Google Says New AI Model Outperforms ChatGPT

Improve the News Foundation -

Facts

  • Google has announced the new Gemini model of its artificial intelligence (AI) chatbot named Bard, claiming it outperforms ChatGPT and offers 'advanced reasoning' across multiple formats. Gemini, the first new release since tech firms attended an AI safety summit in the UK last month, outperformed ChatGPT's latest model, GPT-4, on 30 out of 32 benchmark tests, including reasoning and image understanding.1
  • The first version of Gemini comes in three varieties, including Gemini Ultra for 'highly complex tasks,' Gemini Pro for 'a wide range of tasks,' and Gemini Nano, which can be used for specific tasks and on mobile devices. The model can also program code in languages such as Python, Java, and C++.2
  • Gemini Pro was able to outperform human experts on benchmarks including problem-solving, math, physics, history, law, medicine, and ethics — the first AI model to do so. Gemini could also accurately link a video of a person acting out a scene from the 'Matrix' to the movie.3
  • According to Google, Gemini Pro was also able to outperform ChatGPT's free version in six out of eight benchmarks, including math and language understanding, with Gemini Ultra, the most powerful model, able to outperform the more powerful paid version, GPT-4. However, Gemini Ultra won't be made publicly available until early 2024.4
  • Gemini Pro has been released in the US, but Ultra will first undergo external safety testing, with the results shared with the US government, before its release. Not even the Pro version has been released in the UK or European Economic Area but the company is in talks with 'local regulators.' Google did admit that the model is still experiencing false answers, known as 'hallucinations.'1
  • Meanwhile, the latest version of Bard has been released in English in over 170 countries and territories, with the company saying it will include more languages and places later on.2

Sources: 1Guardian, 2Forbes, 3France 24 and 4Pcmag.

Narratives

  • Narrative A, as provided by Business Insider. While the jury is still out on whether GPT-4 or Bard has a higher potential, Google's Gemini model seems to to outperforming its OpenAI counterpart at the current moment. Not only has it outscored the competition in 30 of 32 benchmarks, but it can respond to code, audio, images, and videos, not all of which can be processed by GPT. To top it all off, Bard's Gemini model is now the only AI model easily integrated into smartphones.
  • Narrative B, as provided by New York Times. While Bard has certainly progressed since launching, ChatGPT's earlier release and Microsoft's significant investment propelled it ahead in user base and technological advancement. Google's hurried and flawed announcement of Bard further cemented ChatGPT's lead, while key personnel departures from Google due to safety concerns hindered its progress. Technological superiority doesn't matter if Google can't steal users from OpenAI.

Predictions