Reverse engineer Github copilot
recently undertook the task to reverse engineering the Github Copilor VS Code plugin. He shares his finding in a great blog post and a live demo.
Follow
Sergios Karagiannakos
@KarSergios
Author of Deep Learning in Production (amzn.to/3oa50Vj) | 
Founder of AI Summer | 
Machine Learning EngineerSergios Karagiannakos’s Tweets
Breaking news: Google invests $300mn in artificial intelligence start-up Anthropic
1
What do all these mean? It means that I can now write the same piece of code in 4 hours instead of 2 days. And if you ask me, that's a big win.
Show this thread
In my experience, they are also quite good at creating complex commands such as bash scripts, git commands, or complex SQL queries. But should not be followed blindly.
1
Show this thread
They can make your code more compact and readable.
1
Show this thread
They can also help to familiarize yourself with new tools, libraries, or best practices for writing code.
1
Show this thread
With the correct prompts, you can use them to analyze and debug code.
1
Show this thread
They can create dummy and mock data for testing.
1
Show this thread
They can help tremendously in creating unit tests, logging functionality, and documentation.
1
Show this thread
They are excellent for writing mundane and boilerplate code. No more writing regular expressions or react boilerplate.
1
Show this thread
𝐖𝐢𝐥𝐥 𝐂𝐡𝐚𝐭𝐆𝐏𝐓 𝐚𝐧𝐝 𝐂𝐨𝐩𝐢𝐥𝐨𝐭 𝐭𝐚𝐤𝐞 𝐩𝐫𝐨𝐠𝐫𝐚𝐦𝐦𝐞𝐫'𝐬 𝐣𝐨𝐛?
No! But they give an incredible productivity boost if you utilize them correctly. In the past few weeks, I have used both of them a lot when coding. Here are my two cents:
🧵👇
1
2
3
Show this thread
[AI Learning Mondays] - Efficient Python Tricks and Tools for Data Scientists - Effective Python for Data Scientists Link: hubs.ly/Q01y3W-B0
4
13
A great blog on efficient self-attention
Quote Tweet
[AI Learning Wednesdays] - An excellent write up on!!! This blog post aims to provide a comprehensive overview of the different types of efficient attention with intuitive explanations. #machinelearning #deeplearning
Link:
hubs.ly/Q01y3WMH0
1
[New article] - Learn Pytorch now from scratch: Training your first deep learning models step by step. A fully comprehensive tutorial by #deeplearning #ai #ml
4
15
In this new article, we explain the basic idea neural fields and explore the two most promising architectures: neural radiance fields (NeRFs) and instant neural graphics primitives with multiresolution hash encoding
4
15
1
4
32
Of course, no one falls perfectly into one category.
But understanding what you want, your strengths, and your weaknesses is imperative.
1
Show this thread
7) The architect: Let's build better software together
8) The CTO: Let's solve an important problem together
9) The visionary CEO: We're going to change the world
⬇️
1
1
Show this thread
3) The freelancer: I just don't want a boss
4) The dev evangelist: I have something important to say
5) The manager: How can I provide the most value?
6) The business founder: Let's build the company I always wanted to work at
⬇️
1
Show this thread
Each depends on the individual traits (skill, connection, vision) and yields different long-term results (income, stress, free time, satisfaction, impact).
1) The lone coder: Please leave me alone so I can code
2) The ladder climber: Climb the ladder and wait your turn
⬇️
1
Show this thread
𝐂𝐚𝐫𝐞𝐞𝐫 𝐭𝐫𝐚𝐜𝐤𝐬 𝐟𝐨𝐫 𝐒𝐨𝐟𝐭𝐰𝐚𝐫𝐞 𝐄𝐧𝐠𝐢𝐧𝐞𝐞𝐫𝐬
Some years ago I found one of the best talks I have ever seen on career advancement in the tech industry. According to there are 9 tracks of career progression.
youtube.com/watch?v=TrLDU6
🧵⬇️
1
3
Show this thread
I can't even begin to describe how honored I am that my name is included in the same list with all these people. Thank you
2
1
8
Excited to announce Imagen Video, our new text-conditioned video diffusion model that generates 1280x768 24fps HD videos! #ImagenVideo
imagen.research.google/video/
Work w/
58
1,011
3,415
Show this thread
The Illustrated Stable Diffusion
jalammar.github.io/illustrated-st
New post!
Over 30 visuals explaining how Stable Diffusion works (diffusion, latent diffusion, CLIP, and a lot more).
30
563
2,183
Show this thread
[New Article] How diffusion models work: the math from scratch
Diffusion models are a new class of state-of-the-art generative models that generate diverse high-resolution images.
theaisummer.com/diffusion-mode
#machinelearning #deeplearning #python #art #ai
For more details🧵⬇️
5
182
627
Show this thread
Data leakage is the use of label information in the model training process which is not available at prediction time.
Sources of data leakage according to include:
🧵⬇️
1
3
7
Show this thread
1/7 Bias-variance tradeoff is probably the single most important concept in all of machine learning.
Thread: 🧵⬇️
8
40
217
Show this thread
5/5 Make sure you understand the challenges of deploying ML systems in production using Chip Huyen's book (amazon.com/Designing-Mach)
👇
1
2
12
Show this thread
4/5 Brush up your SQL and data processing skills. Useful platforms for that are Leetcode (leetcode.com/explore/learn/), Hackerank (hackerrank.com/domains/sql), and stratascratch (stratascratch.com)
👇
1
4
4
Show this thread
3/5 Make sure your ML fundamentals are strong. Stanford CS229 Machine Learning Course is the best place for this: youtube.com/playlist?list=
👇
1
2
5
Show this thread
2/5 Practice your coding problems. Leetcode (leetcode.com) is the industry standard. Do check out neetcode.io for a shorter list of problems with their explanations.
👇
1
2
3
Show this thread
Are you preparing for a machine learning interview? Here is my 5-step process:
1/5 Master statistics and probabilities. Khan academy( (youtube.com/watch?v=uhxtUt) and Statquest (youtube.com/c/joshstarmer/) are my go-to places.
🧵👇
5
56
279
Show this thread
I am delighted to announce that the camera-ready version of my new book, "Machine Learning: Advanced Topics", is finally available online for free at probml.github.io/book2 ( will publish the hard copy in 2023.)
43
945
3,992
Show this thread
New lectures on the Geometric Deep Learning course by , , , and . And from a quick look, they seem amazing. Can't wait to watch the full course.
1
30
155
MedicalZoo is a Pytorch library for 3D multi-modal medical image segmentation. It has over 1.1K stars and 225 forks.
If you are interested in open-source contributing, feel free to ping us, join our discord server or directly open a new pull request.
10
44
Awesome online book on Bayesian approaches.
A good read for students reviewing maths for ML.
bayesrulesbook.com
11
360
1,621
Very nice paper on MLOps. It's worth a look
Quote Tweet
Machine Learning Operations (MLOps):
Overview, Definition, and Architecture
arxiv.org/abs/2205.02302
A collection of (mostly) technical things every software developer should know.
2