Known as AI6, it is a structured study group organised in 50+ cities across the globe, including Bangalore, Lagos, Toronto, Singapore and Sunnyvale. It is community-driven and free-to-attend*.
AI Saturdays (AI6) is catalysed by Nurture.AI. The first cycle of AI6 began in January 2018 and ended in May 2018. Over 5,000 participants from 50 cities worldwide were part of the AI6 movement. Motivated by the overwhelming demand and positive feedback from the AI6 community, we decided to organise a second cycle for AI6 so we can continue delivering value to those who want to learn AI.
We will be kickstarting the second cycle of AI6 in August 2018 and are currently accepting ambassador applications.
*any joining fees are strictly limited to covering basic operational costs.
AI Saturdays is really exposing me to the resources and network relevant for growth. The AI6 ambassadors in Lagos took their time to explain in detail every area in the lecture videos that needs to be explained.
After the AI6 program, I have gained confidence to approach any AI problem, research on it and solve it. I also learnt to read AI related papers and publications that are necessary for development.
Thank you AI Saturdays (AI6).
I enjoyed and learned a lot of new ways of doing things and learning on a continuous basis. Importantly, I came to know interesting people and network with like-minded people. The voluntary instructors (ambassadors) were a great source in making understand various concepts.
Thanks for the great opportunity provided to us to learn Deep learning and I look forward to attending more events like this.
AI6 is a wonderful community whether you come as a beginner or an expert. I can't imagine better ways to improve on it. The content was wonderful, the organisation and preparation was on target. What I liked best was the incredible interactive sessions we had, the community that developed with all members, and the incredible learning process of working in teams. It was an amazing learning experience. It has really been an eye-opener for me as a student and also helped me grow as a developer.
Being one of the founding ambassadors of AI Saturdays, Dhaka (Bangladesh) chapter, I had the opportunity to get to know an exceptional group of like-minded people in my home city who are interested in AI education, with whom I formed a dedicated AI learning community in Dhaka. This had been a first of its kind initiative in our country. It would not have been possible without the support from Nurture.ai, who provided us with much needed educational resources and insights on conducting efficient sessions.
We hope to see the network expand to more cities and bring together more communities of learners and practitioners to benefit from shared knowledge under a common platform.
All testimonials have been edited for brevity.
See what the community has accomplished
Using Artificial Neural Networks, Zaim coded up a model to predict the stock market. He was pleasantly surprised to achieve more than 90% accuracy using just a one layer of an artificial neural network involving 1.5 million data points.
He believes the critical success factor was putting a significant effort in feature engineering; one also needs to have a mental model of how things will work instead of blindly trying out different things. This is especially important for complex problems, where pulling a proverbial brick will collapse the whole structure. If there is more time, he would have tried the same approach on other publicly available data such as health records that will more likely have a significant impact.
Since graduating from AI Saturdays, he has been working on a novel Artificial Intelligence algorithm that can see the solution space.
When we were simultaneously learning cs224n and cs231n in AI Saturdays, I got interested in intersections between the two courses. These intersections are very close to the AI we all dream of achieving - an AI that can understand both visual and language signals and respond to them. I thought of implementing an AI of that sort; and to make the task challenging I decided to implement the whole thing from scratch.
I learned how life was before these amazing deep learning frameworks. Implementing RNN, LSTM, CNN from scratch was challenging but at the same time was a great learning experience. Implementing backward pass used to take a lot of time and focus. Sometimes I spent hours understanding and deriving the backward pass by taking manual gradients. After all that I can say that it was worth doing.
If I had more time, I would also have implemented the cutting edge architectures in place of an LSTM like AWD-LSTM, GRU etc. I would also have tried attention mechanisms, which would have improved my results greatly.
See Divyansh’s project here.
The problems faced by other Virtual Assistants were that they took a huge amount to data and computational power to train. However the VANI was trained using much lesser data and the responses which it generated were still considerate.
This is a sequence-to-sequence model, written from scratch in tensorflow. Additionally, its application was not at all domain specific as the model can be applied to machine translation, text summarization, etc. The data was collected by web scraping through reddit and similar sites, on which it was trained.
The Assistant is capable of chatting with the user through texts and messgages. It also responds to trigger actions:
See their project here.
Motivation of this project came from the question of how a machine can work intelligently like a human. See their project here.
An Adversarial Autoencoder on Tensorflow for clustering mental health issues with datasets from ABIDE et al, part of international project AI-ON Psychiatric Disorders Manifold.
We provide you a community-centered learning framework, which can be adapted to your local chapter.
Weekly study sessions are conducted in each participating city and are facilitated by local ambassadors.
We believe in learning by doing - our lesson plans come with hands-on coding assignments.
AI6 participants are encouraged to participate and contribute to discussions on the AI6 forums.
This is what an AI6 meetup might look like every week. As every local chapter is bound to be unique, ambassadors have the autonomy to decide on what works best. The AI6 community has compiled a list of useful lesson materials here.
10am - 12pm
Choose at least one of the following MOOCs:
Fast.AI / Coursera Machine Learning / Deep Learning Specialization / Standford Computer Vision (CS231n) / Stanford Natural Langague Processing (CS224n) / UCL Reinforcement Learning / Berkeley CS294 / MOOC of your choice.
12pm - 1pm
1pm - 3pm
|Work on assignments / Code AI Project|