Google is enhancing its Gemini platform by introducing a file upload feature for its custom AI chatbots called “Gems”. Available to users with a Gemini Advanced subscription under Google Workspace, this feature allows the AI to refer to uploaded files, enabling responses that are more contextually informed and aligned with specific user needs.
The new capability sets Gemini apart in the customizable AI landscape, where it competes with OpenAI’s Custom GPTs, by offering real-time integration with Google Workspace and dynamic updates.
This development comes as Google prepares its new AI browser assistant, Jarvis, set for limited release in December 2024.
Gemini AI’s New Customization Feature Explained
Gems allow users to personalize their interactions with AI by tailoring the chatbot’s behavior and expertise to specific needs, such as business tasks, personal projects, or creative endeavors. The new ability to upload files into Gems represents a major step in refining how users interact with AI tools.
By adding up to 10 documents, such as PDFs, Google Docs, or CSVs, users can create chatbots that generate responses based on precise reference material. This helps businesses automate tasks like drafting content or handling HR queries, improving efficiency and reducing manual input.
- Document and text files, such as TXT, DOC, DOCX, PDF, RTF, DOT, DOTX, HWP, HWPX and Google Docs
- Data Files, such as XLS, XLSX, CSV, TSV and Google Sheets
Updates made to linked Google Docs or Google Sheets will be automatically recognized by Gems, while other document types need to be re-uploaded to reflect any changes.
The addition of file uploads in Gems broadens the scope of AI utility in workplace and educational settings. For instance, companies can ensure that digital content adheres to corporate guidelines by embedding style manuals, while educational institutions might use course materials to generate lesson plans that align with specific curricula. This added capability reduces repetitive prompting and enables smoother user interactions.
Pre-Made Gems and Mobile Advancements
Google is expanding beyond customizable Gems with the release of six pre-made versions tailored for professional use. These include the Marketing Insights Gem, which assists in understanding customer behavior, and the Sentiment Analyzer that helps interpret customer feedback over time.
Google Gemini Gems vs. OpenAI Custom GPTs
Google’s Gemini Gems and OpenAI’s Custom GPTs both cater to users looking for customizable AI solutions, but they come with notable distinctions. One standout feature of Gemini Gems is their integration with Google Workspace, which allows seamless real-time access to files stored in Google Drive, Docs, and Gmail.
This eliminates the need for manual document uploads—something that OpenAI’s Custom GPTs still require. This integration gives Gems an edge for users deeply embedded in Google’s ecosystem.
Gems are designed to handle more demanding tasks, thanks to their higher token limit of 15,000 tokens, in contrast to Custom GPTs, which are typically limited to 4,096 tokens. This makes Gemini Gems more adept at managing complex, multi-layered workflows.
However, Custom GPTs offer advantages in terms of flexibility when it comes to sharing. Users can share Custom GPTs publicly via URLs or feature them in the GPT store, a level of accessibility that Gemini Gems do not provide; Gems remain private and cannot be shared publicly or within teams.
Another point of comparison lies in how external documents are handled. Custom GPTs allow the integration of knowledge base files directly into their system, making them versatile for referencing outside documents. On the other hand, while Gemini Gems integrate dynamically with Google’s services, they lack the capability to embed external knowledge bases as effectively.
Jarvis AI: Google’s Next AI Tool for Chrome
While Gemini continues to evolve, Google is also preparing to introduce Jarvis, an AI assistant built for Chrome. Details of this project came to light following an accidental listing on the Chrome Web Store, revealing a prototype intended for internal use. Jarvis will handle tasks like form-filling and booking services directly in the browser, using the Gemini 2.0 model to interpret web page elements and execute user commands seamlessly.
Jarvis is designed to focus specifically on browser tasks, distinguishing it from tools like Anthropic’s Claude, which can interact across multiple desktop applications. This AI agent aims to simplify online workflows, appealing to users who spend extensive time in web-based environments. Google’s trial launch for Jarvis is planned for December 2024.
Competing AI Agents and Industry Landscape
Google’s recent moves with Gemini and Jarvis come amidst growing competition from other AI developers. Anthropic has expanded its Claude model to include Computer Use, enabling tasks beyond browser-based functions. Available through platforms like Amazon Bedrock and Vertex AI, Claude is marketed as a more flexible solution for businesses needing robust desktop automation.
Meanwhile, Microsoft has bolstered its position with Copilot agents designed for use in enterprise software such as Dynamics 365. These agents automate tasks specific to business processes, from lead categorization to financial data reconciliation, enhancing productivity in professional settings.
Data Privacy and Security Considerations
As AI tools evolve, the handling of user data remains a primary concern. For The Gemini Gems file upload feature, Google states that any documents uploaded are managed securely and kept separate from training datasets. Companies also have the option to delete uploaded data at their discretion.
This approach aims to balance functionality with privacy, particularly in light of past issues faced by rivals, such as Microsoft’s Copilot+ capturing screenshots without user consent, leading to delays in its release.
Affordable AI Solutions for Developers
Alongside these upgrades, Google recently launched Gemini 1.5 Flash-8B, a cost-efficient AI model designed for high-volume use. This model supports up to 4,000 requests per minute, allowing developers to handle tasks from customer service interactions to large-scale translations. By providing a budget-friendly option, Google is making its AI tools more accessible for developers seeking reliable performance at a lower cost.
AI Integration Across Google’s Ecosystem
Gemini’s reach extends beyond standalone applications, embedding into tools like Google Maps and Gmail. Users benefit from advanced features such as lane guidance and real-time weather alerts powered by Gemini AI, showcasing Google’s ongoing commitment to integrating AI across its platforms.
