Arc Browser browses for you

And: Bard gets image generation | AI2's new open-source LLM - OLMo 7B

Author

AI Periscope
February 01, 2024

Exploring below the surface of AI headlines.

Summaries | Insights | Points of Views

In Today’s Edition

Arc Browser

Image source: engadget

Summary - The Arc browser, developed by The Browser Company, is introducing AI-powered features designed to revolutionize how users interact with the web by blending the functionality of a browser, search engine, and individual websites into one seamless experience. The upcoming Instant Links feature allows users to quickly access search results, while Live Folders will aggregate updates from selected sites.

Buoy points:

  • Arc Search App Integration: The app leverages AI to create custom webpages that aggregate diverse internet sources in response to user queries, embodying the concept of a browser that "browses for you."

  • Instant Links Feature: By pressing shift and enter during a search, users can automatically open the top search result, streamlining the process of finding specific information or content, like video trailers or recipes.

  • Live Folders for Real-Time Updates: This feature acts like an advanced RSS feed, automatically updating a folder with new content from selected sites, indicating a move towards more dynamic and personalized content curation.

  • Arc Explore's Advanced AI: Utilizing Large Language Models (LLMs), Arc Explore aims to merge browsing, searching, and website interaction into a singular, efficient process, showcasing a more ambitious use of AI to enhance web navigation.

  • User Experience Focus: Despite the innovative AI integration, there's an acknowledgment of the challenge in altering entrenched web browsing habits, highlighting the user experience's central role in the adoption of new technologies.

  • Future Potential for AI in Web Browsing: The developments hint at the untapped potential of AI in transforming web interactions, suggesting a future where AI could further personalize and streamline online experiences.

POV - This is interesting and it’s going into the “keep an eye on this” category for me. My take? Google just announced their initial set of AI features for Chrome. If this seamless and more dynamic web browsing experience gets traction, I wouldn’t be surprised with either an acquisition by Google, or Arc getting muscled out. What are your thoughts?

Bard Image Gen

Image source: Google

Summary - Google's Bard chatbot has introduced AI-powered image generation capabilities, leveling the playing field with its rival, ChatGPT Plus, which includes DALL-E 3 integration for image creation. This new feature utilizes Google's Imagen 2 model and is part of Bard's enhancements powered by the Gemini Pro large language model. With a focus on responsible AI use, Google has implemented safety measures and watermarking to address concerns over content misuse.

Buoy points:

  • AI Image Generation to Bard: Google has updated Bard to generate images using the Imagen 2 text-to-image model, catching up with the capabilities of ChatGPT Plus.

  • Responsible AI Use: Google emphasizes responsible AI development, integrating watermarking into images generated by Bard to signal they are AI-created and establishing guardrails to prevent misuse.

  • Launch of ImageFX: Alongside Bard's new feature, Google has released ImageFX, an experimental tool that allows users to create images with simple text prompts.

  • Expansion of Bard's Language Support: Bard now supports over 40 languages and is available in more than 230 countries.

  • Accessibility and Cost: Unlike ChatGPT Plus, Bard's image generation feature is available at no cost, making advanced AI capabilities more accessible to a wider audience.

POV - Since it just dropped, I’ve only had limited time to mess with Bard’s image gen. Thus far, ChatGPT Plus wins ($20/mo), though it’s a ‘paid vs. free’ comparison. It is safe to say that Bard will improve quickly.

Open-source AI

Summary - AI2's release of the open-source large language model (LLM) OLMo 7B represents a significant shift back towards transparency in AI research. By sharing everything from training data to source code on platforms like GitHub and Hugging Face, AI2 aims to foster an environment where the AI community can freely access, critique, and improve upon their work, addressing the problems that have arisen from the proprietary nature of recent supersized LLMs like GPT-4.

Buoy points:

  • Open Source as a Standard: AI2's initiative to release OLMo 7B and its associated components openly challenges the recent trend of treating AI models as proprietary, aiming to restore the collaborative ethos that characterized early AI research.

  • Comprehensive Transparency: OLMo 7B's release includes not just the model but also the training and pretraining data, source code, and intermediate checkpoints, offering unprecedented visibility into the model's workings.

  • Addressing AI Challenges: By providing full access to OLMo's components, AI2 enables researchers to tackle issues like model "hallucinations" and bias more effectively, as understanding the connection between training data and model output becomes possible.

  • Environmental and Efficiency Benefits: Open sourcing AI models like OLMo could significantly reduce the redundancy in AI research, thereby lowering the computational and environmental costs associated with independently developing similar technologies.

POV -The transparency and accessibility inherent in open-source could foster innovation through collaboration. And through collaboration, perhaps a well-needed set of standards and guardrails can emerge.