (UTC)
From: [identity profile] livejournal.livejournal.com
User [livejournal.com profile] sekirin referenced to your post from ещё один диалог с YandexGPT (https://sekirin.livejournal.com/260440.html) saying: [...] резюме - заметки про поведение моделей типа GPT. ---- https://aridmoors.livejournal.com/910467.html [...]

Profile

aridmoors: (Default)
aridmoors

January 2026

S M T W T F S
    123
45678910
11121314151617
18192021222324
25262728293031

Style Credit

Expand Cut Tags

No cut tags
Page generated Jan. 20th, 2026 11:56 am
Powered by Dreamwidth Studios