Did a thread on this. Around 10% that answered still found GPT-4 better
LLMS
-
Gaudi 2 Achieves 28% Faster Inference Speed Than A100
By
–
On Stable Beluga 2.5 70B, our fine-tuned version of Llama 2 70B, Gaudi 2 achieved 28% faster inference speed for tokens/second per accelerator versus the A100. Read the full analysis and learn how our findings underscore the need for alternatives in compute solutions here:
-
Build Enterprise Applied AI Systems at Scale with Generative AI
By
–
Use #GenerativeAI from @AbacusAI to Build End-to-End Enterprise Applied #AI Systems at scale: https://t.co/m6xtTyEAdK = the world's first end-to-end #MLOps and #LLMOps platform where AI builds Applied AI agents and systems!
— Kirk Borne (@KirkDBorne) 11 mars 2024
——#MachineLearning #ML #DataScience #EnterpriseAI #LLMs pic.twitter.com/KkIBWsFr3HUse #GenerativeAI from @AbacusAI to Build End-to-End Enterprise Applied #AI Systems at scale: https://
abacus.ai = the world's first end-to-end #MLOps and #LLMOps platform where AI builds Applied AI agents and systems!
——
#MachineLearning #ML #DataScience #EnterpriseAI #LLMs -
Accessing X Data for Open Source LLM Development
By
–
+1 Open Source LLM I assume the access to X data will be through their API and will be pricy as hell for any 3rd party willing to build on top of it.
-
GPT-5 release enables GPT-6 discussion
By
–
The sooner GPT-5 is released, the sooner we can start talking about GPT-6.
-
7B Language Models Show Strong Mathematical Capabilities
By
–
Common 7B Language Models Already Possess Strong Math Capabilities Li et al.: https://
arxiv.org/abs/2403.04706 #ArtificialIntelligence #DeepLearning #MachineLearning -
Pi Update: Voice Quality Improvements Compared to ChatGPT
By
–
Is Pi better after the latest update? The closest I’ve gotten to feeling that presence is with ChatGPT voice but it still feels a bit more like a walkie talkie vs a seamless phone call.
-
Fine-tuning Transformers: Leveraging OSS Checkpoints for Model Adaptation
By
–
@drummatick the power of these transformer models often come from the data it’s trained on, I’d try fine tuning from an OSS checkpoint. The power of the transformer architecture is that (a) it’s easy to adapt to different problems/modalities, (b) lends itself well to broad pre
-
Claude 3 Shows Superior Reliability Over GPT-4 on Perplexity
By
–
After 100s of queries myself as a user, with Claude 3 (Opus and Sonnet) as the default model on Perplexity, I’m yet to see a hallucination. I couldn’t have said this for GPT 4. And seeing similar reports from others like @sharifshameem and @rauchg who are switching.
-
AI Hallucinations Predicted in 2001: Updated List
By
–
i predicted hallucinations in 2001; here is another list from two years ago today