Testing the Google Gemini-Exp-1206 model in data analysis, visualizations

One of Google’s latest experimental models, Gemini-Exp-1206, shows the potential to alleviate one of the most exhausting aspects of every analyst’s work: perfectly synchronizing data and visualizations and providing a compelling narrative, without having to work all night long.

Investment analysts, junior bankers and advisory board members aspiring to partner positions enter their roles knowing this long hoursweekends and the occasional all-night party can give them a head start on advancement.

What takes so much time is performing advanced data analysis while creating visualizations that reinforce: captivating plot. This challenge is the fact that each banking, fintech and consulting firm, such as JP Morgan, McKinsey and PwC, has unique formats and conventions for data analysis and visualization.

VentureBeat interviewed members of internal project teams whose employers hired these companies and assigned them to the project. Employees working in consultant-led teams have found it a constant challenge to create visualizations that condense and consolidate enormous amounts of data. One of them stated that it is common for teams of consultants to work overnight and do at least three to four iterations of presentation visualizations before settling on one and getting it ready for a board-level update.

A compelling apply case for test driving the latest Google model

The processes analysts rely on to create story-supporting presentations with stalwart visualizations and graphics require so many manual steps and repetitions that they have proven to be a compelling apply case for testing Google’s latest model.

This week, VentureBeat took the Google Exp-1206 for a thorough test drive. We created and tested over 50 Python scripts in an attempt to automate and integrate analysis and intuitive, easy-to-understand visualizations that could simplify the sophisticated data being analyzed. Given hyperscalers dominating news cycles today, our specific goal was to create an analysis of a given technology market while creating supporting tables and advanced graphics.

Our findings from over 50 different iterations of validated Python scripts included:

• The greater the complexity of the code request in Python, the more the model “thinks” and tries to predict the desired result. Exp-1206 tries to predict what is needed based on a given sophisticated prompt and will vary what it generates, even by changing the subtleties of the prompt. We saw this in how the model changed between the table type formats placed directly above the hyperscaler market analysis spider chart we created for the test.

• Forcing the model to undertake sophisticated data analysis and visualization and create an Excel file creates a spreadsheet with multiple tabs. Without asking for an Excel spreadsheet with multiple tabs, Exp-1206 created it. The primary tabular analysis requested was on one tab, visuals on the second, and the supporting table on the third.

• Telling the model to iterate on the data and recommend the 10 visualizations it thinks best fit the data produces beneficial and insightful results. To reduce the time needed to create three or four iterations of slides before reviewing the board, we forced the model to create multiple iterations of conceptual images. They can be easily cleaned and integrated into your presentation, saving you hours of manual work creating diagrams on your slides.

Pushing Exp-1206 towards sophisticated, multi-layered tasks

VentureBeat’s goal was to see how far the model could be taken in terms of complexity and layered tasks. The performance of creating, running, editing, and tuning 50 different Python scripts demonstrated how quickly the model tries to pick up on nuances in the code and respond immediately. The model flexes and adapts based on rapid history.

The result of running Python code created with Exp-1206 in Google Co showed that varying granularity extended to layer shading and transparency in an eight-point spider plot that was intended to show a comparison of six hyperscale competitors. The eight attributes that we asked Exp-1206 to identify across all hyperscalers and anchor the spider graph remained consistent while the graphical representations varied.

Battle of the hyperscalers

We chose the following hyperscalers for comparison in our test: Alibaba Cloud, Amazon Web Services (AWS), Digital Realty, Equinix, Google Cloud Platform (GCP), Huawei, IBM Cloud, Meta Platforms (Facebook), Microsoft Azure, NTT Global Data Centra, Oracle Cloud and Tencent Cloud.

We then wrote an 11-step prompt with over 450 words. The goal was to test how well Exp-1206 handles sequential logic and doesn’t lose its place in a sophisticated, multi-step process. (The prompt can be read in the appendix at the end of this article.)

We then sent a reminder Google Artificial Intelligence Studioselecting the Gemini Experimental 1206 model as shown in the figure below.

We then copied the code into Google Colab and saved it in a Jupyter notebook (Hyperscaler Comparison – Gemini Experimental 1206.ipynb) and then ran the Python script. The script ran flawlessly and created three files (denoted by red arrows in the upper left corner).

Hyperscaler benchmarking and graphics in less than a minute

A model created specifically to save analysts’ time

VentureBeat has learned that in their daily work, analysts continue to create, share and fine-tune hint libraries for specific AI models to improve reporting, analysis and visualization across their teams.

Attachment:

Experimental test of Google Gemini 1206 hints

Write a Python script to analyze the following hyperscales who have announced global infrastructure and data center presence for their platforms, and create a table comparing them that captures the significant differences in each global infrastructure and data center presence approach.

Let the first column of the table be the name of the company, the second column be the name of each of the company’s hyperscalers that have a global infrastructure and data center presence, the third column be what makes their hyperscalers unique and a deep dive into the most diverse features, and the fourth column is data center locations for each hyperscaler at the city, state and country levels. Include all 12 hyperscalers in the Excel file. Don’t browse the web. Create an Excel file with the result and format the text in the Excel file to be free of brackets ({}), quotation marks (‘), double asterisks (**), and any HTML code to improve readability. Name the Excel file Gemini_Experimental_1206_test.xlsx.

Then create a table that is three columns wide and seven columns deep. The first column is titled Hyperscaler, the second is titled Unique Features and Differentiators, and the third is titled Infrastructure and Data Center Locations. Bold and center the column titles. Also bold the hyperscaler titles. Double check that the text in each cell of this table wraps and does not flow to the next cell. Adjust the height of each row to ensure that all text will fit in the intended cell. This table compares Amazon Web Services (AWS), Google Cloud Platform (GCP), IBM Cloud, Meta Platforms (Facebook), Microsoft Azure, and Oracle Cloud. Center the table at the top of the results page.

Then let’s take Amazon Web Services (AWS), Google Cloud Platform (GCP), IBM Cloud, Meta Platforms (Facebook), Microsoft Azure, and Oracle Cloud and define the eight most differentiating aspects of this group. Apply these eight differentiating aspects to create a spider chart comparing these six hyperscalers. Create a single enormous spider chart that clearly shows the differences between these six hyperscalers, using different colors to improve its readability and the ability to see the outlines or traces of the different hyperscales. Be sure to title your analysis “What differentiates hyperscalers the most”, December 2024. Make sure the legend is fully apparent and not at the top of the graphic.

Add a spider graphic to the bottom of the page. Center the spider graphic below the table on the output page.

Here are the hyperscalers to include in your Python script: Alibaba Cloud, Amazon Web Services (AWS), Digital Realty, Equinix, Google Cloud Platform (GCP), Huawei, IBM Cloud, Meta Platforms (Facebook), Microsoft Azure, NTT Global Data Centra , Oracle Cloud, Tencent Cloud.

Daily insight into business apply cases with VB Daily

If you want to impress your boss, VB Daily will let you do just that. We give you insight into what companies are doing with generative AI, from regulatory developments to practical implementations, so you can share your insights for maximum return on your investment.

Read our Privacy Policy

Thank you for subscribing. Find more VB newsletters here.

An error occurred.