Artificial Intelligence | Google's Gemini Apex: New AI Model, Old Questions
Senior Technology Correspondent · AI, startups & MeitY policy
Quick summary
Google DeepMind today launched Gemini Apex, an advanced large language model that understands video, audio, and text in real-time. But critical details like pricing for India and training data transparency remain unclear.
Google DeepMind has a new offering: , it announced Gemini Apex. This is a big step up for its main large language model (LLM) series. Think of LLMs as the smart software behind tools like ChatGPT. They understand and create human-like text.
Apex promises a lot. The company says it can process information in real-time. It understands different types of data at once: video, audio, and text. This is called 'multi-modal' understanding. It should allow the AI to handle complex tasks.
Google aims for 'advanced real-world applications.' This could mean AI helping with live translations or understanding complex business presentations. Imagine an AI watching a video call, listening to the speaker, and reading shared documents all at once.
What It Does Differently
The core change in Gemini Apex is how it handles information. Older LLMs often struggled to combine different kinds of data seamlessly. Apex is designed to do this much better. It should react faster too.
This means less waiting for the AI to process your request. For developers in India, this speed and multi-modal ability could open new doors. They might build apps that interact with users more naturally, say, by understanding spoken Hindi and a video simultaneously.
That said, specific features for Indian languages weren't highlighted in the announcement. Nor was pricing for our market. Local startups, like SynapseFlow, which just raised $200 million for enterprise AI, will be watching closely.
The Unanswered Questions
The announcement was polished. The details, less so. Google DeepMind talked about capabilities but shared little about the 'how'.
One big question: What data was used to train Gemini Apex? , the EU Parliament passed new rules. These rules say generative AI developers must be open about their training data. This includes copyrighted material.
Google didn't mention this transparency in its Apex announcement. It's a critical point for trust. Without knowing this, it's hard to judge the model's fairness or legal standing.
Benchmarks, which are tests to show how well an AI performs, were also missing. We don't know how Apex compares to rivals, or even to older Gemini models. Cost and availability for developers in India remain unconfirmed.
So, we have a shiny new AI model with big promises. But the fine print, the real-world performance, and the ethical considerations are still largely in the dark. As always, the proof will be in the actual use, not just the announcement.
Key Takeaways
- Google DeepMind launched Gemini Apex , a new large language model.
- Apex promises improved real-time processing and seamless understanding of video, audio, and text.
- The announcement lacked specifics on training data transparency, performance benchmarks, and pricing for India.
People also ask
- What is multi-modal understanding?
- AI processes and understands diverse information concurrently, such as text, images, and sound.
- Does this AI work in Indian languages?
- Still unclear: The official announcement didn't specify support or performance for Indian languages like Hindi or Marathi.
- What is an LLM?
- An LLM, a large language model, trains on vast text to understand and generate human language.
- So what now?
- Developers need to await further technical details from Google, including availability and performance benchmarks.