Image Courtesy: Unsplash
Google has introduced its long-awaited AI model, Gemini, – a year after the launch of OpenAI's ChatGPT -- with its top bos calling it its "most capable and general AI model yet."
Launched Wednesday, Gemini, which will be integrated into Google products like its search engine, debuted in over 170 countries, including the US.
Initially, it's released as an upgrade to Google’s chatbot, Bard. However, the Bard upgrade won't be available in the UK and Europe until Google obtains approval from regulators.
Gemini comes in three versions and is "multimodal," meaning it can understand text, audio, images, video, and computer code all at once.
"Gemini Ultra’s performance exceeds current state-of-the-art results on 30 of the 32 widely-used academic benchmarks. With a score of 90.0%, Gemini Ultra is the first model to outperform human experts on MMLU," said Google CEO Sundar Pichai
Demis Hassabis, the CEO of DeepMind (the London-based Google unit behind Gemini), described it as their most complex and ambitious project yet, requiring an enormous effort.
Alphabet mentioned producing three Gemini versions, each designed for different processing power needs, with the most robust version suitable for operating in data centres, and the smallest efficient for mobile devices.
Jeff Dean, chief scientist at Google Deepmind and Google Research, claimed Gemini Ultra is the first model to achieve human-expert performance on MMLU across 57 subjects with a score above 90 per cent.
It also achieved a new state-of-the-art score of 62.4 per cent on the new MMMU multimodal reasoning benchmark, outperforming the previous best model by more than 5 percentage points, he said.
"Impressive," said X boss Elon Musk on his platform.
Gemini underwent testing in various areas like General, Reasoning, Math, and Code to gauge its text-handling capabilities. Additionally, image, video, and audio tests were conducted to assess its multimodal skills. Gemini outperformed ChatGPT in most tests except for "HellaSwag," focused on everyday common sense reasoning.
Max Tegmark from MIT in a post on X claimed that GeminiAi got the top MMLU score.
Gary Marcus, a prominent AI expert, in his post said that from a commercial standpoint, GPT-4 is no longer unique. However, he said, Gemini's performance was less than mindblowing.
"That's a huge problem for OpenAI, especially post drama, when many customers are now seeking a backup plan," he wrote.
Marcus also questioned if Large Language Models, or LLMs, were plateauing in development, pointing to hints from influential figures and the absence of GPT-5 despite high demand.
Comments