Productivity booster or job killer?
People adept with this tool are in high demand and often well-compensated.
Microsoft hasdevelopeda framework called SpreadsheetLLM that uses large language models for analyzing and interpreting spreadsheet data.
It solves this by serializing data andincorporatingcell addresses, values, and formats into a data stream.
The tool includes a component that compresses spreadsheets for LLMs.
SpreadsheetLLM does have some limitations in its current form.
For example, it ignores cell background colors, which may convey some meaning to the sheet.
It also lacks semantic-based compression for cells containing natural language.
Weaknesses aside, it comes off as highly effective.
In tests, it outperformed traditional approaches by 25.6 percent in GPT-4’s in-context learning setting.
Additionally, SheetCompressor reduces token usage for spreadsheet encoding by 96 percent, significantly decreasing computational costs.
SpreadsheetLLM is also very good at spreadsheet table detection, which is fundamental to spreadsheet understanding.
This tool could streamline data processing in several industries.
It is impossible to estimate the job impact since the AI/LLM industry is still burgeoning.
Positions in many sectors rely heavily on spreadsheets, particularly inMicrosoft Excel.
These roles are highly valued, with commensurate compensation.
A study of nearly 27 million job listingsfoundthat Microsoft Excel skills were the most sought-after software skill.
Perhaps more intriguing is the model’s ability to work with both structured and unstructured spreadsheet data.
However, SpreadsheetLLM is not ready for a public launch.
It’s still in the research phase and too raw to be incorporated into commercial products like Microsoft Excel.
