Media is too big
VIEW IN TELEGRAM
Китайская Unitree Robotics тестирует своих четвероногих питомцев (кожаные мешки поплатятся!).
Перспектива удешевления при массовом выпуске и оснащения такой штуки стрелковым оружием и взрывчаткой, очевидно, не за горами.
🔗 x.com/UnitreeRobotics
Перспектива удешевления при массовом выпуске и оснащения такой штуки стрелковым оружием и взрывчаткой, очевидно, не за горами.
🔗 x.com/UnitreeRobotics
Media is too big
VIEW IN TELEGRAM
Занятная WASM-демка. Для перемещения частиц нужно изменять размеры окна браузера (или просто его перемещать).
🔗 grantkot.com
🔗 grantkot.com
Высокий суд Лондона освободил под залог основателя WikiLeaks Джулиана Ассанжа, и он вылетел на родину в Австралию. В тюрьме журналист провел 1901 день, ожидая экстрадиции в США, где ему могли присудить 175 лет тюрьмы по обвинению в шпионаже. Освобожден Ассанж в рамках сделки с Минюстом США, по которой он согласился частично признать свою вину.
🔗 forbes.ru
🔗 forbes.ru
Анонимный телеграм-канал
Интересно, как быстро доступная всем генерация видео нейросетями преодолеет тот разрыв в качестве между сегодняшними дерганными «пластилиновыми мультиками» и нормальным, стабильным, приближенным к кинематографическому, видеопотоком без черри-пикинга? У картинок…
Media is too big
VIEW IN TELEGRAM
Huggingface: Introducing the Open LLM Leaderboard v2
We burned 300 H100 to re-run new evaluations like MMLU-pro for all major open LLMs!
Some learning:
- Qwen 72B is the king and Chinese open models are dominating overall
- Previous evaluations have become too easy for recent models, much like grading high school students on middle school problems
- There are indications that AI builders have started to focus on the main evaluations too much at the expense of model performances on other ones
- Bigger is not always smarter
🔗 huggingface.co/.../open_llm_leaderboard
🔗 huggingface.co/.../blog
#llm
We burned 300 H100 to re-run new evaluations like MMLU-pro for all major open LLMs!
Some learning:
- Qwen 72B is the king and Chinese open models are dominating overall
- Previous evaluations have become too easy for recent models, much like grading high school students on middle school problems
- There are indications that AI builders have started to focus on the main evaluations too much at the expense of model performances on other ones
- Bigger is not always smarter
🔗 huggingface.co/.../open_llm_leaderboard
🔗 huggingface.co/.../blog
#llm
Анонимный телеграм-канал
Из Rabbit R1 (работающего под управлением Android 13) выковыряли apk-шку приложения и запустили ее на обычном смартфоне. Соответственно, ждем запуска Doom на самом Rabbit R1. 🔗 emergetools.com/deep-dives/rabbit-r1 #AIslop #LLM
🤣 Вишенкой на торте (это очень смешно):
Rabbit data breach: all r1 responses ever given can be downloaded
Rabbit inc has known that we have had their elevenlabs (tts) api key for a month, but they have taken no action to rotate the api keys.
On may 16, 2024, the rabbitude team gained access to the rabbit codebase and found several critical hardcoded api keys in its code. these keys allow anyone to:
* read every response every r1 has ever given, including ones containing personal information
* brick all r1s
* alter the responses of all r1s
* replace every r1’s voice
🔗 rabbitu.de/articles
#AIslop #LLM
Rabbit data breach: all r1 responses ever given can be downloaded
Rabbit inc has known that we have had their elevenlabs (tts) api key for a month, but they have taken no action to rotate the api keys.
On may 16, 2024, the rabbitude team gained access to the rabbit codebase and found several critical hardcoded api keys in its code. these keys allow anyone to:
* read every response every r1 has ever given, including ones containing personal information
* brick all r1s
* alter the responses of all r1s
* replace every r1’s voice
🔗 rabbitu.de/articles
#AIslop #LLM
😁1
Анонимный телеграм-канал
🤣 Вишенкой на торте (это очень смешно): Rabbit data breach: all r1 responses ever given can be downloaded Rabbit inc has known that we have had their elevenlabs (tts) api key for a month, but they have taken no action to rotate the api keys. On may 16,…
AI hardware maker Rabbit's CEO said the company has now sold 130,000 of its R1 units and that usage of the R1 is doubling every two weeks.
🔗 forbes.com
🔗 forbes.com
Introducing GPT-4o mini in the API
Today we introduced GPT-4o mini—our new affordable and intelligent small model that’s significantly smarter, cheaper, and just as fast as GPT-3.5 Turbo—and launched it in the API. Here’s what you need to know:
* Intelligence: GPT-4o mini outperforms GPT-3.5 Turbo in textual intelligence (scoring 82% on MMLU compared to 69.8%) and multimodal reasoning.
* Price: GPT-4o mini is more than 60% cheaper than GPT-3.5 Turbo, priced at $0.15 per 1M input tokens and $0.60 per 1M output tokens (roughly the equivalent of 2500 pages in a standard book).
* Modalities: GPT-4o mini currently supports text and vision capabilities, and we plan to add support for audio and video inputs and outputs in the future.
* Languages: GPT-4o mini has improved multilingual understanding over GPT-3.5 Turbo across a wide range of non-English languages.
With its low cost and latency, GPT-4o mini works well for high-volume tasks (e.g., passing a full code base or conversation history to the model), cost-sensitive tasks (e.g., summarizing large documents), and tasks that require fast responses (e.g., customer support chatbots). Like GPT-4o, GPT-4o mini has a 128k context window, supports up to 16k output tokens per request, and has a knowledge cut-off date of October 2023. We plan to launch fine-tuning for GPT-4o mini in the coming days.
🔗 openai.com
Today we introduced GPT-4o mini—our new affordable and intelligent small model that’s significantly smarter, cheaper, and just as fast as GPT-3.5 Turbo—and launched it in the API. Here’s what you need to know:
* Intelligence: GPT-4o mini outperforms GPT-3.5 Turbo in textual intelligence (scoring 82% on MMLU compared to 69.8%) and multimodal reasoning.
* Price: GPT-4o mini is more than 60% cheaper than GPT-3.5 Turbo, priced at $0.15 per 1M input tokens and $0.60 per 1M output tokens (roughly the equivalent of 2500 pages in a standard book).
* Modalities: GPT-4o mini currently supports text and vision capabilities, and we plan to add support for audio and video inputs and outputs in the future.
* Languages: GPT-4o mini has improved multilingual understanding over GPT-3.5 Turbo across a wide range of non-English languages.
With its low cost and latency, GPT-4o mini works well for high-volume tasks (e.g., passing a full code base or conversation history to the model), cost-sensitive tasks (e.g., summarizing large documents), and tasks that require fast responses (e.g., customer support chatbots). Like GPT-4o, GPT-4o mini has a 128k context window, supports up to 16k output tokens per request, and has a knowledge cut-off date of October 2023. We plan to launch fine-tuning for GPT-4o mini in the coming days.
🔗 openai.com
Openai
GPT-4o mini: advancing cost-efficient intelligence
Introducing the most cost-efficient small model in the market
Banks, airlines, railways and media outlets hit by global outage linked to Windows PCs
Businesses including banks, airlines, railways, telecommunications companies, TV and radio broadcasters, and supermarkets have been taken offline after blue screen of death error screens were seen on Windows workstations across the globe.
Planes have been grounded as several airports are hit by a global IT outage, with Windows PCs shutting down.
🔗 news.sky.com
Businesses including banks, airlines, railways, telecommunications companies, TV and radio broadcasters, and supermarkets have been taken offline after blue screen of death error screens were seen on Windows workstations across the globe.
Planes have been grounded as several airports are hit by a global IT outage, with Windows PCs shutting down.
🔗 news.sky.com