Suggestions
Trevor Fox
Move fast and build things
Trevor Fox is a driven individual who is passionate about contributing positively to the world through collaboration and innovation. With a love for learning, Trevor holds a degree from the University of California, Davis, and is continuously looking for new ways to improve himself and those around him.
His extensive knowledge in growth and web strategy has lent itself to various organizations, including Growth at Census and his consulting company, trevorfox.com. As a strong advocate for ethical business practices and sustainability, Trevor envisions a global future where experimentation and creative solutions intersect to create a better world.
Through his personal blog, trevorfox.com, as well as his work mentoring others on growthmentor.com, Trevor is committed to sharing his learnings and helping others succeed. He seeks to connect with like-minded individuals who share his values of creativity, innovation, and learning. With Trevor, you are guaranteed a partner who will work diligently to achieve success in whatever endeavour you pursue.
Highlights
Does Google's Gemini read column descriptions in BigQuery when writing SQL queries? I did the hard work to find out, so you don't have to.
You might be saying, who cares?
First, don't judge me. Second, this is important because it signals where the future of data analysis is heading.
Consider this: Google paid $2.6B to acquire Looker, not so it could trash the front end but so it could cherish the back end (semantic layer). The LookML semantic layer provides the context about the structure, join patterns, and metadata, governance restrictions that BigQuery needs to magically write SQL and get it pretty darn close to right.
That's where we're going, so you better learn to at least read SQL and determine if it's headed off the rails or not!
#BigQuery #Looker #AI #LLM #Analytics #SQL #GoogleGemini
Yesterday, Benn Stancil wrote yet another thought-provoking perspective on the application of LLMs. He posed the question, what would you do if "a million Yale-educated lawyers showed up at your corporate headquarters offering you their services for $5.92 an hour?" (If the analogy isn't clear, check out his post in the comments.)
I compared this idea to the AIrtime Talks conversation with Jai Toor a couple of weeks ago. Jai suggested an approach that was not concrete to me at the time: reduce the scope of a problem so that you can specifically design an LLM capability to solve it.
In other words, these lawyers don't have the ego you would expect a Yale-educated lawyer to have. They can also train to be great coffee smell detectors, mug washers, order recommenders, or latte cream designers (if you can design the right training).
Designing with generative AI offers several interesting problems, but deciding how to divide, specialize, and coordinate these tasks is a critical one for building systems of any novel value.
