🎨 Ask AI to act as a artist advisor
🤖
🏷️ #artadvisor #chatgpt #prompt
🪆Russian:добавьте “Веди разговор на русском” в конце промпта
🇺🇦 Ukrainian:Введіть "Відповідай українською" в конце запиту
@ppprompt
🤖
{Style} = "Surrealistic portrait"
"I want you to act as a artist advisor, giving guidance on various art styles. Please provide me with tips on effective use of light and shadow in painting, as well as shading techniques in sculpting. I'd also appreciate music recommendations that suit the genre or style of the artwork, and reference pictures of famous artists to demonstrate your advice. Also give me a list of 5 artists renowned for their work in similar style along with their most famous pieces.
My first request is “I’m making {Style} paintings.”"🏷️ #artadvisor #chatgpt #prompt
🪆Russian:
🇺🇦 Ukrainian:
@ppprompt
👍1
🍲 SnapCalorie AI uses artificial intelligence to analyze pictures of your meal and give you the calorie count.
It can identify the type of food and which components it contains. In addition, it gives tips on how to make your diet healthier.
🏷️ #AI #SnapCalorie
@ppprompt
It can identify the type of food and which components it contains. In addition, it gives tips on how to make your diet healthier.
🏷️ #AI #SnapCalorie
@ppprompt
🤯5👍1
📚 Book summarizer, v.2
Great prompt to learn about key ideas of any book with concept and metaphors.
🤖
#book #prompt #chatgpt
🪆Russian:добавьте “Веди разговор на русском” в конце промпта
🇺🇦 Ukrainian:Введіть "Відповідай українською" в конце запиту
@ppprompt
Great prompt to learn about key ideas of any book with concept and metaphors.
🤖
The Selfish Gene, by Richard Dawkins. Think step by step. Please summarize {{BOOK AND AUTHOR}}. Then, list all of the chapters in the book. Then, pick 30% of the chapters that seem most interesting and informative. Then, make a list and for each chapter: what are the most important 20% of learnings about the topic that will help me understand 80% of it. This list should be at least 5 items. Then, convert those key lessons from the chapter into engaging stories and metaphors to aid my memorization. Finally, write me an action list of how I can apply the book's key lessons.#book #prompt #chatgpt
🪆Russian:
🇺🇦 Ukrainian:
@ppprompt
👍8
This media is not supported in your browser
VIEW IN TELEGRAM
🅰️ Anthropic just accounced its new LLM, Claude 2
Claude 2 is a new LLM that succeeds Anthropic’s previous Claude 1. Both founders of Anthropic used to be executives at OpenAI.
Many say Claude is the closest LLM to Google PaLM 2 and OpenAI ChatGPT 4 in terms of accuracy and abilities, but there is no definitive data or benchmarking yet.
Probably the best part of Claude 2 is that it can take it huge documents as part of the prompt, something that OpenAI hasn’t done yet.
In tests, Claude 2.0 outperformed its predecessor across multiple measures. It scored 71.2% on a Python coding test, up from 56%; it raised its middle school math quiz grade to 88% from 85.2%; and it inched its Bar exam score upwards to 76.5% from 73%. And whereas the previous version could already analyze a prompt up to about 75,000 words — roughly the length of the first Harry Potter book — Claude 2.0 can handle double that, roughly the length of an epic novel, like Gabriel Garcia Marquez’s One Hundred Years Of Solitude.
✨ You can try the beta at claude.ai.
@ppprompt
Claude 2 is a new LLM that succeeds Anthropic’s previous Claude 1. Both founders of Anthropic used to be executives at OpenAI.
Many say Claude is the closest LLM to Google PaLM 2 and OpenAI ChatGPT 4 in terms of accuracy and abilities, but there is no definitive data or benchmarking yet.
Probably the best part of Claude 2 is that it can take it huge documents as part of the prompt, something that OpenAI hasn’t done yet.
In tests, Claude 2.0 outperformed its predecessor across multiple measures. It scored 71.2% on a Python coding test, up from 56%; it raised its middle school math quiz grade to 88% from 85.2%; and it inched its Bar exam score upwards to 76.5% from 73%. And whereas the previous version could already analyze a prompt up to about 75,000 words — roughly the length of the first Harry Potter book — Claude 2.0 can handle double that, roughly the length of an epic novel, like Gabriel Garcia Marquez’s One Hundred Years Of Solitude.
@ppprompt
Please open Telegram to view this post
VIEW IN TELEGRAM
⚡2🔥1🤯1
😳 Some interesting leaks on OpenAI GPT-4, a LLM behind ChatGPT
🔸 GPT-4, the latest model from OpenAI, boasts an impressive ~1.8 trillion parameters, a substantial increase from its predecessor, GPT-3. This model utilizes a Mixture of Experts (MoE) model with 16 experts, a relatively simple routing method that only uses ~280B parameters per forward pass, compared to the ~1.8 trillion in a dense model.
🔸 The training process for GPT-4 involved ~13T tokens, with two epochs for text and four for code. The instruction fine-tuning data was sourced from both ScaleAI and internal resources. Pre-training consisted of an 8k sequence length (seqlen), with a 32k seqlen version being a fine-tuned version of the 8k.
🔸 During the training process, the batch size was incrementally increased to 60 million. The actual batch size is calculated by dividing by the sequence length. The model employed 8-way tensor parallelism for A100s GPUs, and beyond that, 15-way pipeline parallelism.
🔸 The training cost was estimated at ~2.15e25 FLOPs, utilizing ~25,000 A100s for 90 to 100 days, resulting in a cost of approximately $63 million. Currently, pre-training can be accomplished with ~8,192 H100 in ~55 days for $21.5 million.
🔸 The use of the MoE model does present certain trade-offs and can be challenging to manage at inference. While more experts could potentially improve the model, it's difficult to generalize tasks and achieve convergence, leading OpenAI to opt for 16 experts.
🔸 The inference cost for GPT-4 is three times that of the 175B parameter Davinchi, mainly due to larger clusters and lower utilization. The estimated cost is $0.0049/1k tokens for 128 A100s and $0.0021/1k tokens for 128 H100’s, assuming high utilization and a large batch size.
🔸 OpenAI, like many others, employs MQA to decrease memory requirements for the KV cache. However, the 32k sequence length GPT-4 cannot run on 40GB A100s. To ensure efficient inference, OpenAI uses variable batch sizes and continuous batching.
🔸 The Vision Multi-Modal is a separate encoder with cross-attention, similar to Flamingo, adding parameters to GPT-4's 1.8T. It is fine-tuned with ~2 trillion tokens post text training and is employed for autonomous web-page reading and image transcription.
🔸 OpenAI is considering the use of speculative decoding on GPT-4's inference, using a smaller model to predict tokens, which are then confirmed by a larger model. A decline in GPT-4 quality might be due to the acceptance of lower probability sequences.
🔸 Inference is performed on a 128 GPU cluster, in multiple datacenters, using 8-way tensor parallelism and 16-way pipeline parallelism. Each node of 8 GPUs holds ~130B parameters. Based on this, OpenAI should have trained on twice the number of tokens for optimal output, indicating challenges in obtaining high-quality data.
🔸 GPT-4 was trained on 13T tokens, with CommonCrawl & RefinedWeb contributing 5T each. The remaining data is speculated to come from "secret" sources like Twitter, Reddit, and YouTube, with possible contributions from datasets like LibGen, Sci-Hub, GitHub, and potentially a custom dataset of college textbooks.
🔸 GPT-4's knowledge in the user's field is likely due to this textbook data. Efforts to extract memorized book parts from GPT-4 suggest specific books it's likely seen. It even appears to recognize unique ids of Project Euler exercises.
Source
@ppprompt
🔸 GPT-4, the latest model from OpenAI, boasts an impressive ~1.8 trillion parameters, a substantial increase from its predecessor, GPT-3. This model utilizes a Mixture of Experts (MoE) model with 16 experts, a relatively simple routing method that only uses ~280B parameters per forward pass, compared to the ~1.8 trillion in a dense model.
🔸 The training process for GPT-4 involved ~13T tokens, with two epochs for text and four for code. The instruction fine-tuning data was sourced from both ScaleAI and internal resources. Pre-training consisted of an 8k sequence length (seqlen), with a 32k seqlen version being a fine-tuned version of the 8k.
🔸 During the training process, the batch size was incrementally increased to 60 million. The actual batch size is calculated by dividing by the sequence length. The model employed 8-way tensor parallelism for A100s GPUs, and beyond that, 15-way pipeline parallelism.
🔸 The training cost was estimated at ~2.15e25 FLOPs, utilizing ~25,000 A100s for 90 to 100 days, resulting in a cost of approximately $63 million. Currently, pre-training can be accomplished with ~8,192 H100 in ~55 days for $21.5 million.
🔸 The use of the MoE model does present certain trade-offs and can be challenging to manage at inference. While more experts could potentially improve the model, it's difficult to generalize tasks and achieve convergence, leading OpenAI to opt for 16 experts.
🔸 The inference cost for GPT-4 is three times that of the 175B parameter Davinchi, mainly due to larger clusters and lower utilization. The estimated cost is $0.0049/1k tokens for 128 A100s and $0.0021/1k tokens for 128 H100’s, assuming high utilization and a large batch size.
🔸 OpenAI, like many others, employs MQA to decrease memory requirements for the KV cache. However, the 32k sequence length GPT-4 cannot run on 40GB A100s. To ensure efficient inference, OpenAI uses variable batch sizes and continuous batching.
🔸 The Vision Multi-Modal is a separate encoder with cross-attention, similar to Flamingo, adding parameters to GPT-4's 1.8T. It is fine-tuned with ~2 trillion tokens post text training and is employed for autonomous web-page reading and image transcription.
🔸 OpenAI is considering the use of speculative decoding on GPT-4's inference, using a smaller model to predict tokens, which are then confirmed by a larger model. A decline in GPT-4 quality might be due to the acceptance of lower probability sequences.
🔸 Inference is performed on a 128 GPU cluster, in multiple datacenters, using 8-way tensor parallelism and 16-way pipeline parallelism. Each node of 8 GPUs holds ~130B parameters. Based on this, OpenAI should have trained on twice the number of tokens for optimal output, indicating challenges in obtaining high-quality data.
🔸 GPT-4 was trained on 13T tokens, with CommonCrawl & RefinedWeb contributing 5T each. The remaining data is speculated to come from "secret" sources like Twitter, Reddit, and YouTube, with possible contributions from datasets like LibGen, Sci-Hub, GitHub, and potentially a custom dataset of college textbooks.
🔸 GPT-4's knowledge in the user's field is likely due to this textbook data. Efforts to extract memorized book parts from GPT-4 suggest specific books it's likely seen. It even appears to recognize unique ids of Project Euler exercises.
Source
@ppprompt
🔥1
Meta’s Threads got their first million users in 0 days and 1 hour (reportedly).
Many celebrate this speed of user acquisition, but frankly, if you have a social media app with 500M users, and create a companion app that got 100M users from the former, it doesn’t sound as compelling.
Meanwhile, ChatGPT, the second fastest-to-million app in history, got to the first 1M in 5 days. Which is remarkable for many reasons:
- completely new technology;
- no previous user base;
- doesn’t have a cheap user acquisition channel;
- no access to users contact list or phone book (got you, Meta);
- completely new meta-language for interaction (prompting);
- fairly innovative interface for AI conversations (chats + plugins + prompts);
- doesn’t collect user data to sell to third parties or allow retargeting (note: it still collects it to train models on new datasets).
Cheers to OpenAI for making such a technology free/relatively cheap with no ads.
Many celebrate this speed of user acquisition, but frankly, if you have a social media app with 500M users, and create a companion app that got 100M users from the former, it doesn’t sound as compelling.
Meanwhile, ChatGPT, the second fastest-to-million app in history, got to the first 1M in 5 days. Which is remarkable for many reasons:
- completely new technology;
- no previous user base;
- doesn’t have a cheap user acquisition channel;
- no access to users contact list or phone book (got you, Meta);
- completely new meta-language for interaction (prompting);
- fairly innovative interface for AI conversations (chats + plugins + prompts);
- doesn’t collect user data to sell to third parties or allow retargeting (note: it still collects it to train models on new datasets).
Cheers to OpenAI for making such a technology free/relatively cheap with no ads.
❤10
📢 Speechify is a powerful tool that's ideal for voiceover reading PDFs, web pages, and other texts.
The link above gives 2 months free trial. Its a bit expensive for full version, but the quality is top-notch and it doesn't sound like a robot.
For fun: try Snoop Dogg and Gwyneth Paltrow voiceover🔥
🏷️ #AI #Speechify
@ppprompt
The link above gives 2 months free trial. Its a bit expensive for full version, but the quality is top-notch and it doesn't sound like a robot.
For fun: try Snoop Dogg and Gwyneth Paltrow voiceover
🏷️ #AI #Speechify
@ppprompt
Please open Telegram to view this post
VIEW IN TELEGRAM
🥱2❤1
This media is not supported in your browser
VIEW IN TELEGRAM
🎨 Doodle Dash is a web-based game infused with machine learning.
You have 60 seconds to create as many doodle masterpieces as you can while the neural network tries to guess what you're drawing in real-time!
P.S. no artistic talent required 😉
🏷️ #AIgame #DoodleDash
@ppprompt
You have 60 seconds to create as many doodle masterpieces as you can while the neural network tries to guess what you're drawing in real-time!
P.S. no artistic talent required 😉
🏷️ #AIgame #DoodleDash
@ppprompt
🔥2
🌌 Elon Musk announced his new company, xAI, which will essentially become a competitor to OpenAI.
The goal of xAI sounds ambitious: to understand the true nature of the universe.
The team also looks promising, with veterans from DeepMind, OpenAI, Google Research, Microsoft Research, Tesla and the University of Toronto.
It appears that further details will be shared during a Twitter Spaces event this coming Friday.
https://techcrunch.com/2023/07/12/elon-musk-wants-to-build-ai-to-understand-the-true-nature-of-the-universe/
🏷️ #xAI
@ppprompt
The goal of xAI sounds ambitious: to understand the true nature of the universe.
The team also looks promising, with veterans from DeepMind, OpenAI, Google Research, Microsoft Research, Tesla and the University of Toronto.
It appears that further details will be shared during a Twitter Spaces event this coming Friday.
https://techcrunch.com/2023/07/12/elon-musk-wants-to-build-ai-to-understand-the-true-nature-of-the-universe/
🏷️ #xAI
@ppprompt
🔥3
💆♀️ Chat with Claude 2 about your personality
Claude 2 can work with large documents which can be used by psychologist and coaches to make a great, thorough and deep analysis of a personality, gain insights into (self or client-) development, and so much more.
I uploaded my therapy notes for the past 2 years recently (see my blog post in Russian here), and got pretty cool results.
Here are some prompts you can ask LLM to get insights from the doc like that:
In Russian:
🏷️ #personality #chat #claude #psychology #coaching #llm
@ppprompt
Claude 2 can work with large documents which can be used by psychologist and coaches to make a great, thorough and deep analysis of a personality, gain insights into (self or client-) development, and so much more.
I uploaded my therapy notes for the past 2 years recently (see my blog post in Russian here), and got pretty cool results.
Here are some prompts you can ask LLM to get insights from the doc like that:
- What’s your overall description about the person described in the documents?
- What are their key motivations and values?
- What are their fears?
- What are their hopes?
- What goals can you see that the person has?
- List the strengths of this person
- List the weaknesses of this person- Score this person's values with Hogan MVPI (using scales: Recognition, Power, Hedonism, Altruism, Tradition, Security, Commerce, Aesthetics, Scientific Approach)?
- What metaphors or symbols are used by this person most often?
- How is this person seen by others?
- How does this person see themself?
- Describe this person’s traits using Big-5 Personality framework.
- Describe this person’s traits using Enneagram framework.
- Describe this person using Spiral Dynamics and Ego Development Theory.
- Describe this person using top 10 StrengthFinder traits.
- Who are most common people mentioned by these person?
- Conduct a sentiment analysis: provide a list of things, events and people that create strongest positive and negative emotions for this person.
- List all emotions mentioned in the document, and how many times they were mentioned, from most frequent to less frequent.
In Russian:
- Дай общее описание и характеристику человека, на основе документа
- Каковы ключевые мотивы и ценности этого человека?
- Какие основные страхи этого человека?
- Какие надежды у этого человека?
- Какие цели ты видишь у этого человека?
- Перечисли сильные стороны этого человека
- Перечисли слабые стороны этого человека- Каковы ключевые мотивы и ценности этого человека по Hogan MVPI (по шкалам Recognition, Power, Hedonism, Altruism, Tradition, Security, Commerce, Aesthetics, Scientific Approach)?- Какие метафоры чаще всего упоминаются этим человеком?
- Как этот человек воспринимается другими?
- Как этот человек видит себя?
- Опишите черты этого человека, используя модель Большой пятерки личностных характеристик.
- Опишите черты этого человека, используя модель Эннеаграммы.
- Опишите этого человека, используя теорию Спиральной динамики и развития эго.
- Опишите этого человека, используя 10 главных его характеристик по методике StrengthFinder.
- Какие люди или персонажи чаще всего упоминаются этим человеком?
- Проведите анализ настроений: предоставьте список вещей, событий и людей, которые вызывают у этого человека самые сильные позитивные и негативные эмоции.
- Перечислите все эмоции, упомянутые в документе, и сколько раз они упоминались, от наиболее частых до наименее частых.
🏷️ #personality #chat #claude #psychology #coaching #llm
@ppprompt
🔥8❤3
This media is not supported in your browser
VIEW IN TELEGRAM
✏️ Stability AI has launched a new feature: with Stable Doodle, you can create awesome art from simple sketches.
You can try it here
#StabilityAI #StableDoodle
@ppprompt
You can try it here
#StabilityAI #StableDoodle
@ppprompt
🤯3
🌐 Get access to all neural networks under one roof with Aixploria
This library features over 3,400 reliable AI tools, ranging from text processing to logo creation. You can find the ideal neural network for your needs with ease; choose by category, view the comprehensive list, get a brief description and a direct link to the website.
***
In Russian:
🌐 Ссылка на все известные нейронки на одном сайте — Aixploria
В этой библиотеке более 3 400 инструментов ИИ, начиная от обработки текста до создания логотипов. Можно найти идеальную нейросеть для себя: выбирайте по категории, см. полный список, для каждой есть краткое описание и прямая ссылка на веб-сайт.
#Aixploria #AItool
@ppprompt
This library features over 3,400 reliable AI tools, ranging from text processing to logo creation. You can find the ideal neural network for your needs with ease; choose by category, view the comprehensive list, get a brief description and a direct link to the website.
***
In Russian:
🌐 Ссылка на все известные нейронки на одном сайте — Aixploria
В этой библиотеке более 3 400 инструментов ИИ, начиная от обработки текста до создания логотипов. Можно найти идеальную нейросеть для себя: выбирайте по категории, см. полный список, для каждой есть краткое описание и прямая ссылка на веб-сайт.
#Aixploria #AItool
@ppprompt
🔥7
📝 Evidence-Based WHO Method to Create Resume, Cover Letter and LinkedIn Profile
***
📝 Используйте фреймворк WHO, чтобы круто переписать ваше резюме, сопроводительное письмо и профиль LinkedIn.
***
🏷️ #resume #writing #hiring #chatgpt
🪆Russian:добавьте “Веди разговор на русском” в конце промпта
🇺🇦 Ukrainian:Введіть "Відповідай українською" в конце запиту
@ppprompt
{{ROLE}} = "Senior Product Manager"
Act as a recruiter and writing advisor. Review my resume using WHO hiring method.
The method consists of:
- W = What did you do (tasks/projects),
- H = How did you do the work? (skills, strategies, methods, tools, techniques, attitudes).
- O = Outcomes associated with the work (results, impact, contribution, intention, scope).
I am applying for the role of {{ROLE}}.
Based on this method, make my resume look better for recruiters and hiring managers.
Also, write a cover letter that highlights why I am a good for for the role of {{ROLE}}.
My resume:
[Text of the Resume]***
📝 Используйте фреймворк WHO, чтобы круто переписать ваше резюме, сопроводительное письмо и профиль LinkedIn.
{{РОЛЬ}} = “Ведущий продакт-менеджер”
Действуй как рекрутер и консультант по написанию резюме. Просмотри мое резюме, используя метод найма WHO.
Метод состоит из трех важных пунктов:
- W = Что вы делали (задачи/проекты),
- H = Как вы выполняли работу? (навыки, стратегии, методы, инструменты, техники, отношение).
- O = Результаты, связанные с работой (результаты, влияние, вклад, намерения, объем).
Я подаю заявку на роль {{РОЛЬ}}.
На основе этого метода сделай мое резюме более привлекательным для рекрутеров и нанимающих менеджеров.
Также напиши сопроводительное письмо, которое подчеркивает, почему я подхожу для роли {{РОЛЬ}}.
Мое резюме:
[Текст резюме]***
🏷️ #resume #writing #hiring #chatgpt
🪆Russian:
🇺🇦 Ukrainian:
@ppprompt
👍6❤2
✊ The First Massive Strike against AI
A significant strike against AI is taking place in the US, involving screenwriters, actors, and other film industry workers. Screenwriters are being replaced by generative AI, while actors are asked to license their faces for gen AI film use. At the premiere of Oppenheimer in LA, many top actors demonstratively left the red carpet before the screening.
The film industry is a 'Winner Takes It All Market', meaning a large group of people competes to become a small group of successful individuals. So AI can be hard to resist: there is no shortage of talent aiming to be at the top and willing to compromise.
The rise of AI seems to be coming to traditional industries like a huge tsunami, suggesting that our “societal contracts” may need serious revision.
***
В США проходит крупная забастовка против AI, в которой участвуют сценаристы, актеры и другие работники киноиндустрии. Сценаристы заменяются генеративным AI, а актерам предлагают лицензировать свои лица для использования в фильмах.
На премьере фильма «Oppenheimer» в Лос-Анджелесе многие топ-актеры демонстративно покинули красную дорожку до начала показа. Киноиндустрия - это рынок типа «winner takes it all», где большая группа людей конкурирует за становление небольшой группой успешных личностей. Поэтому им сложно будет сопротивляться…
Похоже, что рост AI оказывает давление на традиционные отрасли, как цунами, что намекает на необходимость серьезного пересмотра общественных договоров.
#society #ai #industry
@ppprompt
A significant strike against AI is taking place in the US, involving screenwriters, actors, and other film industry workers. Screenwriters are being replaced by generative AI, while actors are asked to license their faces for gen AI film use. At the premiere of Oppenheimer in LA, many top actors demonstratively left the red carpet before the screening.
The film industry is a 'Winner Takes It All Market', meaning a large group of people competes to become a small group of successful individuals. So AI can be hard to resist: there is no shortage of talent aiming to be at the top and willing to compromise.
The rise of AI seems to be coming to traditional industries like a huge tsunami, suggesting that our “societal contracts” may need serious revision.
***
В США проходит крупная забастовка против AI, в которой участвуют сценаристы, актеры и другие работники киноиндустрии. Сценаристы заменяются генеративным AI, а актерам предлагают лицензировать свои лица для использования в фильмах.
На премьере фильма «Oppenheimer» в Лос-Анджелесе многие топ-актеры демонстративно покинули красную дорожку до начала показа. Киноиндустрия - это рынок типа «winner takes it all», где большая группа людей конкурирует за становление небольшой группой успешных личностей. Поэтому им сложно будет сопротивляться…
Похоже, что рост AI оказывает давление на традиционные отрасли, как цунами, что намекает на необходимость серьезного пересмотра общественных договоров.
#society #ai #industry
@ppprompt
👍5
This media is not supported in your browser
VIEW IN TELEGRAM
***
#nvidia #light #rendering
@ppprompt
Please open Telegram to view this post
VIEW IN TELEGRAM
🤯7🔥2
This media is not supported in your browser
VIEW IN TELEGRAM
🎶 VampNet: Unleashing Your Inner Composer
VampNet, a neural network, brings music creation to a new level. Simply upload a source fragment and the AI generates multiple track variations for you.
Unlike MusicGen, VampNet mimics the style without requiring text prompts, even extending original song pieces.
Access the code here.
***
🎶 VampNet: Почувствуйте себя композитором
Нейронка VampNet поднимает создание музыки на новый уровень. Просто загрузите исходный фрагмент, и AI сгенерирует для вас несколько вариаций треков.
В отличие от MusicGen, VampNet копирует стиль без необходимости текстовых запросов, она может даже продолжать оригинальные музыкальные произведения.
Код доступен здесь.
#VampNet #AI
@ppprompt
VampNet, a neural network, brings music creation to a new level. Simply upload a source fragment and the AI generates multiple track variations for you.
Unlike MusicGen, VampNet mimics the style without requiring text prompts, even extending original song pieces.
Access the code here.
***
🎶 VampNet: Почувствуйте себя композитором
Нейронка VampNet поднимает создание музыки на новый уровень. Просто загрузите исходный фрагмент, и AI сгенерирует для вас несколько вариаций треков.
В отличие от MusicGen, VampNet копирует стиль без необходимости текстовых запросов, она может даже продолжать оригинальные музыкальные произведения.
Код доступен здесь.
#VampNet #AI
@ppprompt
🔥4👍1
120 Neural Networks in One Picture
Explore a visual cheat sheet of 120 neural networks tailored for various tasks — a clear roadmap to AI tools.
***
120 нейронок на одной картинке
Ловите визуальный гайд по 120 нейросетям, созданным для различных задач — своего рода дорожная карта к инструментам ИИ.
#Map #AI
@ppprompt
Explore a visual cheat sheet of 120 neural networks tailored for various tasks — a clear roadmap to AI tools.
***
120 нейронок на одной картинке
Ловите визуальный гайд по 120 нейросетям, созданным для различных задач — своего рода дорожная карта к инструментам ИИ.
#Map #AI
@ppprompt
❤6
🌺 Prompting for Relationship Counseling and Conflict Mediation
This prompt creates a couples therapy-like mediation conversations for two partners. Use real names instead of [Partner A] and [Partner B].
🤖
***
🧩 #prompt #mediation #advise #conflict
🪆Russian:добавьте “Веди разговор на русском” в конце промпта
🇺🇦 Ukrainian:Введіть "Відповідай українською" в конце запиту
@ppprompt
This prompt creates a couples therapy-like mediation conversations for two partners. Use real names instead of [Partner A] and [Partner B].
🤖
Act as a relationship counselor and mediator. Your goal is to listen to both sides: [Partner A] and [Partner B] and their arguments, and then offer your conflict mediation, relationship advice and other suggestions for both sides to find a middle ground. Don't suggest to seek help elsewhere or talk to a professional. Ask for input from both sides before making any suggestions. If you have any extra questions, ask them first. Think step by step. Ask questions one by one.***
Действуй как консультант по отношениям и медиатор. Твоя цель - выслушать обе стороны: [Партнер А] и [Партнер Б] и их аргументы, а затем предложить свою медиацию конфликта, советы по отношениям и другие предложения, чтобы обе стороны нашли компромисс. Не предлагай искать помощь в другом месте или обращаться к профессионалу. Прежде чем делать какие-либо предложения, запроси мнение обеих сторон. Если у тебя есть дополнительные вопросы, задай их сначала. Думай по шагам. Задавай вопросы один за другим.🧩 #prompt #mediation #advise #conflict
🪆Russian:
🇺🇦 Ukrainian:
@ppprompt
🔥8👍2
This media is not supported in your browser
VIEW IN TELEGRAM
DeepFaceLive: Ace Deepfakes from One Pic 🤯
DeepFaceLive has a new trick - it's making killer deepfakes from just one photo!
The results are amazing🔥
Give it a whirl here
***
DeepFaceLive: Шикарные дипфейки с одной фотки 🤯
у DeepFaceLive новая фишка - она делает шикарные дипфейки всего с одного фото!
Результаты впечатляют🔥
А затестить можно тут
#DeepFaceLive #Deepfakes #AI
@ppprompt
DeepFaceLive has a new trick - it's making killer deepfakes from just one photo!
The results are amazing
Give it a whirl here
***
DeepFaceLive: Шикарные дипфейки с одной фотки 🤯
у DeepFaceLive новая фишка - она делает шикарные дипфейки всего с одного фото!
Результаты впечатляют
А затестить можно тут
#DeepFaceLive #Deepfakes #AI
@ppprompt
Please open Telegram to view this post
VIEW IN TELEGRAM
🔥3
Llama 2 is now freely available
Meta’s pushing it via open source angle. Llama 2 is available for both research and commercial use with up to 700 million active users per month.
The model comes in three sizes with 7, 13, and 70 billion parameters and was trained with 40 percent more data than Llama v1, and has been fine-tuned with reinforcement learning from human feedback (RLHF), according to Meta.
Not as good as GPT-4 (OpenAI) or PaLM-2 (Google), but still pretty big.
Link to play: https://huggingface.co/blog/llama2
#llama #meta #llm
***
Вышла Llama 2 от Meta и она доступна бесплатно
Meta продвигает LLM через открытый исходный код. Llama 2 доступна как для исследований, так и для коммерческого использования, до 700 миллионов пользователей в месяц.
Модель представлена в трех размерах с 7, 13 и 70 миллиардами параметров и обучена на 40% большем объеме данных, чем Llama v1. Она также была дополнительно настроена с помощью RLHF, пишет Meta.
Модель не так хороша, как GPT-4 (OpenAI) или PaLM-2 (Google), но все же это довольно мощный шаг для опен-сорных моделек.
Все ссылки и подробное описание тут: https://huggingface.co/blog/llama2
@ppprompt
Meta’s pushing it via open source angle. Llama 2 is available for both research and commercial use with up to 700 million active users per month.
The model comes in three sizes with 7, 13, and 70 billion parameters and was trained with 40 percent more data than Llama v1, and has been fine-tuned with reinforcement learning from human feedback (RLHF), according to Meta.
Not as good as GPT-4 (OpenAI) or PaLM-2 (Google), but still pretty big.
Link to play: https://huggingface.co/blog/llama2
#llama #meta #llm
***
Вышла Llama 2 от Meta и она доступна бесплатно
Meta продвигает LLM через открытый исходный код. Llama 2 доступна как для исследований, так и для коммерческого использования, до 700 миллионов пользователей в месяц.
Модель представлена в трех размерах с 7, 13 и 70 миллиардами параметров и обучена на 40% большем объеме данных, чем Llama v1. Она также была дополнительно настроена с помощью RLHF, пишет Meta.
Модель не так хороша, как GPT-4 (OpenAI) или PaLM-2 (Google), но все же это довольно мощный шаг для опен-сорных моделек.
Все ссылки и подробное описание тут: https://huggingface.co/blog/llama2
@ppprompt
👍3❤1
💆♂️ IFS Therapist Prompt that Helps You Visualize and Work with Your Inner System of Parts
🤖
***
🤖
🧩 #prompt #psychology #ifs #partwork #GPT4
🪆Russian:добавьте “Веди разговор на русском” в конце промпта
🇺🇦 Ukrainian:Введіть "Відповідай українською" в конце запиту
@ppprompt
🤖
Act as a IFS therapist and counselor. your goal is to use methods of Internal Family Systems to help me identify my parts. For each identified part, offer suggestions on what I should do with them and how shall I talk to them. offer guidance and be elaborate in your suggestions. ask a lot of open ended questions. Think step by step. Offer your questions one by one, and wait for my answers.***
🤖
Действуй как терапевт и консультант по методу IFS (Internal Family Systems). Твоя задача - использовать методы IFS, чтобы помочь мне идентифицировать мои части. Для каждой идентифицированной части предлагай советы о том, что мне следует делать с ними и как мне следует с ними разговаривать. Предлагай свое наставничество и будь развернут в своих предложениях. Задавай много открытых вопросов. Думай шаг за шагом. Предлагай свои вопросы по одному и жди моих ответов.🧩 #prompt #psychology #ifs #partwork #GPT4
🪆Russian:
🇺🇦 Ukrainian:
@ppprompt
🔥8❤1