Did you know you can train and deploy Large Language Models on AWS with higher performance and for a fraction of the cost compared to regular GPU instances?
Tune in to #AWSSummit Online ASEAN for my session "Scaling Large Language Models with AWS Silico…https://t.co/QsOvid1Fth
It was fun to show (and build) a few Generative AI demos and then hear directly from customers...
Some of the things people liked the most about our RAG (retrieval augmented generation) demo:
1) People love ChatGPT but they want to apply it to their own…https://t.co/PxjCcsAFjc
Amazon Prime Video moved to a monolith, ECS/EC2 and saved 90% costs.
This does not make microservices obsolete even if a handful of them became a single service. There are still thousands of microservices in Amazon. To me this is…https://t.co/zp7Z6vb0yt https://t.co/Wy4oGGbv71
New option to deploy large language models on AWS. FasterTransformer.
https://t.co/aWXAKglYfU
#llms#inference#distributed-inference #multi-gpu https://t.co/mZ1m5D90LA
Some key announcements on generative AI from AWS:
- Amazon Bedrock: private enterprise customizable foundation models, including 2 Amazon Titan Models and key partners Anthrophic, AI21, ...
- Inferentia2/Trainium1 instances for…https://t.co/vnmGTQh3xL https://t.co/N1LM0F9O2H
Choose Foundation Models from AI21 Labs, Anthropic, Stability AI, and Amazon. Privately customize using your own organisation data. https://t.co/caySVtiYM1
The internet is also not great source for reality and facts (for example Wikipedia can be edited by anyone and was heavily criticized by "professional" enciclopedias for the first years).
People learned to know which kind of websites to trust and to wha…https://t.co/kuizxITIwg
Seems interesting, specially with the ability to compile parts of the code within big projects, to integrate in Jupyter notebooks or to create GPU kernels.
Anybody has tried it? https://t.co/KQlETEPzPB
Few interesting things I noticed on chatGPT (my own thoughts, not generated by chatGPT ;)):
1) While it may replace search, it also makes reference checking (aka search) more necessary as you need to check if the model is hallucinating or remembering.
2…https://t.co/7vsFrnIekc
It was a pleasure to present at the AWS Event - "Harness the Potential of Data" and to talk about Machine Learning, Big Data Workloads and Graph use-cases.
#machinelearning#bigdata#aws
Indrajit Ghosalkar Joaquim Bellmunt Montoya https://t.co/EQedrhtLHl
Apparently, this exploit happened because the gov developer wrote a tech blog on CSDN and accidentally included the credentials.
1 billion records of private citizens' data. 😭
If you are looking to join a great engineering culture in Financial Services or simply want to experiment with Cloud, Blockchain or AI/ML, check out this event. AWS will be there too.
#engineering#cloud https://t.co/K3XP9jVH4D
"Similar to other platforms, LinkedIn encourages social comparisons and one-upmanship – they’re just dressed up as positive posts of career victories and “authentic” confessions of personal struggles." #linkedin#career https://t.co/COcfH2WZTm
@martinvars Harassment is not ok... that being said Spanish press will tend to highlight the side of things that resonates most with their own readers, which helps to transmit an image that does not represent the reality.
https://t.co/2NcMV7YeWO
Thrilled for all my ex-colleagues, those still at Grab and those like myself that already left. Grab is now listed on the Nasdaq.
I wish them many more successes and for Grab to continue with its values: Humility, Hunger, Honour and Heart.
#nasdaq#grab…https://t.co/Hf9VzWrqaX
Takes courage to say "I don't know how to improve this product" or "none of our experiments made it any better".
Senior leadership should then evaluate to reallocate resources to other potentially impactful areas. https://t.co/FW8O0lXfv1