Ilya Shabanov Profile picture
Dec 19, 2022 β€’ 7 tweets β€’ 11 min read β€’ Read on X
Struggling with R or Python?

Leverage @OpenAI and #ChatGPT to learn and solve your data analysis challenges.

I am just blown away 🀯 ! Here are a few use cases of what you can do.

#AcademicTwitter #sciencetwitter @OpenAcademics @AcademicChatter @PhDVoice @PostdocVoice

⬇️
@OpenAI @OpenAcademics @AcademicChatter @PhDVoice @PostdocVoice First I asked it a straight forward question to load a CSV file with "width, length, depth and weight in R": Image
@OpenAI @OpenAcademics @AcademicChatter @PhDVoice @PostdocVoice Ask for Ideas on what to do with the data.

ex "How could I visualise the four dimensional data?" Image
@OpenAI @OpenAcademics @AcademicChatter @PhDVoice @PostdocVoice Ask to show and EXPLAIN a special technique that you would like to apply to your data

I asked about "how to do a PCA analysis and what it means". Image
@OpenAI @OpenAcademics @AcademicChatter @PhDVoice @PostdocVoice Copy paste code, that you do not understand.

That one is just magic - i copy pasted some stuff from the internet. Compare screenshot and link. It's better than the person explaining it.

buff.ly/3FyhJsa

I attached the plot the code generates for comparison. ImageImage
@OpenAI @OpenAcademics @AcademicChatter @PhDVoice @PostdocVoice Ask why YOUR code is not working by copy pasting it.

Here i made a small example where I was using float probabilities as indices and it was crashing.

I is indeed the 3. reason and then I asked it to fix it for me. And it actually fixed MY code, not some generic code 🀯 . ImageImage
@OpenAI @OpenAcademics @AcademicChatter @PhDVoice @PostdocVoice I am completely blown away by how good it is. It 100% has the potential to revolutionise the way we work.

Even if you have no clue of programming - the AI can give you clues what you can or cannot do. Before that, the job was with the 1 person in the lab who knows R πŸ˜„

β€’ β€’ β€’

Missing some Tweet in this thread? You can try to force a refresh
γ€€

Keep Current with Ilya Shabanov

Ilya Shabanov Profile picture

Stay in touch and get notified when new unrolls are available from this author!

Read all threads

This Thread may be Removed Anytime!

PDF

Twitter may remove this content at anytime! Save it as PDF for later use!

Try unrolling a thread yourself!

how to unroll video
  1. Follow @ThreadReaderApp to mention us!

  2. From a Twitter thread mention us with a keyword "unroll"
@threadreaderapp unroll

Practice here first or read more on our help page!

More from @Artifexx

Apr 27
Plagiarism is a serious problem in academia.

Even a misplaced citation can massively damage your reputation.

Luckily AI is great at spotting plagiarism. Here is how:
πŸ‘‡ Image
1. Plagiarism is complex

If you look at the Harvard Guide it identifies 6 types of plagiarism.

"Uncited paraphrase" might be the most difficult to spot.

Here you might mention established findings from older papers without citing them. Image
2. Check for Plagiarism

After you have written your manuscript (yourself!). You might still end up with a few unintended instances of plagiarism.

This is where @teampaperpal comes in.

Make an account on paperpal . com and click "Plagiarism Check" in the sidebar. Image
Read 7 tweets
Apr 2
Don't have time to read a paper in detail?

Here is how to extract relevant information instantly:
πŸ‘‡ Image
The setup:

You have a rather large and relevant conceptual paper.

There are others and you are not sure if it's worth reading it.

You are however more interested in the "big picture" not the methods and details.
1. Upload the PDF to MyAIDrive(dot)com

Copy the link to the PDF.

MyAiDrive will analyze the PDF and make the results available in ChatGPT. Image
Read 8 tweets
Mar 26
Want to remember every paper you read?

Replace Zotero with a reference map and leverage your spatial memory.

Here is how:
πŸ‘‡

1. Reference manager vs map

Take a look at this screenshot: Which one looks more approachable and interesting? On the right is Zotero displaying all your papers in an endless list. On the left is a reference map.

Reference maps lay out your papers or PDFs in 2D on an endless surface called a "canvas" or "whiteboard". There are many tools that are capable of doing it: Obsidian is an obvious choice, Heptabase is great too, DrawIO is more complex but also good.

2. Zooming in and out / Finding things

Using the scroll wheel or pinch gesture you can fluidly navigate between the bird's eye view and the detailed view with your own notes on a single paper. Left: Detail, Right: All Papers.

To find papers you "fly up" and then "land on" the paper you are looking for. It feels incredibly natural and easier than scrolling through a list.

3. Why it works: Spatial Memory
Humans evolved moving around as hunter gatherers and spatial memory is a key trait needed for navigation. You leverage it by laying out your papers in a landscape, not a list. Your papers gain location and relation.

4. Headers and Topics: Location
Now that your papers are on a landscape or map you can have "countries" on this map. Every country is a topic, further subdivided in sub-topics. Here is the "Machine Learning Country" in the far south west of my map:

I can refer to "papers in the south west" - this is spatial memory being leveraged to remember where things are.

5. Semantic Connections: Relation

The next step is to build the "roads" between locations on the map. Simply draw an arrow and write on it what this relation signifies.

In the above example Swenson 2020 (top) wrote "the trait-demography relationship is weak (Yang 2018)".
So I read Yang 2018 (left) and added a connection. Later I found that (Lynn 2023) suggested a few solutions and linked those two as well.

By just looking at this map you can immediately write a sentence for your literature review. A narrative emerges and synthesis begins.

Summary:

Lay out papers on a spatial canvas using e.g. Obsidian instead of Zotero. Remembering them will be much easier because you can use your spatial memory. Synthesis starts happening automatically when you annotate connections between papers.

Do you do something like this?
Share a screenshot with us!Image
Image
Image
Image
This is one of the methods you can learn about in my upcoming webinar:

In this post I mentioned Heptabase and Obsidian as tools to achieve this results.

Here is a comparison between those two:

effortlessacademic.com/note-taking-to…
Read 4 tweets
Mar 14
6 Months ago my core tools barely included any AI.

Today things have shifted - a short update to this post:
1. Writing with 3 AI GPTs

When writing I consider 3 things:
- Finding citations to support the argument
- Putting ideas effectively on paper
- Skimming papers for logical relevance

ChatGPT can now do all of these with the help of GPTs:

effortlessacademic.com/a-complete-gui…
2. Semantic Search with SciSpace

Search for meaning not keywords. This is what AI enables us to do!

Check out this post:

Read 4 tweets
Mar 12
Webinar Apr 6th:
Literature Review & Academic Writing with AI

β†’ Find the most impactful literature quickly
β†’ Uncover reference gaps
β†’ Aid your writing process faithfully & ethically with AI

Link:

All details below:
πŸ‘‡effortlessacademic.com/elr3-webinar/Image
In this video I explain why the major changes in AI are such a big deal for literature reviews:
1. Semantic Literature Search

Semantic search allows you to obtain a comprehensive reading list on any literature review topic in under 1 minute.

This is brand new and released in January 2024.

We will look deeply at the caveats & solutions of this new method.
Read 11 tweets
Mar 5
How do you conduct a lit-review in 2024?

❌ Old way: Searching for papers by keywords on Google
βœ… New way: Semantic search using AI tools like SciSpace

Here is a demo:
πŸ‘‡ Image
1. AI Search

Internet search replaced tedious searches in libraries and journals.

Similarly, semantic search will replace the internet search.

AI allows to search answers to questions directly, rather than using keywords.

SciSpace is one such semantic search tool. Let's try!
2. @scispace_ on the GPT store

SciSpace is one of the top 10 GPTs (AI assistants) worldwide.

It harnesses the conversational abilities of ChatGPT and combines them with a database of 200 Million academic papers.

Start a conversation with this AI assistant:
Read 12 tweets

Did Thread Reader help you today?

Support us! We are indie developers!


This site is made by just two indie developers on a laptop doing marketing, support and development! Read more about the story.

Become a Premium Member ($3/month or $30/year) and get exclusive features!

Become Premium

Don't want to be a Premium member but still want to support us?

Make a small donation by buying us coffee ($5) or help with server cost ($10)

Donate via Paypal

Or Donate anonymously using crypto!

Ethereum

0xfe58350B80634f60Fa6Dc149a72b4DFbc17D341E copy

Bitcoin

3ATGMxNzCUFzxpMCHL5sWSt4DVtS8UqXpi copy

Thank you for your support!

Follow Us!

:(