Focus AI #104
#EuroPython #Renault #DataViz #Bloom
Focus AI, the latest AI news in 3 minutes
Edition #104 - November 18th 2023
Feel free to subscribe or to share it.
This week:
🧑🏫 Destination AI: Introduction to Artificial Intelligence
🚗 Renault and Google team up to develop a ‘software defined’ vehicle
🦐 Those cool image generators are perpetuating bias
🗺️ BLOOM: Hundreds of researchers make an open source GPT3 using a French supercomputer
And about code:
[🖼️ Poster] Data Viz
[📰 Post] No, You Don’t Need MLOps
[🧑🏫 Class] Advanced NLP
[🗣️ Product] Deepgram: Speech-to-text
[🎤 Conf] Videos from EuroPython 2022
== News ==
🧑🏫 Destination AI: Introduction to Artificial Intelligence
This course aims to demistify AI. OpenClassRoom and Institut Montaine made it. If you are getting bored by people asking what is AI, just send them this course. There are 3 parts
what AI really is (and what it's not!)
the impact of artificial intelligence on the workplace and on society
behind the scenes of an artificial intelligence project and get familiar with concepts like deep learning and machine learning.
Along the way, you will be hearing directly from some of the most influential leaders in AI today. They will be answering questions about where AI is today and where it's going in the future.
🚗 Renault and Google team up to develop a ‘software defined’ vehicle
French automaker Renault is partnering with Google to develop its cars like a tech company makes software.
The pact, which expands on a previous collaboration between the two firms, will see Renault commit to making what it calls a “software defined” vehicle using technology from Google’s cloud division.
Using artificial intelligence, the two companies plan to create a “digital twin” of a new vehicle. Digital twins aim to replicate physical objects in a virtual setting so that they can be tested and monitored before real-world deployment.
🦐 Those cool image generators are perpetuating bias
Researchers with Stanford University, Columbia University, Bocconi University, and the University of Washington have studied some of the biases that manifest in image generation models, like Stable Diffusion and DALL-E. The research, unsurprisingly, finds that these image generators both perpetuate biases and, more troublingly, amplify them (as in, they tend towards displaying more acute biases than the underlying datasets used to train the models).
🗺️ BLOOM: Hundreds of researchers make an open source GPT3 using a French supercomputer
Hundreds of researchers from around the world spent a year training a GPT3-style model called 'BLOOM', then released the models and code, and now they've released a research paper documenting the model and training process. Overall, BLOOM is a big deal - though the BLOOM model isn't the best available language model you can get, the fact BLOOM was developed at all is a milestone in AI research, showing how distributed collectives can come together to train large-scale models.
Where the compute came from: BLOOM is also an example of nationalistic AI ambitions: "The compute for training BLOOM was provided through a French public grant from GENCI and IDRIS, leveraging IDRIS’ Jean Zay supercomputer" - in other words, some parts of the French government essentially sponsored the compute for the model. French AI startup HuggingFace led a lot of the initial work, though "in the end, over 1200 people registered as participants in BigScience", spanning 38 distinct countries.
"Training BLOOM took about 3.5 months to complete and consumed 1,082,990 compute hours. Training was conducted on 48 nodes, each having 8 NVIDIA A100 80GB GPUs (a total of 384 GPUs)".
== Code & Tools ==
Let’s dive in into more technical news.
[🖼️ Poster] Data Viz
Data Viz leads you to the most appropriate graph for your data. It links to the code to build it and lists common caveats you should avoid.
[📰 Post] No, You Don’t Need MLOps
In this article, the author takes a back-to-first-principles approach to the MLOps field, talks about the original problems, how they can be elegantly addressed in today’s ML frameworks, and why any additional complexity beyond these keep-it-simple (KISS) solutions is unnecessary.
[🧑🏫 Class] Advanced NLP
Graham Neubig's "Advanced NLP" is one of the best resources you'll find for current state-of-the-art techniques and algorithms in modern NLP.
Follow the links for the slides and an awesome collection of readings and resources.
[🗣️ Product] Deepgram: Speech-to-text
Understand real-world audio with one easy speech API.
Unlock voice intelligence with deep learning. Get powerful transcription with understanding for diarization, summarization, language detection, topic detection, and more.
[🎤 Conf] Videos from EuroPython 2022
Videos from EuroPython 2022, held online and in Dublin, Ireland from 11 - 17 July 2022!
#EuroPython #Renault #DataViz #Bloom
A big thanks to our sources: https://jack-clark.net/, https://www.actuia.com/, https://thevariable.com/news/, https://techcrunch.com/, https://read.deeplearning.ai/the-batch/
What about you? Have you noticed something else?
Don’t miss the next news!
Have a good week-end,
Maxime 🙃 from Toulouse, France with 🌺