Google Gemini AI Outperforms GPT 3.5, to Be Integrated Into Google Bard

Google Gemini AI Outperforms GPT 3.5, to Be Integrated Into Google Bard

Google Gemini AI Outperforms GPT 3.5 And Other AI Models

Google, on Wednesday, rolled out Gemini AI — the tech giant’s latest and most advanced AI model. Capable of processing audio, video, images, and text, Gemini has outperformed OpenAI’s GPT 3.5 on a majority of benchmarks.

In a major upgrade to the tech giant’s generative AI chatbot Google Bard, Google also announced that it will now be powered by Gemini AI. This would allow Bard to rival ChatGPT more effectively with enhanced reasoning, understanding, and planning capabilities. The upgrade will be implemented in two phases, Google revealed.

Gemini Outshines Rivals in Benchmark Tests

Capable of receiving different kinds of data and delivering output in the form of both text and images, the 32k context window puts Gemini AI ahead of the competition.

Google announced that the AI model will be released in three sizes for flexible scalability — Nano, Pro, and Ultra.

  • The smallest model, Gemini Nano, is designed to run on mobile devices. The model is split into two parts, with 1.8 billion parameters for Nano-1 and 3.25 billion parameters for Nano-2.
  • Gemini Ultra, on the other hand, is a data center-level AI model capable of handling complex tasks that involve reasoning and processing several data types.
  • The medium-sized model, Gemini Pro, seeks to offer more versatility and can efficiently handle a broader range of tasks.

Gemini AI was put to a series of benchmark tests to find out where it stands compared to rival AI models OpenAI, Anthropic, X, and Meta.

While Gemini Pro couldn’t surpass GPT-4, it still performed better than GPT-3.5 and all the other models. 

Evaluating Gemini’s abilities, the testers concluded that “the performance of Gemini Pro outperforms inference-optimized models such as GPT-3.5 and performs comparably with several of the most capable models available, and Gemini Ultra outperforms all current models”.

In 8 out of 10 tasks, Gemini Ultra outperformed GPT-4, Claude, Grok-1, and Llama-2. A majority of these tasks were text-based, such as machine translation, Python coding, math problems, text comprehension, etc.

However, the Gemini team warned that just like other LLMs, the new AI model still suffers from a number of limitations. These include hallucinations and difficulties in counterfactual reasoning, logical deductions, and causal understanding.

Bard to Become Significantly More Powerful With Gemini Integration

For now, Google will upgrade Bard with a specially modified version of Gemini Pro. This upgraded version will initially be released in English in over 170 countries and territories. More languages will follow soon, along with the AI model’s release in the UK and the EU, stated Google.

Sissie Hsiao, Vice-President and General Manager of Assistant and Bard at Google, said that while Gemini Pro will initially power only text-based prompts in Bard, the coming months will see it extended to support multimodal input.

This is the biggest single quality improvement of Bard since we’ve launched.Sissie Hsiao

In 2024, Google will be introducing Bard Advanced, which will be powered by the most powerful Gemini model. In addition to receiving inputs in the form of text, images, audio, and video, Bard Advanced will possess multimodal reasoning capabilities. It will also be capable of understanding, explaining, and generating high-quality code.

While Gemini will likely give Bard an edge over ChatGPT, some experts believe that considering GPT 3.5 was released over a year ago, this upgrade is more of a catch-up. It remains to be seen if Bard can actually surpass ChatGPT in terms of popularity in the future.

Read More

Leave a Reply

Your email address will not be published. Required fields are marked *