So, to sum it up, MSFT now has:
• OpenAI’s CEO
• OpenAI’s President
• The guy who lead GPT-4
• The guy who made GPT-4 work
• The guy who polished models for release
💀
Mark Tenenholtz
7,157 posts
Mark Tenenholtz
@marktenenholtz
Member of Technical Staff @perplexity_ai
Cincinnati, OH
Joined October 2019
- Python 3.11 is out! 🎉 This is one of the most exciting releases in a while, including significant speed upgrades and better error messages. Here's what's new:
- DALLE-2 was paywall-released recently by an extremely well-funded company. Just yesterday, a group of independent researchers released their own model (Stable Diffusion) that you can use in a few lines of code for FREE. The speed of the ML research community is insane 🤯
- Hello, I've been using Matplotlib for 7+ years and I still google how to do everything except plt.plot()
- Do I have access to a debugger? Yes. Do I know how to use a debugger? Absolutely. So how do I debug my code? print('here') ... print('here2') ... print('here3') ... print('here4')
- “Our center detection model has 95% accuracy” The results: x.com/BillyM2k/statu…
- Zillow’s home buying business lost them $500,000,000, 25% of their stock value, and 25% of their workforce. How did this happen to a company with so much data on housing prices? Bad model evaluation. Here’s the fatal error they made that you must avoid when deploying models🧵
- Expectation: zoom is now training models on your telemetry data Reality: they’re training GPT-4 on your laptopI’m sorry… zoom is doing what 😵💫
- I've shared tons of free and inexpensive material for learning machine learning. Combined, their value is easily greater than a $145,092 machine learning degree. Here are the best ones:
- Google has launched Carbon: a successor to C++. It matches 100% of the performance of C++ and aims to provide a significantly better developer experience. I think this is huge news for ML. Here's why you should care (even if you've never written a line of C/C++ in your life):
- I am at least 3-5x more productive using ChatGPT to code. Not only am I faster writing code I'm familiar with, but I've even shipped apps in tech stacks I'd never used before. Here's my process, the prompts I use, and why it all works:
- Best AI advice I can give you right now: Learn how to train LLMs now. Closed source models are winning now, but small, task-specific, open-source models are the future.
- Over the last couple of years, I've spent 1000's of hours building ML models. Truth is, after using dozens of models/architectures, 99% of them are a waste of time. I start most problems with 1 of 6 architectures. Here are the best models for a strong baseline 🧵
- Someone implemented a framework to allow you to train PaLM (Google’s 540B parameter LM) using the same reinforcement learning strategy as ChatGPT! Open source strikes again. To start, here’s the part where you initially pre-train the model:
