They didn’t even disclose this info in “PaLM 2 Technical Report” https://
arxiv.org/abs/2305.10403
LLMS
-
PaLM 2 Technical Report Missing Information Disclosure
By
–
-
PaLM 2 Leak: Parameters and Tokens Revealed, But Key Details Missing
By
–
There’s much talk about Google’s PaLM 2 details being leaked. But seems the only thing “leaked” is unsurprising info: the model has 360B params trained on 3.6T tokens, rather than important details like dataset and architectural tricks. What’s the big fuss?
https://
cnbc.com/2023/05/16/goo
gles-palm-2-uses-nearly-five-times-more-text-data-than-predecessor.html
… -
Tiny Language Models Generate Coherent Text With Simple Vocabulary
By
–
TinyStories: How Small Can Language Models Be and Still Speak Coherent English? A tiny 10M-param transformer can generate paragraphs of coherent text and reason, when trained on synthetic stories limited to only words that a 3-4 year-old would understand.
-
Rodney Brooks Interview on GPT-4 and AI Topics
By
–
A transcription of part of an interview I did with Glenn Zorpette @Electric_Genie editor of @IEEESpectrum a couple of weeks ago. GPT and other topics. https://
spectrum.ieee.org/gpt-4-calm-down -
Vector Matching Engine for GPT-4 Applications with Langchain
By
–
We built Vector Matching Engine. You can use it alongside Langchain for GPT-4 applications. With Vector Matching Engine, you can search large amounts of vector embeddings. Ideal for language, image, and recommender models. https://
abacus.ai/vectormatching -
MedPALM2 Achieves Major Milestone in Medical AI Licensing Exam
By
–
And there's #MedPALM2 that just jumped a big notch forward over prior LLM performance for the US medical licensing exam equivalent (passing =60%) https://
arxiv.org/pdf/2305.09617
.pdf
… -
Update on Google Bard AI summarization and source-matching features
By
–
ICYMI: Google Bard AI got improved summarization and source-matching capabilities unveiled TLRD; – Summaries got better – It has access to the recent search index now – Links to the source are not really there – Facts are still made-up mostly https://
testingcatalog.com/google-bard-ai
-update-improved-summarization-and-source-matching-capabilities-unveiled/
… -
Early Bird Data Science Conference Passes Available in Europe
By
–
For just a few days more, passes to the leading #DataScience training conference in Europe start at just £99. Don’t miss out on 2 days of hands-on learning in #MachineLearning, #DeepLearning, #NLProc, #MLOps, #LLMs, #AI and more. Get your pass now: https://
bit.ly/3MxGk54 @_odsc -
Google PaLM 2 Training Uses Five Times More Text Data
By
–
Google's PaLM 2 uses nearly five times more text data than predecessor https://
cnbc.com/2023/05/16/goo
gles-palm-2-uses-nearly-five-times-more-text-data-than-predecessor.html
… -
LLM Applications Discussion at Redis Days Conference
By
–
To say I'm excited to discuss LLM applications with @tchutch94 at @Redisinc Days next week would be an understatement
— Harrison Chase (@hwchase17) 17 mai 2023
It's been awesome to work with Tyler on the open source project, and I'm looking forward to a fun conversation 🙂https://t.co/qOFYwBZMrg pic.twitter.com/Fngfsp8643To say I'm excited to discuss LLM applications with @tchutch94 at @Redisinc Days next week would be an understatement It's been awesome to work with Tyler on the open source project, and I'm looking forward to a fun conversation 🙂 https://
redis.com/redisdays/