The #Nazi salute was the final straw - I’m no longer using X. If you’re a #student or #researcher with questions about #astrophysics or interested in discussing projects, positions, or collaborations, feel free to reach out to me on #Bluesky. Cheers!
🎉 @FortuneMagazine has recognized us as no. 6 on the 2024 #Future50 list! 🎉
SambaNova continues to push the boundaries of what's possible with AI, and we’re excited to unlock new possibilities for our customers and the world at large. 🚀
https://t.co/PJ11xw7WJo
#Fortune50 #GenerativeAI #AI
We teamed up with @huggingface to launch a new integration that lets devs deploy #ChatGPT-like interfaces with a single button click, reducing deployment time from hours to minutes. ⏰
Want to try? Learn more in the article 👇
https://t.co/bflK6oquTE
For those of you who asked me where else could we go to find life, here are a few possibilities, except these are very very very FAR!
#exoplanets#nasa#habitable
Beautiful launch! Falcon heavy with Europa Clipper off unveiling the mysteries of an enormous ocean under the icy crust of Jupiter��s moon Europa, our best chance for finding the makings of life beyond Earth 👽
#nasa #europaclipper #jpl
NASA’s Europa Clipper is launching soon! So exciting, can’t wait for the results.
Here is the link in case anyone wants to watch the launch: https://t.co/fv1pOulIxD
#Nasa#europaclipper#launch
Try it out here for blazing fast @OpenAI's O1 style test time scaling powered by @SambaNovaAI cloud API on your favorite opensourced models from @AIatMeta .
Many thanks to @_akhaliq, who put tremendous works in making the gradio app working! 🤗
https://t.co/29LN5onzic
SambaNova has launched their cloud inference platform, now allowing developers to access Llama 3.1 8B, 70B and 405B on their custom AI chips
@SambaNovaAI has set a record for inference on @AIatMeta's Llama 3.1 405B - serving the model in native 16-bit precision and achieving 132 output tokens per second.
SambaNova's API inference offering is powered by their SN40L custom AI chip that uses their Reconfigurable Dataflow Unit architecture. The SN40L is manufactured on a 5nm TSMC process and includes a unique combination of DRAM, HBM3 and SRAM on each chip. This is the first time that SambaNova is making inference on their custom chips widely available to developers.
For the first time, this means that Llama 3.1 405B is available with output speeds that are faster than the frontier model APIs offered by leading AI labs (GPT-4o, Gemini 1.5 and Claude 3.5 Sonnet). This is a significant moment for frontier-class open weights AI.
Summary of SambaNova output speeds, as measured for Artificial Analysis leaderboards:
‣ Llama 3.1 8B: 1066 tokens per second
‣ Llama 3.1 70B: 461 tokens per second
‣ Llama 3.1 405B: 132 tokens per second
All three endpoints are now listed on Artificial Analysis. SambaNova’s pricing is competitive for all models (input/output per million token prices are $0.1/$0.2 for 8B; $0.6/$1.2 for 70B; $5/$10 for 405B). SambaNova’s endpoints for all three models currently support a context window of 8192 tokens.
See below for further analysis and links to our full results.
Happy 25th anniversary, Chandra!🥳 25 new images to celebrate our special day, from nearby supernova remnants to groups of galaxies more than 10 billion light-years away! Experience them all at: https://t.co/8tqJTvR7Dn #Chandra25
25 new images to celebrate Chandra's 25th anniversary! Launched into space on July 23, 1999, Chandra continues to discover new phenomena and examine old mysteries from within our Solar System out to nearly the edge of the observable Universe. 🎉https://t.co/8tqJTvR7Dn #Chandra25
@whereisyvette@CenterForAstro@uoregon Congratulations Yvette! It was so lovely meeting you at CfA and the Scotch hours, hopefully see you again in Oregon.
Woohoo! 🤩🚀 What a fantastic first launch of #Ariane6! Absolutely breathtaking to witness this powerhouse in action. And with it, Europe regains the ability to independently deliver valuable payloads into Earth orbit and all the way to the #Moon 🌕 Huge congratulations to all the brilliant and hard-working people behind this success! 🥳 #GoArianeGo
Super grateful to have been a part of the @ROGAstronomers Think Space Lecture series and #BritishScienceWeek. Thank you to the ~200 people who participated in excellent discussion after. Check out my talk, Diving into the Dark: #SupermassiveBlackHoles -
https://t.co/RQdDODG5Ob
«If you jump into a black hole, your mass energy will be returned to our Universe, but in a mangled form, which contains information about what you were like, but in an unrecognisable state. It is like burning an encyclopedia. Information is not lost, if one keeps the smoke and the ashes. But it is difficult to read»
— Stephen Hawking, Information Loss in Black Holes, 2005
Prepare for today’s eclipse with a journey through this observatory! Explore the mesmerizing corona and dynamic CMEs to deepen your understanding of our favorite star https://t.co/VQMR3oTu8S
Ps: Avoid challenging the sun to a staring contest without protection! #SolarEclipse2024
@ROGAstronomers Thank you so much for giving me this wonderful opportunity to talk about my work and discuss the workings of the universe with your audience. Looking forward to it!