🍊 This Week's Juice: Using AI for Data Analysis, a Smartphone AI Desktop Robot, and 3 AI Detectors!

Sponsored by

Hey friends!

AI is everywhere now.

It’s in various industries, affecting the way we live and work.

In today’s issue, I want to widen your mind.

And show you how AI is being used in multiple industries.

Here’s this week’s AI juice:

Let’s get this started!

- Justin 🍊

Read time: 5 minutes

1. AI for Data Analysis: Julius AI 🧐

As a data analyst, I love playing around with data.

Especially my OWN data.

I recently extracted my Apple Watch fitness data for a portfolio project in Python.

(P.S. I’m looking for a role as a data analyst now; please let me know if you are hiring and are interested in my work!)

But, I wanted to see if Julius AI, a proclaimed “AI Data Analyst,” could beat me in their data analysis skills.

Essentially, I’m making it a human vs AI contest.

How to Analyse Data with Julius AI

I created an account with Julius AI and uploaded my data.

That was not bad at all; I got a sneak peek of my data within seconds. (My file wasn’t too big)

Then I gave it this prompt:

It immediately got to work in Python.

a. Pre-Processing Steps

It determined the data types:

And proceeded to offer some next steps:

b. Data Cleaning

It then attempted to perform cleaning actions, but it hit an error when converting text like “1h 18” into a float.

It couldn’t understand that 1h 18 could be converted into 78 minutes using a string split and converting hours into minutes and adding them up.

c. Summary Statistics

In any case, with the flawed cleaning of data, it proceeded to drop a summarized table of statistics.

While this is great for a glance into the data, we need some graphs!

d. Basic EDA with Charts

Subsequently, it ran multiple data visualizations using matplotlib and seaborn libraries in Python.

The first chart was a histogram of my Active Calories.

With this, I was able to observe a right-skewed histogram, which showed that I was burning approximately 380kcal of Active Calories a day.

Next, I got a weird chart showing my heart rate distribution.

The data wasn’t cleaned properly, so the graph is whack.

Then, it gave me a histogram of my Exercise Minutes, peaking at around 10 minutes.

This made sense.

To push it for further analysis, I asked it for a time series analysis of my Heart Rate:

I found it weird that it imported libraries and the data again. This meant that each response was instanced.

It hit an error when trying to plot the data because the data type was a string.

It was nice that they gave a proposed solution and even AI-generated responses to click on to continue the conversation.

I clicked on one of the responses.

I told it to convert the ranges into a single average value and plot a time series graph.

But it failed to create a proper graph again.

Let’s stop here. 😆 

My Thoughts

I may be biased, but my human-coded portfolio project showed more promise than this…

I managed to get a proper analysis with graphs and even predictions.

This experiment proves that the current state of AI data analysts is just not competent enough to handle slightly complex data.

In the real world, data is MUCH more dirty than this.

This goes to show that data analysts won’t get replaced by AI.

But don’t get me wrong,

Julius AI is a fantastic tool for crazily quick data EDA.

But it needs to be well cleaned up before any analysis can be done.

Also, I was using the free plan, but you may get better results from their paid plans that offer the latest models (GPT-4, Claude).

2. Smartphone AI Desktop Robot: LOOI 🤖

I’m always a sucker for cool gadgets.

Recently, I came across this cute little robot.

It’s called LOOI.

LOOI is an AI-powered desktop robot that uses your smartphone as a screen.

And it’s cheaper than the Rabbit R1 or Humane AI Pin.

Based on reviews, it’s really responsive because it is powered by ChatGPT.

It has a great build.

What’s neat is this:

The seamless transition between taking your phone off and putting it back on.

It works for MagSafe-compatible phones only, like the latest iPhones.

Furthermore, some reviewers say it doesn’t hear prompts as well and functions are limited.

3. 3 AI Content Detectors ✍🏻

In this rampant world of AI use, it’s hard to find original and trustworthy content.

Especially when Large Language Models (LLMs) hallucinate.

They lie super confidently.

Using these “lie detectors” is good practice when reading content online.

Here are 3 AI content detectors you need to try:

✍🏻 1/ Copyleaks

Copyleaks is a comprehensive AI content detection suite with over 99% accuracy.

It provides features like plagiarism detection, AI-generated source code identification, and a writing assistant, all backed by military-grade security and trusted by leading organizations.

It’s free, and enhanced models require you to sign up.

✍🏻 2/ Writer

Writer is a quick and easy AI content detector that detects AI content within seconds.

It’s free and requires no subscription!

✍🏻 3/ GPTZero

GPTZero is an advanced AI detection tool designed to identify text generated by AI models like ChatGPT, GPT4, and others, tailored especially for educational settings.

It features a comprehensive dashboard for deep text analysis, integration with Google Docs, and an API for organizational use, enhancing transparency and trust in digital content.

You can use it for free!

That's all, folks - see you next Wednesday!

If you have any questions or feedback, reply to this email. (I'll personally reply! :)L