is there a plan to incorporate ai (like chatgpt) to ask questions and allow for natural language creation?
examples of questions:
how do i convert 10 rows into one row and 10 columns?
how do i take a sample of the first three rows of each subcategory?
which transforms do i use to clean email addresses and then find out how many times each email address appears?
an example of how i would see a natural language command used:
paste from clipboard. only use the first 10 rows to create a list sorted first by date and then by distance to omaha, nebraska.
how does one go about creating a âgoodâ set of transforms on the canvas? how do i measure efficiency?
i see the % memory consumption as an indicator. but i donât know how to use that to drive a decision. do i use less nodes? do i break the dataset into fewer datasets and process them serially?
how do i refactor what i have? i donât know if the 50 steps on my canvas could have been done with just 15 steps by a more experienced user.
what do i look for or practice to get better at using edt?
or is there a plan to have a persistent ai bot (a la microsoftâs clippy) that will constantly evaluate and then suggest to me that i could do this another/better way? or maybe a button i could click that would activate the âbrice evaluatorâ and then evaluates my canvas and makes suggestions. e.g. âyou can do this section of 5 transforms with this one transform and these parametersâŚâ
There are loads of things we want to do with EDT, including: more input and outputs (including databases and web APIs), basic visualization, better use of Javascript, scheduler, usability and performance improvements etc. Currently a natural language doesnât figure high in that list.
Natural language is highly ambiguous and far from an ideal medium for describing data transform problems. Expect to see endless hype about it from others though.
Also, with some experience, you will hopfully be able to âconvert 10 rows into one row and 10 columnsâ in a similar time to typing the phrase in natural language.
Finally we would probably have to change to a monthly subscription to cover the cost of using ChatGPT.
As @olaf says it is really a matter of reading the documentation/forum and practice.
Also it doesnât matter that much if you take a few more steps to get the data you need than the theoretical minimum. As long as you end up with the data you need in a reasonable time.
Everyone hated âclippyâ didnât they? So no!
That would probably be harder and more consuming to program than creating Easy Data Transform in the first place - and that has been hard enough!