BREAKING: OpenAI is testing a new system prompt for the ChatGPT desktop app which is focused on screen shared context. "You are assisting a user on the desktop"… ChatGPT is instructed to ask the user to "SHARE_YOUR_SCREEN_PLEASE" which would trigger a native pop-up
LLMS
-
User reports repetitive word generation in LLM blog writing
By
–
it still adds "enhanced" 100+ times into any blog post I ask it to write – my benchmark is super simple
-
First Attempt at DSPy Agents from Scratch
By
–
A first attempt at DSPy Agents from scratch https://
bit.ly/3zzKIN5
#AI #MachineLearning #DeepLearning #LLMs #DataScience -
SWE-bench: New Benchmark for Complex AI Reasoning Tasks
By
–
Esta semana escucharemos bastante sobre el SWE-bench ya que parece se convertirá en el nuevo benchmark para evaluar las capacidades de razonamiento y de resolución de tareas complejas de la próxima generación de modelos. Ya en Marzo os decía
-
Company Achieves Breakthrough via Fine-tuned OpenAI Model
By
–
El salto logrado por esta compañía es notable, y lo han conseguido a partir de hacer un fine-tuning del modelo de OpenAI con datos de cómo Ingenieros de Software razonan y resuelven este tipo de tareas.
-
RAG Systems Advances: Multi-modal Support and Production Monitoring
By
–
In the past year, RAG systems have advanced significantly. It now features: •Multi-modal documents and responses
•LLM routing based on query complexity
•Multi-lingual support
•Improved accuracy and reduced hallucinations
•Feedback and monitoring in production. A recent -
Singular Value Decomposition for Recommendation Systems
By
–
On Singular value decomposition and it’s use in recommendation systems https://
bit.ly/3zzKuFJ
#AI #MachineLearning #DeepLearning #LLMs #DataScience -
Microsoft scientist predicts continued AI capability growth ahead
By
–
Si os queréis subir a un tren del hype, mejor que sea este: el de Eric Horvitz, jefe científico de Microsoft, quien adelanta que en los próximos 18 meses no vamos a ver un "plateu" de las capacidades de la IA, sino más bien todo lo contrario.pic.twitter.com/lt0bxLBBGe
— Carlos Santana (@DotCSV) 12 août 2024Si os queréis subir a un tren del hype, mejor que sea este: el de Eric Horvitz, jefe científico de Microsoft, quien adelanta que en los próximos 18 meses no vamos a ver un "plateu" de las capacidades de la IA, sino más bien todo lo contrario.
-
VITA: Open-Source Interactive Omni Multimodal LLM Development
By
–
VITA
— AK (@_akhaliq) 12 août 2024
Towards Open-Source Interactive Omni Multimodal LLM
discuss: https://t.co/RaJdxkTEbo
The remarkable multimodal capabilities and interactive experience of GPT-4o underscore their necessity in practical applications, yet open-source models rarely excel in both areas. In… pic.twitter.com/G9h0gzLC59VITA Towards Open-Source Interactive Omni Multimodal LLM discuss: https://
huggingface.co/papers/2408.05
211
… The remarkable multimodal capabilities and interactive experience of GPT-4o underscore their necessity in practical applications, yet open-source models rarely excel in both areas. In -
Gemma Scope: Open Sparse Autoencoders for Gemma 2
By
–
Gemma Scope
— AK (@_akhaliq) 12 août 2024
Open Sparse Autoencoders Everywhere All At Once on Gemma 2
discuss: https://t.co/hLAdvC1FJC
Sparse autoencoders (SAEs) are an unsupervised method for learning a sparse decomposition of a neural network's latent representations into seemingly interpretable features.… pic.twitter.com/aXO8Ihm6TvGemma Scope Open Sparse Autoencoders Everywhere All At Once on Gemma 2 discuss: https://
huggingface.co/papers/2408.05
147
… Sparse autoencoders (SAEs) are an unsupervised method for learning a sparse decomposition of a neural network's latent representations into seemingly interpretable features.