GPT is all the rage in the machine 🤖
(Image: Evening Standard) How do you think ChatGPT got so smart?
Asking questions to GPT is a great way to both connect disparate data sources, get help with coding, and remix the words and pixels of the world in a novel way.
Though there are quite a few technical issues and ethical problems with this technology.
Sometimes an elegant illustration is the easiest way to share your thoughts and feelings with the world.
Can your machine data artist already do this effectively?
Machine learning
When "AI magic" doesn't cut it, data analysis in code is done in languages like R: here showing the RStudio interface with an open Data Package, graphing tools, and a console for SPARQL queries. If you take an introductory class in Data Science, this is probably what you will learn.
Jupyter notebooks (pictured above) as well as Observable are used with many languages (R, Python, Julia, JavaScript) by professional data scientists to develop models, make small visualizations, and share their results and methods online.
Typically data is used not in code but in dashboards, such as ORNL Vista shown above. Compact, dense information displays are useful for trained professionals to make real-time course corrections. For most people, they are overwhelming (even if kind of nice to look at).
Image from optimizing weights (TowardsDataScience)
To understand data science, you need to understand that statistics is a process, and there is no "right answer" - just the probability of an answer being correct. Much of data science involves creating models, and improving them with better data or algorithms. And it the big question, is whether your algorithm is getting truly smarter, or just leading you further into the rabbit hole of hallucination and bias.
Image from machine learning explained (MIT)
It is important to understand that while Machine Learning is a very powerful tool, it is not appropriate everywhere and all the time.
Right now, we are applying it in many places where we could save time and money with much simpler or more appropriate solutions. With GPT the lines get even more blurry, and inaccessible, unreliable pseudo-analytics will be everywhere. Stand on the side of the light!