There’s a fun theory I saw about degraded ChatGPT performance: It’s worse because the system prompt says it’s December (holidays), in the same way that “tipping” it $200 makes it better.
GENERATIVE AI
-
Llama3 Release Expected February 2024 with 120B Model
By
–
Februar 2024 Llama3, different model sizes, best one 120b as leaks predict so far.
-
Poe Android: Default Bot Selection and Retry Features
By
–
New on Android: You can now set your default bot on Poe to any of the bots in your list. Visit the settings page and select your preferred bot to set it as your default.
— Poe (@poe_platform) 8 décembre 2023
We’re also rolling out a retry button, enabling you to regenerate results from any bot with one click. pic.twitter.com/dsNEUkYusxNew on Android: You can now set your default bot on Poe to any of the bots in your list. Visit the settings page and select your preferred bot to set it as your default. We’re also rolling out a retry button, enabling you to regenerate results from any bot with one click.
-
Mistral Releases 8-Head MoE Model Available via Torrent
By
–
1) Mistral just released their 8-head MoE model. 2) Pretty wild that major AI models are being released as torrent magnets on this platform.
-
Intel Gaudi-2 Accelerates Model Training and LLM Services
By
–
🗲#IntelAmbassador From demonstrating 3 x faster model training on Gaudi-2, through to Intel ’s LLM-as-a-service, the #IntelInnovation event signposted just how the next era of digital transformation is already upon us & it's empowered by 5 tech superpowers 💪
— Sen. Sally Eaves (@sallyeaves) 8 décembre 2023
🔸In this new ‘top… pic.twitter.com/nmGPtnzLIs🗲#IntelAmbassador From demonstrating 3 x faster model training on Gaudi-2, through to Intel ’s LLM-as-a-service, the #IntelInnovation event signposted just how the next era of digital transformation is already upon us & it's empowered by 5 tech superpowers In this new ‘top
-
RAG Performance Limits and Multi-Assistant Solutions
By
–
RAG performance will degrade beyond that point, better to use multiple assistants today but it’s possible we’ll expand the limit in the future!
-
Groq Presents LPU Inference Engine with Superior GenAI Performance
By
–
We can only speak for #Groq – Groq™ is a #GenAI solutions company & the creator of the LPU™ Inference Engine, the fastest language processing accelerator on the market, delivering 10x better performance compared to GPUs. We call that #GroqSpeed. #LLMs run #BetterOnGroq. #GroqOn
-
GroqOn: Shared Link on X to Tempura V1
By
–
#GroqOn
https://x.com/TempuraV1/status/1731086948435972249
… -
Groq’s Mission: Democratizing AI Access While Preserving Human Agency
By
–
"[Our mission is] to make sure that everyone has access to AI and that we all can benefit from it … we continue to have human agency. And so we need to be essential to the infrastructure to help guide that.” — @JonathanRoss321
. Read more here:
