One of Google’s latest experimental models, Gemini-Exp-1206, shows the potential to alleviate one of the most grueling aspects of any analyst’s job: getting their data and visualizations to sync up perfectly and provide a compelling narrative without having to work all night. Investment analysts, junior bankers, and members of consulting teams aspiring for partnership positions take on their roles, fully aware that long hours, weekends, and pulling the occasional all-nighter might give them an inside edge on a promotion. What burns so much of their time is getting advanced data analysis done while simultaneously creating visualizations that reinforce a compelling storyline.
Making this more challenging is that every banking, fintech, and consulting firm, like J.P. Morgan, McKinsey, and PwC, has unique formats and conventions for data analysis and visualization. VentureBeat interviewed members of internal project teams whose employers had hired these firms and assigned them to the project. Employees working on consultant-led teams said producing visuals that condense and consolidate the massive amount of data is a persistent challenge. One noted it was common for consultant teams to work overnight and do a minimum of three to four iterations of a presentation’s visualizations before settling on one and getting it ready for board-level updates.
Gemini-Exp-1206 has now emerged as a potential game-changer that could streamline this cumbersome process significantly. Launching the model earlier in December, Google’s Patrick Kane outlined its capabilities. “Whether you’re tackling complex coding challenges, solving mathematical problems for school or personal projects, or providing detailed, multistep instructions to craft a tailored business plan, Gemini-Exp-1206 will help you navigate complex tasks with greater ease,” Kane explained. Google noted the model’s enhanced performance in more complex tasks, including math reasoning, coding, and adherence to a series of instructions.
Testing the Potential
VentureBeat took Google’s Exp-1206 model for a thorough test drive this week by creating and testing over 50 Python scripts in an attempt to automate and integrate analysis alongside intuitive and easily understood visualizations that could simplify the complex data being analyzed. With hyperscalers dominating today’s news cycles, our specific goal was to create an analysis of the technology market while also generating supporting tables and advanced graphics.
Our findings revealed various aspects after thorough evaluation using over 50 different iterations of verified Python scripts. The greater the complexity of a Python code request, the more the model “thinks” and tries to anticipate the desired result. Exp-1206 attempts to intelligently predict what’s needed from a complex prompt and adjusts its output based on the nuance of the request. This behavior was evident as the model alternated between formats of table types directly above the spider graph of the hyperscaler market analysis created for the test.
Forcing the model to handle complex data analysis and visualization and to produce an Excel file resulted in a multi-tabbed spreadsheet. Without explicitly requesting an Excel spreadsheet with multiple tabs, Exp-1206 created one. The primary tabular analysis was in one tab, visualizations on another, and an ancillary table on the third. Moreover, telling the model to iterate on the data and recommend the top 10 visualizations it deems best suited for the data delivered beneficial, insightful results.
Pushing the Model Further
Our aim was to see how far the model could be pushed in terms of complexity and handling layered tasks. The performance of the model in creating, running, editing, and fine-tuning 50 different Python scripts highlighted how rapidly it adapts to prompt history and picks up on code nuances. The model displays significant flexibility based on previous inputs, continually refining its process.
Results from running Python code created with Exp-1206 in Google Colab demonstrated that even nuanced details, such as shading and translucency of layers in an eight-point spider graph, were effectively managed. The spider graph was designed to show how six hyperscaler competitors compare, where the eight attributes asked for alignment across all hyperscalers and anchored the graph consistently, even though graphical representations varied.
VentureBeat selected the following hyperscalers for a comparative analysis 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 Centers, Oracle Cloud, and Tencent Cloud. By writing an 11-step prompt consisting of over 450 words, our goal was to see how well Exp-1206 could handle sequential logic without losing its place in a complex multistep process.
Creating Comparative Tables and Graphs
We submitted the prompt to Google AI Studio and selected the Gemini Experimental 1206 model. The code was then copied into Google Colab, saved into a Jupyter notebook (Hyperscaler Comparison – Gemini Experimental 1206.ipynb), and run. The Python script executed flawlessly, creating three files denoted with red arrows in the upper left corner.
The first series of instructions asked Exp-1206 to create a Python script comparing 12 different hyperscalers by their product name, unique features, differentiators, and data center locations. The result was an Excel file that included a formatted spreadsheet to fit the required columns and took less than a minute to generate. The next set of commands asked for a table comparing the top six hyperscalers, which was represented in HTML format by the model.
The final set of commands focused on creating a spider graph to compare the top six hyperscalers. Exp-1206 was tasked with selecting eight criteria for the comparison and completing the plot. These commands were converted into Python, and the model created the required file, which was then provided in the Google Colab session.
Streamlining Analysts’ Work
Google’s latest experimental model, Gemini-Exp-1206, aims to ease one of the toughest aspects of an analyst’s job: aligning data and visualizations to create a strong narrative without working overtime. Investment analysts, junior bankers, and consultants striving for promotions understand that their roles demand long hours, weekends, and sometimes all-nighters. These grueling hours are often spent conducting advanced data analysis while also generating visualizations that effectively tell the story.
Adding to this challenge are the varying data analysis and visualization standards upheld by each banking, fintech, and consulting firm, such as J.P. Morgan, McKinsey, and PwC. VentureBeat spoke with internal project team members who reported that producing visuals that synthesize vast amounts of data is a constant obstacle. One employee mentioned that it’s typical for consultant teams to work through the night and perform at least three to four iterations of visualizations before finalizing a presentation for board-level updates.
Enter Gemini-Exp-1206, which promises to revolutionize this laborious process. Launched in December, Google’s Patrick Kane highlighted the model’s capabilities: “Whether you’re addressing complex coding challenges, solving math problems for school or personal projects, or giving detailed, multistep instructions to create a custom business plan, Gemini-Exp-1206 will help you tackle intricate tasks more easily,” Kane said. Google emphasized the model’s improved performance in handling complex tasks such as mathematical reasoning, coding, and following a sequence of instructions.
