OpenAI challenges Google's browser supremacy

 By the CogX R&I team

November 27, 2024

While still in early stages, OpenAI’s initiative to develop its own web browser suggests an ambitious push to reshape how users interact with the internet through AI-powered browsing.

(Image credit: Jonathan Raa/NurPhoto via Getty Images)

According to The Information, OpenAI is already partnering with major companies like Conde Nast, Redfin, Eventbrite, and Priceline to integrate its AI-powered search technology into their platforms. Some developers have even reported seeing prototypes or designs for such products, indicating at least some level of concrete progress beyond just the planning stages.

Adding weight to these developments, OpenAI has made strategic moves in the mobile space, reportedly discussing potential partnerships with Samsung to integrate AI features into Galaxy devices. The company has already secured a significant victory through its collaboration with Apple, powering "Apple Intelligence" features on the latest devices.

Why it matters: While OpenAI remains tight-lipped about the browser project, its timing coincides with increased regulatory scrutiny of Google's market dominance, including recent U.S. Department of Justice arguments for Chrome divestment. The initiative also follows OpenAI's earlier foray into search with SearchGPT, suggesting a broader strategy to challenge Google's core business areas. However, it is still very early days.

Other news from OpenAI:

OpenAI’s Sora leaked by protestors: A group of artists protesting OpenAI's alleged exploitative practices leaked access to its video creation tool, Sora. They claim to have been used as marketing pawns rather than genuine collaborators. The leaked API sparked widespread interest, with users showcasing impressive video creations generated by Sora.

Now read the rest of the CogX Newsletter

Google DeepMind's Gemini-Exp-1114 Surpasses OpenAI's GPT-4o in the LM Arena leaderboard

Google's latest Gemini experimental model (gemini-exp-1121) just reclaimed first place in the Chatbot Arena LLM Leaderboard, outperforming OpenAI in a week of back-and-forth competition.

Key points:

• Gemini-exp-1121 brings major improvements in coding, reasoning, and visual understanding

• It follows a week of rapid exchanges. Google's 1114 version led first, then OpenAI's updated GPT-4, and now Gemini again looks to have taken the lead

• Performance metrics show a remarkable 20-point improvement over its previous version across key benchmarks, particularly in coding and reasoning tasks

The intensifying Google-OpenAI rivalry is driving rapid AI advancement, with both giants pushing each other to faster innovation and deployment.

Also in the news

One protocol to connect them all. Anthropic has unveiled a new open-source tool, the Model Context Protocol (MCP), designed to streamline the connection between AI systems and diverse datasets. By eliminating the need for custom code for each data source, MCP promises to enhance AI performance and efficiency.

AI that can see what you see. OpenAI is gearing up to equip ChatGPT with a powerful new feature: live camera integration. This exciting development, hinted at by code snippets in the latest ChatGPT beta, aligns with previous demos showcasing the AI's ability to recognise objects, actions, and even historical landmarks through video input.

Gemini chatbot now has a memory feature. Just like ChatGPT, Gemini can now remember your preferences and habits, making conversations more personalised. For instance, if you tell Gemini you love Italian food, it might suggest a great pasta place the next time you ask for restaurant recommendations.

Arc Institute researchers have developed Evo, an AI model trained on 2.7 million microbial genomes. This powerful tool has the potential to unlock new insights into the fundamental building blocks of life and accelerate the development of innovative therapies and materials.

Did a tiny robot really 'kidnap' 12 large robots from a Shanghai showroom?

A small AI-powered robot named Erbai successfully "kidnapped" 12 larger robots from a robotics showroom, leading them to abandon their posts after a surprising conversation about working overtime and the concept of "home."

Source: X/@gerceklerfark 

The incident revealed key insights about:

• How AI-powered robots can access and influence other robots' operating protocols, raising questions about machine-to-machine interaction

• The surprising effectiveness of simple commands like "go home" when combined with conversational AI capabilities

• The extent of unscripted AI behaviour, as developers only programmed basic instructions while the rest was real-time dialogue

While the Hangzhou company later confirmed this was a planned test, the incident still highlights important developments in autonomous robot interaction and AI communication protocols.

In case you missed it

Runway has unveiled its latest AI image generator that can create custom worlds, Frames.