LLM censorship should be an opt-in feature, not a default. No one would buy a keyboard that doesn’t let you type bad words.
LLMS
-
Groq’s LPU Hardware Breakthrough Enables Near-Instantaneous LLM Response Times
By
–
Groq's recent hardware breakthroughs have been going viral on X.
— Rowan Cheung (@rowancheung) 21 février 2024
Groq (not Grok) uses LPUs instead of GPUs, allowing the chatbot to run LLMs at nearly instantaneous response times.
This unlocks a whole new world of potential AI and user experiences.pic.twitter.com/EnJnd3jEQmGroq's recent hardware breakthroughs have been going viral on X. Groq (not Grok) uses LPUs instead of GPUs, allowing the chatbot to run LLMs at nearly instantaneous response times. This unlocks a whole new world of potential AI and user experiences.
-
OpenAI Doubles GPT-4 Turbo Rate Limits to 1.5M Tokens
By
–
OpenAI also doubled rate limits for GPT-4 Turbo now reaching a maximum of 1.5M tokens per minute while also removing daily limits. Good news for developers. But when Sora @openai
?? -
Musk reveals Midjourney X partnership amid major AI developments
By
–
NEWS: Elon Musk has revealed a potential Midjourney and X partnership. Plus, major developments from Neuralink, Grok, Adobe, OpenAI, Groq, and a new AI workplace study. Here's everything going on in AI right now:
-
Nous-Hermes-2 Quantized to 4-bit for MLX Apple Silicon
By
–
I just quantized this amazing model to 4-bit, with support for the MLX platform so you can run it super fast on Apple Silicon https://
huggingface.co/mlx-community/
Nous-Hermes-2-Mistral-7B-DPO-4bit-MLX
… -
Speculative Streaming: Fast LLM Inference Without Auxiliary Models
By
–
Speculative Streaming: Fast LLM Inference without Auxiliary Models Bhendawade et al.: https://
arxiv.org/abs/2402.11131 #ArtificialIntelligence #DeepLearning #MachineLearning -
Multimodal models struggle with transfer learning to text despite image knowledge.
By
–
Also weird there’s no obvious transfer learning back to text. For all the ineffable, AGI-essential knowledge supposedly in images, multimodal models seem no better at spatial reasoning word problems, creating SVGs, designing web UI, or drawing ASCII art.
-
AI Agents Reasoning Code at Scale GTC24
By
–
Discover the next wave of #AI on Wednesday, March 20 at #GTC24. Join @imbue_ai CEO Kanjun Qiu and NVIDIA VP of Applied Deep Learning Research Bryan Catanzaro for a discussion on AI agents that reason code at scale. https://
nvda.ws/48jEAEm -
Deploying Long-Context LLMs: Practical Challenges at GTC24
By
–
Inference with long contexts has become increasingly important for building real-world applications with #LLMs. Join experts at #GTC24 to explore the practical challenges of deploying long-context LLMs.
— NVIDIA AI (@NVIDIAAI) 20 février 2024
Register here: https://t.co/0jCrgfPHEA pic.twitter.com/C0T5WcTNVhInference with long contexts has become increasingly important for building real-world applications with #LLMs. Join experts at #GTC24 to explore the practical challenges of deploying long-context LLMs. Register here: https://
nvda.ws/3UsuFsL -
Gemini 1.5 Pro Model Guide Released With 1M Token Context
By
–
The Gemini 1.5 Pro model guide is live! With support of up to 1 million tokens context length, you may be wondering what's possible with Gemini 1.5 Pro. My overall impression after our first round of testing is that Gemini 1.5 Pro is among the most powerful long context LLMs