After Andrew Ng's original Stanford AI Coursera, what? | Hacker News Reader

After Andrew Ng's original Stanford AI Coursera, what?

41 pointsreactspa8 comments

In 2017, "a friend"* did Andrew Ng's original Stanford AI Coursera course.

- it renewed their understanding of Matrix Algebra and more importantly "why" they need it (being from India, they were fortunate that they learned matrices by rote, and the Indian education system ensured that details like "why" on earth they needed to study this thing never had a chance to pollute their pristine mind /sarcasm).

- it gave them an intuitive understanding of how neural networks are a sort of regression (those weights are akin to regression weights).

- it showed them some crazy things, such as how an image is fed into a neural network as a sequence of pixels. They suspect the way humans view images is more holistic... all pixels at the same time.

They did it as a lark. Because pmarca said that he would work on BTC or/and AI if he were young again (they did the BTC Princeton Coursera too).

They're glad they did Ng's course back then. Wonderful teacher, a gift to humanity.

Since then, they've not done anything with AI (they're gratefully and gainfully employed, albeit in an unrelated area). But they kept an eye on developments in the space, such as "tensors".

They're beginning to see some urgency in AI now. The noise is building up to a crescendo.

They want to jump in again, and dedicate every spare hour to learning more. Where it will take them, who knows. With gratitude, they don't currently need it to earn their bread. But who knows why they might need it in future.

However, they don't want to read any more "think pieces" in Bloomberg or Forbes about how AI will affect the future. They want to get busy learning the MVK (minimum viable knowledge) for actually building something useful, and alone.

The "alone" part is important.

So, they need to learn the AI equivalent of dev ops, front end, and back end. But not too much (to start with). Just enough. Give them a curriculum, HN. FLOSS/FOSS of course. Please, and thanks in advance.

----------

* Forgive me for trying Taleb's "a friend" instead of using first person. He says it gets better results.

0. ML by Ng @Stanford

1. Deep Learning Specialization by Ng @Coursera

2. Deep Learning for Coders with fastai and PyTorch by Jeremy Howard

3. Learn PyTorch really well. Either use d2l.ai or the PyTorch book from Manning.

    START MAKING PROJECTS HERE

Now it depends where you want to go.

>> Jump the hype train of LLM?

1. Natural Language Processing specialization @Coursera

2. Multilingual NLP from CMU

3. Learn using HuggingFace with either their book or course.

Your path from here will be visible to you.

>> Jump the hype train of multimodal AI like Stable Diffusion, Dall-E, etc.?

1. Diffusers course from HuggingFace.

2. Plenty of resources out there in the forms of blogs, oss projects, etc.

>> Want Edge AI, Federated Learning, or Deep RL?

There are great resources like:

1. Deep RL course from Hugging Face

2. Deep Reinforcement Learning in Action by Zai, Brown

3. Federated Learning course by Openmined

4. Edge AI course from Harvard at edX

5. Edge AI course from Edge Impulse

There are also cool stuff happening in the Math+DL and Science+DL spaces. There are nice resources for those, too.

A bit of advice, don't take too many courses and focus on building stuff, and take courses when you think that a gap exists in your knowledge.

Deep Learning from NYU is taught by Yann LeCun and Alfredo Canziani, and it is pure gem.

You will feel the need of learning math when you advance.

Mathematics for ML from Imperial College London is great.

So is Strang's Linear Algebra.

Much appreciated, thank you.

For the 'backend' part:

Andrej Karpathy's Zero to Hero

Deep learning for coders part 1 (2022 edition).

Deep learning for coders part 2 (2022 edition), lectures 9-10 are available so far. The rest will apparently be published in a few weeks.

For prompt engineering:

Follow Riley Goodside on Twitter.

For frontend:

Build something with Streamlit or Gradio.

Try www.theschoolofai.in The live course happen only 2 times a year, and is one of the most rigorous out there.

They should ask chatGPT.

Excellent advice. Will do.

Here's what I got:

After completing Andrew Ng's Machine Learning Specialization on Coursera, there are several AI courses you could consider taking next, depending on your interests and career goals:

Deep Learning Specialization: This specialization, also by Andrew Ng, covers the foundations of deep learning, including Convolutional Neural Networks (CNNs) and Recurrent Neural Networks (RNNs).

Artificial Intelligence A-Z™: This course offers a comprehensive overview of the field of AI, covering topics such as supervised and unsupervised learning, reinforcement learning, and artificial neural networks.

Natural Language Processing Specialization: This specialization focuses on the application of machine learning to natural language processing (NLP) tasks, such as sentiment analysis, text classification, and machine translation.

Advanced AI: Deep Reinforcement Learning in Python: This course covers advanced topics in reinforcement learning, including deep reinforcement learning and Q-learning.

Ethics and Society of AI: This course focuses on the ethical and social implications of AI, including topics such as bias in AI, privacy, and the future of work.

It's important to keep in mind that these courses offer different perspectives and depth of coverage, so you may want to consider your interests and career goals when choosing which course to take next.

Try the fast.ai course

Reply on news.ycombinator.com