Published on February 16, 2024, 3:31 pm

Google has introduced its flagship suite of generative AI models, apps, and services called Gemini. Developed by Google’s AI research labs DeepMind and Google Research, Gemini comes in three different versions: Ultra, Pro, and Nano. What sets Gemini apart from other models is that it is “natively multimodal,” meaning it can work with more than just text. It has been trained on various audio, images, videos, codebases, and text in different languages.

Gemini offers a wide range of capabilities. For example, Gemini Ultra can help with tasks such as physics homework by solving problems step-by-step or identifying relevant scientific papers and extracting information from them. It can even generate artwork. However, some of these capabilities are still in development.

Gemini Pro is an improvement over LaMDA (another Google model) in terms of reasoning and understanding capabilities. Independent studies have shown that it outperforms OpenAI’s GPT-3.5 in handling longer and more complex reasoning chains. However, like all large language models, Gemini Pro has its limitations and may struggle with certain math problems.

Gemini Nano is a smaller version of the Gemini Pro and Ultra models that can run directly on some phones instead of relying on remote servers. It powers features like summarizing recorded conversations in the Recorder app and suggesting replies in messaging apps through Gboard.

Google claims that Gemini performs well on benchmarks compared to other models but early user impressions have been mixed. Some users have reported issues with basic factual accuracy, translations, and coding suggestions.

As for pricing, Gemini Pro will cost $0.0025 per character while output will cost $0.00005 per character once it exits preview mode in Vertex AI. Pricing for Gemini Ultra has yet to be announced.

To experience Gemini firsthand, users can access the Gemini apps or use the API via Vertex AI or AI Studio for development purposes. Additionally, Google has incorporated Gemini models into its dev tools for Chrome, Firebase mobile development platform, and Duet AI for Developers.

In conclusion, Google’s Gemini suite of generative AI models offers a wide range of capabilities that go beyond text generation. While it shows promise in some areas, there are still challenges to overcome. As with any technology, it is important to evaluate its capabilities and limitations before fully relying on it.


Comments are closed.