AI is defeated by a blueberry 🫐

Plus: NASA's deep learning model 🌍, key career decisions for junior Data Scientists 📝, and stop using dict[key] to access values in python dictionaries 🛑!

Image by Élisabeth Joly on Unsplash

Happy Friday y’all,

AI may be taking over the world, but this week I took some small comfort from the fact that neither GPT-3.5 nor Bard are able to identify where the letter ‘r’ is in the word ‘blueberry’ 🫐. Sound ridiculous? I thought so too, until I saw AI vs. Blueberry, a thread. I guess our jobs are safe after all?

Image by author

Without further ado, here are 5 things you might have missed this week:

AI News

• 🔥 OpenAI announces better support for JSON (oh, and the whole GPT thing…) - I couldn’t write another edition of AIi5 without mentioning OpenAI’s DevDay. But, while everyone’s raving about the “app store for GPTs” and GPT-4 Turbo’s 128k context window, I got especially excited about the better support for JSON data. It’s the small things, right?

• 🌍 NASA’s deep learning model - NASA and IBM have released a geospatial AI foundation model for working with Earth observation data. The model has been open-sourced on HuggingFace and can be fine-tuned for image segmentation tasks like crop classification and flood identification. Check it out!

Tips & Tricks

• 🛑 Stop Using dict[key] to Access Values in Python Dictionaries! - This is a super simple but useful tip I stumbled upon this week. The problem with using dict[key] is that, if the key doesn’t exist in the dictionary, you’ll raise a KeyError and crash your script (BIG problem if you’re trying to run a script overnight, unattended!). Instead, use dict.get(key, defaultValue): if the key doesn’t exist, your code will return defaultValue instead. Pretty nifty, right?

• ⏩ 20 SQL query optimisation techniques - Whether you’ve been writing SQL for 20 years or 20 seconds, I can guarantee that you’ll find something useful in Dr Milan Milanović’s helpful list. I particularly liked the tip on using an index, and I recommend the fantastically-named “Use the index, Luke” website to understand how this works.

Career Corner

• 📝 3 Key Career Decisions for Junior Data Scientists - All Data Scientists need to answer three questions: (1) do I want to be an IC or a manager? (2) what kind of Data Scientist do I want to be? and (3) what am I made to do? In this article I’ll give you my take on these three questions and explain how I’m making sure I don’t fall into the Default Path Trap.

One more thing…

Could you help me out by filling in the poll below? We’ve had a lot of new sign-ups recently and this will help me make sure I’m pitching the content correctly.