1/ My notes from @scale_AI Transform✍️It covers:
- Building good data
- Future of ML frameworks
- Challenges for scalable deployment
- How to assess ML maturity
Thanks @alexandr_wang and team for organizing the best AI conference of 2021 thus far. Enjoy!
2/ @AndrewYNg: "The single most important thing that an MLOps team needs to do is to ensure consistently high-quality data throughout all stages of the ML project life cycle."
3/ @fchollet
The 4 #DeepLearning Trends That Keras Will Support:
- An Ever-Growing Ecosystem of Reusable Parts
- Increasing Automation
- Large-Scale Workflows In The Cloud
- Real-World Deployment
4/ @chuchenghsieh
The 5-Level Framework To Assess AI Maturity:
- Level 1: Conductor
- Level 2: Practitioner
- Level 3: Craftsman
- Level 4: Adventurist
- Level 5: Pioneer
5/ @snsf
The 5 Challenges In Deploying At Scale @facebookai:
- Scaling Training Data
- Efficient Training and Inference
- Building AI Platforms
- Developing AI Responsibly
- Creating Best Practices
6/ @andyfang: @DoorDash uses AI to scale their marketplace and make it easier for customers to find what they want while making it easier for merchants to position themselves in an online world - ranging from creating a rich item taxonomy to building an optimized delivery menu.
7/ @soumithchintala
The 3 Possible Futures of ML Frameworks:
- Transformers + ConvNets = Stable Architectures
- Scaling a currently niche fundamentally different architecture to show disruptive results
- Data-efficient models via Priors
8/ Drago Anguelov of @Waymo:
- Most interesting perception problems lie in the long-tail.
- To build a robust system, we can build a network that gives a notion of its own confidence in its prediction.
- Waymo Open Dataset is a testbed for behavior prediction systems.
9/ @kevin_scott:
- AI can help tackle climate change, provide better healthcare, and redefine the workforce.
- There is an opportunity to build more efficient algorithms to train large self-supervised language models.
- We have an enormous amount of work to build Responsible AI.
10/ @jeremyphoward and @math_rachel
The 4 prongs of @fastdotai:
- Start with Education
- Move to Research
- Build Software to Implement the Research
- Community Made It Special
12/ If you had experience using tools to support building high-quality training data for the modern AI/ML stack, please reach out to trade notes and tell me more!
DM open🙋
13/ Finally, congrats @scale_AI on the massive $325M Series E funding round!
Recently finished reading @AdamMGrant so I decided to jot down these practical takeaways to improve my rethinking skills. Might be relevant to you as well! 👇
1 - Think Like A Scientist:
When forming an opinion, resist the temptation to preach, prosecute, or politick. Treat your emerging view as a hunch or a hypothesis and test it with data.
2 - Define Your Identity In Values, Not Opinions:
See yourself as someone who values curiosity, learning, mental flexibility, and searching for knowledge. Keep a list of factors that would change your mind.
During this quarantine time, I binge-watched @Stanford#CS330 lectures taught by the brilliant @chelseabfinn. This blog post is a summary of the key takeaways on #Bayesian Meta-Learning that I’ve learned. #AtHomeWithAI
Bayesian meta-learning generates hypotheses about the underlying function, samples from the data distribution, and reasons about model uncertainty. It is suitable for problems in safety-critical domains, exploration strategies for meta-RL, and active learning.
(2/7) 👇
There are various Bayesian Meta-Learners pre-neural-nets: