Colleague, let me upgrade you
Introduce you to some new features
In the past two weeks, three leading Generative AI developers - Open AI, Microsoft, and Google- have announced significant model upgrades. This further reinforces that we are in a full-on GenAI race as technology rapidly expands. As librarians, understanding these advancements and their implications is crucial. Let's investigate how these updates might be practical to our work and discuss a glaring privacy concern.
ChatGPT 4o: The Omni Model
ChatGPT 4o—the ‘o’ stands for ‘omni’—is OpenAI’s latest rendition. It is currently accessible even to free account users for a limited number - about 10 to 15 - daily prompts if you want to try it. This model enables users to interact through voice, video, and text within the same model, much akin to a superpowered Alexa. Imagine the capabilities of a conversational AI that not only chats but also comprehends and processes visual and auditory inputs, like in the 2013 Spike Jonze movie Her. (Side note: it might be a bit too much like the movie Her, as Scarjo is NOT impressed with the likeness of the voice to her own.)
Key Enhancements:
Multimodal Interaction: You can engage with ChatGPT through voice, video, and text.
Speed and Efficiency: Faster than previous versions, providing near-instant responses. (If you need your answer in two seconds versus three.)
Advanced Image Generation: Enhanced capabilities for creating and modifying images, making it a valuable tool for presentations and visual projects.
Practical applications: My first takeaway was that I can immediately use what I need in image generation. The tool can convert large CVS files into PNG files—something Excel could do but faster and more visually appealing. I have always longed for a graphics team, and whenever Excel makes me an underwhelming chart, I might opt for ChatGPT’s data visualization for reports and presentations.
Google's Astra: The Future of AI Assistants
Google's latest, Astra, brings a new level of sophistication. It can handle voice, video, and text inputs in real time; Astra is designed to provide a more natural, conversational experience. Imagine asking Astra to identify objects in a room or understand the context of a discussion without explicit instructions—again, very reminiscent of Her or Space Odyssey.
Key Features:
Contextual Understanding: Astra can quickly identify and understand objects and their context, enhancing its ability to assist users in real time.
Natural Interaction: Designed to interact seamlessly, making conversations with Astra feel more human and less like interfacing with machines.
Practical applications: Last week, I mentioned some libraries' experiments with a neighbouring college - having them build a chatbot. The technology mentioned with these new models could significantly accelerate these bots' functionality and accuracy. Further, technology’s improvements and ability to generate voice prompts could lead us to verbal chatbots for directional information. This would be more accessible to some customers who may not find it easy to read or type due to vision limitations.
Microsoft's CoPilot: The Integrated Assistant
The new CoPilot is interesting because it integrates deeper into the Windows operating system. This upgrade enhances user experience by providing a more intuitive and helpful AI presence on personal computers. This enhancement aims to create a more intuitive user experience.
Key Features:
Windows Recall: This feature helps users resume tasks and navigate workflows by remembering past activities. It will get to know you better to help you. It functions like a photographic memory and is designed to remember everything done on your PC, which could be informative, albeit creepy.
Enhanced Integration: CoPilot’s deeper integration with Windows makes it a powerful tool for personal computers. This integration will help CoPilot manage emails in Outlook, create documents in Word, and even assist with complex data analysis in Excel, all while maintaining the context of what the user is working on.
Of course, a private virtual assistant has many practical applications. It could help us with document creation and management and provide helpful research assistance. Much like the OpenAI and Google models, the technology would also be useful for data analysis and visualization and for automating quick reference responses.
I am getting push notifications and ads on my work laptop for this new offering. The new AI-enhanced Windows PCs will be available from June 18, initially on premium models starting at $999 from manufacturers like Acer, ASUS, Dell, HP, Lenovo, Samsung, and Microsoft's Surface line.
Privacy Considerations
As exciting as these advancements are, they come with significant privacy concerns that we should consider.
Each model requires more access to user data to function effectively. This includes potentially sensitive information if used in a library or any setting. How often are we looking at something private we wouldn’t necessarily want to be saved? How many times would this put our user privacy policies in jeopardy? As GenAI models become more advanced, we should review privacy and device policies. If we download or use a GenAI on our computer, how much of our organization’s and customer’s data is collected in the background as we work?
While this concerns us as an organization and staff, it is also a concern for our customers using their personal or the library’s devices. While most AI models allow users to filter out activities, they don’t want to be tracked, claiming that users have control over what data is stored and used by AI. Generative Artificial Intelligence is still very much in the Wild West. Google’s Astra promises robust encryption to protect user data. However, for historical concerns, if you all remember the perceived security of “incognito” browsing, you’ll know that absolute privacy is challenging to guarantee. Furthermore, does the average user know how to turn off tracking? Education in privacy literacy may be another role that libraries offer to their customer base through programming, one-on-one tech time and resources.
The future of GenAI is promising and exciting, with models like ChatGPT 4o, Astra, and CoPilot leading the charge. These advancements can significantly benefit libraries, enhancing everything from customer interaction to data management. However, innovation must be balanced with vigilance. The privacy concerns associated with these technologies necessitate a thoughtful and proactive approach. By continuously evaluating and updating our privacy practices and educating our customers, we can better embrace AI.