Exploring the Latest Features of Claude 3.5 and What We Can Do With Them

Anthropic’s Claude 3.5 is one of the latest updates in the field of generative AI. This new and improved GenAI chatbot introduces powerful features that extend its usability across various domains. With new features like prompt advising, artifacts, and tools to interact with computer files, Claude 3.5 aims to serve businesses, developers, researchers, and general users alike. Anthropic offers three robust versions of Claude 3.5 – Sonnet, Haiku, and Opus. In this article, we will try out the latest features of the Claude 3.5 Sonnet and explore some of its practical use cases.

New Features of Claude 3.5

Let’s begin by testing out the latest features of Claude 3.5 and see how they can be used in various domains, across sectors.

Here are the features we’ll be exploring in this section:

1. Analytics Tool and Artifacts for Interactive Outputs
2. Visual PDFs
3. Custom Interface and Prompting Advice
4. Enhanced Context Window
5. Computer Files Interaction

1. Analytics Tool and Artifacts for Interactive Outputs

Claude 3.5 now has a dedicated Analytics Tool that can write and run code to process data, run analysis, and produce data visualizations in real-time. The new “Artifacts” feature allows the generation of dynamic outputs, including code previews, animations, and interactive visualizations. These could be in the form of code, SVG graphics, React components, etc. These features are particularly useful for design tasks, code debugging, and creating collaborative presentations.

Let me show you this with a simple example. I’m going to upload a .csv file to Claude 3.5 and ask it to create some graphs and charts to explain the data.

Prompt

“Explain the contents of the csv file through graphs, charts, and other visuals.
1. Show the relationship between population and GDP
2. Create a bar graph showing the top 10 highest GDP countries
3. Create a chart listing the top 10 countries with GDP per capita”

The chatbot first analyzed the csv file using the Analytics Tool and created an interactive scatter plot showing the relationship between population and GDP. The values of each point on the graph pops up as I hover over it. However, the units (million people, billion $, etc) were marked incorrectly, and GDP was wrongly written as Population.

Nonetheless, the interface was dynamic, interactive, and user friendly. The artifacts were generated in a separate panel on the right side, with the explanations given on the chat window on the left. This made the content easier to read and understand.

It then created an interactive bar graph listing the top 10 countries based on GDP. The exact GDP value of each country is shown when I hover over the bar.

The third part of our prompt included calculating the GDP per capita, using the data in the file, and then displaying the top 10 values in the form of a chart. Claude 3.5 did the calculations correctly and listed the countries in a table format.

When asked to “create an image of the world map marking the countries in different colors from red to blue – in ascending order of their GDP per capita.”, Claude 3.5 created the following SVG file.

As you can see, the SVG file does not show the world map and country outlines correctly. However, it does mark the regions in the positions and the respective colours, giving us a basic understanding.

Finally, when asked to “create a JS-based presentation for me to teach this to 8th grade students”, Claude 3.5 Sonnet, created the following interactive ppt:

Use Cases

• Designers can preview and tweak web interface prototypes interactively.
• Researchers can build presentations with real-time visualization of trends.

3. Custom Interface and Chat Suggestions

The Artifacts feature also brings in an interactive interface that splits the screen into two while generating responses. This improves user experience and readability. The dynamic code generation shown in real-time also adds to an upgraded UI/UX.

This split-screen feature for real-time content or code generation is quite similar to GPT-4o with Canvas. You can check out its features and use cases here.

Claude 3.5’s new interface also provides users with a ‘Chat controls’ feature on the right-hand side. This section lists down the different artifacts and responses generated in the selected conversation, for easier transition and version control. It also lets users choose a font under ‘Chat styles’, which includes a Dyslexic-friendly version, making the chatbot more accessible and inclusive.

Prompting Advice is another feature that is a part of this new interface. At the end of each generated response, Claude 3.5 guides users with prompts to ask follow up questions or request edits. This helps users craft more effective prompts, optimizing interactions, and improving the relevance of outputs.

Here’s some prompting advice provided by Claude 3.5 for the responses generated in our earlier example:

Use Cases

• Coders and content creators can easily switch between various types of content, such as blogs, social media posts, press releases, etc. created within the same conversation, thanks to the new interface.
• Students and researchers can know the right prompts to put in to delve deeper into a topic. The suggestions may even cover aspects you may not have thought of.

2. Visual PDFs

Claude 3.5 can now read not just the text, but also images, infographics, and hand-written notes in PDF files. This enhanced reading capability enhances the chatbot’s ability to understand and analyze the visual components of PDF files. You can now interact with scanned copies of textbooks, images in research papers, and survey reports, thanks to Claude’s new ‘Visual PDFs’ feature.

To test this out, I uploaded a 24-page report by BCG titled, “Where’s the Value in AI?” and asked Claude 3.5 Sonnet to create an interactive dashboard based on its contents.

Prompt: “Read the attached PDF and provide a comprehensive analysis and summary.Create a detailed, visually appealing, and interactive dashboard that presents the key findings and insights from the report.”

Here’s the dashboard generated by Claude 3.5 Sonnet:

Similar to our previous example, the chatbot first analyzed the pdf, its charts, graphs, and other images, to give us insights into its data. It then created an interactive dashboard to show actionable insights and key takeaways.

Use Cases

• Business analysts can upload reports with sales dashboards for trend identification and forecasting.
• Handwritten notes can be analyzed and converted to digital format easily.

4. Enhanced Context Window

With a 200,000-token context window, Claude 3.5 is capable of processing large datasets, lengthy documents, or complex dialogues. This feature enables seamless handling of extended workflows, such as analyzing entire research papers or generating comprehensive reports.

Use Cases

• Researchers can summarize larger volumes of content in one go, speeding up the research process. They can also proofread, edit, and format their research papers easily.
• Educators can upload syllabi or academic papers for content summarization or restructuring.

5. Computer Files Interaction

Claude 3.5 can now directly create files in your personal systems. The AI chatbot can access your computer and do simple file generation and organizing tasks based on simple prompts. It can move your cursor, click on buttons, type text on the screen, take screenshots, explore files, and create new ones just like you would do yourself. You can watch a demo of this below:

Please note that this new feature, still in the beta phase, is currently only available in APIs. This could be a groundbreaking application of LLMs which could make day-to-day work simple for anyone who uses a personal computer.

Use Case

• Individuals and organizations across sectors can implement this feature to access, analyze, and edit local files using simple prompts. This would help automate mundane tasks and streamline workflows.

How to Access the Latest Features of Claude 3.5

Conclusion

The latest features of Claude 3.5 surely take it forward in the field of generative AI. From its ability to read visual PDFs and create interactive artifacts to helping developers with computer use, Claude 3.5 is expanding the bounds of AI chatbot applications. These updates set a new benchmark in AI-driven productivity tools. The enhanced context window and new dynamic interface further enhance its appeal, making it a versatile tool for developers, businesses, educators, and researchers alike.

While it still has room for improvement, particularly in the precision of creating artifacts, these innovative features demonstrate Anthropic’s commitment to driving accessibility and efficiency in the AI landscape. As we await the release of Claude 3.5 Opus, it’s only a matter of time till we get to know just how powerful Claude can be.

Frequently Asked Questions

K.C. Sabreena Basheer

Sabreena Basheer is an architect-turned-writer who's passionate about documenting anything that interests her. She's currently exploring the world of AI and Data Science as a Content Manager at Analytics Vidhya.