Data Acquisition and GenAI

How do we get or acquire data, and how could generative

AI improve this process? We sometimes talk about

data that can be captured, like some wild animal. Sometimes we talk about it being collected like a

basket of fruit, but also extracted like oil or even generated

like electricity. We can even talk

about acquiring data like some suspicious sum of

money that landed on our lap. What's important to

remember here is the concept of

recording and encoding. You're transforming

things around you into digital data by recording the information in some way and encoding it in a computer

processable format. Remembering the fact

that data isn't just passively obtained in some

pristine form should serve as a reminder that you

need to think critically about where the data comes from and what it's

supposed to represent. Data doesn't spontaneously

appear to us. We humans are involved in the

process. That's undeniable. But the way we are involved and the extent to which we

are involved is variable. It can be as direct

as counting items, like the number of empty

cups on your desk to using sophisticated tools that have automated processes

of recording data, like a heart rate monitor, or even a Geiger counter, which is an

electronic instrument used for measuring

ionizing radiation. Even things is automated as a Geiger counter or

heart rate monitor have embedded in them decisions that people made about

what to measure. Getting curious about

what it is that's being recorded and what the

potential limitations might be of the

representation that you're capturing is

something that's going to really

improve your ability to generate valuable

insights with data analysis. You might have an idea of a

question you want to ask, and GenAI can help you go

from question to data. Maybe you're interested

in why some areas of the same city are hotter than

others during the summer. You could use a text-to-text

generative AI tool like ChatGPT or Claude to

brainstorm your question, using a prompt like, I'm

interested in understanding why some areas of the same city are hotter than others

during the summer. What kind of data do I need

to answer my question? You take this as a

starting point and then prompt further

for specifics. How can I take

temperature measurements? You dig in critically to understand what

your options are. What are the alternatives? What are the advantages and disadvantages of different

methods of data collection? Depending on the

tool that you use, it may be trained on

data that can give you an answer on where to

find existing datasets. Or if it's integrated

with a web browser, it can get you that information

with a prompt like, what are some publicly

available sources of data that could help

me answer my question? You may want to play around with the terms data repositories

and data bases as well. Both are terms for

collections of data, where a data base

tends to specify usually something more

structured and specific, like an inventory system

used in a warehouse that uses a very particular

consistent format. While a data repository

can contain a wide array of structured and

unstructured collections of different types of data. For example, researchers often save things on

websites like Zanoto, where you can host images, spreadsheets, all kinds of

modalities and formats. Speaking of modalities

and formats, let's break down

those terms next.