Google Colab, the cloud-based notebook application from Google for coding, data analysis, and artificial intelligence, is introducing a new feature called the Data Science Agent. This tool is designed to assist users in efficiently cleaning data, visualizing trends, and deriving insights from their uploaded datasets.
Initially revealed at Google’s I/O developer conference early last year, Data Science Agent was first launched as a standalone tool. However, Google opted to incorporate it into Colab, enabling users to easily access the agent directly from within their Colab notebooks, as explained by Kathy Korevec, Google’s director of product at Google Labs, in a recent interview.
Starting this week, Data Science Agent is available for free within Colab; though, keep in mind that Colab sets limitations on the computational power available to free users. Google offers various paid Colab subscriptions with enhanced features beginning at $9.99.
While the Data Science Agent primarily targets data scientists and applications in AI, it also facilitates identifying API anomalies, analyzing customer information, and generating SQL code. Users simply need to upload their data and pose a question to the agent.
Powered by Google’s Gemini 2.0 AI model family, the Data Science Agent employs “reasoning” tools to assist with tasks such as feature engineering and data cleaning. In an interview with TechCrunch, Korevec noted that Google is continually enhancing the agent’s capabilities through methods like reinforcement learning and incorporating user feedback to improve its performance.
Currently, the Data Science Agent supports analysis of CSV, JSON, or .txt files that are under 1GB in size, handling approximately 120,000 tokens in a single prompt, equating to around 480,000 words.
Korevec mentioned that the Data Science Agent could be integrated into more developer-focused Google applications and services in the future.
“We’re just beginning to explore the possibilities here,” she remarked. “Since it functions as an agent, we can incorporate it into various tools, and I don’t want to restrict access to those who may be hesitant to delve into the code in Colab.”
Compiled by Techarena.au.
Fanpage: TechArena.au
Watch more about AI – Artificial Intelligence
