AI Made Friendly HERE

5 ways Google Gemini can rule AI in 2025

The AI landscape moves fast. Google built momentum in 2024 by iterating on Gemini, introducing NotebookLM, and driving its AI into ever-deeper Workspace integrations. However, Google needs to deliver on its promises to lead in 2025. This article explores how Google can achieve this by surpassing OpenAI’s Sora with its Lumiere video generation technology, improving Imagen, and supercharging NotebookLM with AI video avatars.

Google must move beyond the “agentic AI” hype and deliver tangible productivity enhancements across its product ecosystem. By leveraging the synergy between its Tensor processors and Pixel devices, Google can showcase the power of on-device AI and demonstrate its prowess in Android smartphones.

Related

I tested NotebookLM and saw the promise of something great

NotebookLM has a lot of issues, but the idea behind it has potential

5

Surpass Sora

Google needs to hit big with Lumiere

A robot in a snowstorm by Sora

Sora’s rendition of a robot in a New York City snowstorm

OpenAI’s Sora has captured interest by generating splashy videos of up to 60 seconds in length. Where previous solutions struggled to create complex scenes and realistic movement, Sora has shown promise. While only available to paying GPT-4 subscribers, Sora doesn’t disappoint. It’s fun to use, has a real wow factor, and gives OpenAI bragging rights in this segment.

Google must accelerate its AI video efforts and surpass Sora’s achievements to compete. Lumiere, Google’s upcoming answer to Sora, needs to be impressive. Lumiere can make its mark by delivering a beautiful visual storytelling experience powered by simple text prompts. Leveraging Google’s expertise in areas like Search could give it an edge in context and coherence. Combine that with longer video output capability, and you have a winner. I am eager to see what the company can do here.

4

Overhaul Imagen (again)

Rivaling Dall-E will silence critics

Imagen-Sample-1

Source: Google Research

I’ve been critical of Imagen. I can respect the need for protection against malicious threats like deepfakes. However, it’s harder to swallow an AI image system with a history of exhibiting biases. Still, there are signs Imagen might be getting better.

Imagen was re-released a few months ago, and I’ve had varying results. Most of the images it created either felt safe and watered-down or weren’t interesting. There have been a few exceptions. The other day, it created a beautiful surrealist image when I prompted it to “Create a stunning image of a complete ecosystem.” That experience gave me hope that Google will continuously iterate to improve Imagen.

Now, if it can only figure out how to render compelling images of people without stumbling into a minefield of its own creation. Let’s hope Imagen pulls up to rival Open AI’s Dall-E to help Google win in AI in 2025.

3

Supercharge NotebookLM with AI video hosts

NotebookLM could be even better with video presenters

NotebookLM helped me work smarter in 2024

NotebookLM has impacted my personal and work productivity. It’s an AI notebook that you can train on your personal data. It’s free of internet noise and is an inhumanly fast personal research assistant. One surprise twist is that NotebookLM can take notes, links, and other sources, synergize the information, and create fun and informative virtual podcasts. The depth of conversations it creates and interactions between the hosts are uncanny. It delivers an unmatched level of dynamic realism.

The one thing NotebookLM doesn’t have is the ability to output video podcasts featuring virtual presenters. Adding the ability to output videos hosted by virtual presenters would be awesome. NotebookLM creates amazing dialogue quality from bland source materials. So why not bolt on avatar video capability, such as those found in applications like Synthesia? This would make the content more entertaining and provide new options for content creators and marketers to bang out good stuff quickly. NotebookLM offers a unique value proposition, and Google can raise the bar further by adding video.

2

Prove that the ‘agentic era’ is more than a catchphrase

Agentic AI sounds great, but Google needs to deliver on the hype

Google-IO-keynote-Gemini-for-Workspace-Hero

Google

The “agentic era” of AI, characterized by proactive and collaborative AI systems, holds promise, such as managing complex supply chains. However, agentic solutions have felt more like an amped-up version of Assistant than a revolution. With its full-stack ownership and reams of data to infer context and intent, Google is uniquely positioned to prove that the agentic era is more than a snazzy catchphrase or a new way to control smart speakers.

The future of agentic AI for personal applications needs to include proactively solving more complex tasks across Workspace and beyond. For example, it needs to execute productivity value adds such as proactively drafting emails, proposing meeting times, and collaborating on preparation.

In the last week, I’ve noticed that Gemini scans my Workspace more frequently and independently for context and information. This is an interesting development as it potentially blends the traditional run-of-internet knowledge of Gemini with the personal dataset capabilities of NotebookLM. So far, the results of these automatic queries haven’t been remarkable. I’ve also noticed Gemini scanning my personal data when I want to find an internet-based answer to my query. Let’s hope Google’s ah-ha moment for agentic AI is around the corner.

1

Show us what’s possible with Tensor and Pixel 10

Owning the whole stack is a huge competitive advantage

Four Google Pixel phones arranged on a pink surface next to playing cards

Google’s Tensor chip and the Pixel smartphones that incorporate it provide a huge market advantage. Owning the full stack, from software to hardware, gives Google unique data insights, optimization opportunities, and interfaces for its AI. By integrating hardware and software, Tensor allows for innovative features like live translation, advanced image processing, and personalized user experiences.

8:45

this is the future of smartphones thumbnail

Watch our review

Google Pixel 9: This is the future of smartphones

Our look at the Google Pixel 9 in full

Now, Google needs to leverage its processors and smartphones to deliver a complete AI ecosystem. It must showcase Google’s unique capabilities to address latency, privacy, and connectivity concerns by extending AI processing beyond the cloud. However, Tensor faced criticism for not living up to its potential. Some pointed to performance gaps compared to competitors and an absence of groundbreaking, AI-powered features.

I want to see Google double down on Tensor’s AI capabilities in the Pixel 10. This means delivering improvements in speed and efficiency, along with innovative features that showcase the power of on-device AI.

What’s next for Google AI?

To lead in the rapidly evolving AI landscape, Google must capitalize on its strengths in 2025. This means surpassing OpenAI’s Sora with its Lumiere video generation technology, realizing the full potential of Imagen, and supercharging NotebookLM with AI video avatars. Furthermore, Google needs to move beyond the hype of “agentic AI” and deliver tangible productivity enhancements across its product ecosystem. Finally, by fully leveraging the synergy between its Tensor processors and Pixel devices, Google can showcase the power of on-device AI and solidify its position as the AI pacesetter.

Originally Appeared Here

You May Also Like

About the Author:

Early Bird