Michael Rowe

Trying to get better at getting better

Gemini: Google’s newest and most capable AI model

Update (09/12/2023): Yesterday evening I came across these two posts, one by Gary Marcus and another by The Decoder, which highlight some of the issues with the demo videos that Google released. Note that these concerns are based on information that Google released themselves, so it’s not about them trying to hide anything. From the Google post: “For the purposes of this demo, latency has been reduced and Gemini outputs have been shortened for brevity.” Marcus makes the point that “Friends don’t let friends take demos seriously.” Well worth keeping in mind.


“From natural image, audio and video understanding to mathematical reasoning, Gemini Ultra’s performance exceeds current state-of-the-art results on 30 of the 32 widely-used academic benchmarks used in large language model (LLM) research and development…. With a score of 90.0%, Gemini Ultra is the first model to outperform human experts on MMLU (massive multitask language understanding), which uses a combination of 57 subjects such as math, physics, history, law, medicine and ethics for testing both world knowledge and problem-solving abilities.”

On 06 December, Google released their newest AI model, called Gemini. Early demonstrations suggest that this is a significant step in generative AI, with an emphasis on multimodality. Gemini Pro is now the model used by Bard and Gemini Nano will be running on Pixel 8 phones.

Here’s some reading.

  • Gemini – Google DeepMind. (2023, December 08). Retrieved from https://deepmind.google/technologies/gemini/#introduction
  • Knight, W. (2023). Google’s Gemini Is the Real Start of the Generative AI Boom. WIRED. Retrieved from https://www.wired.com/story/google-gemini-generative-ai-boom
  • Learn more about Gemini, our most capable AI model. (2023, December 08). Retrieved from https://blog.google/technology/ai/gemini-collection
  • Pichai, S. (2023). Introducing Gemini: our largest and most capable AI model. Google. Retrieved from https://blog.google/technology/ai/google-gemini-ai
  • Rakowski, B. (2023). Pixel 8 Pro — the first smartphone with AI built in — is now running Gemini Nano, plus more AI updates coming to the Pixel portfolio. Google. Retrieved from https://blog.google/products/pixel/pixel-feature-drop-december-2023
  • Wikiwand – Gemini (language model). (2023, December 08). Retrieved from https://www.wikiwand.com/en/Gemini_(language_model)

Yesterday Google published a series of videos on YouTube, demonstrating the capabilities of Gemini. I’m posting a few of them here for easy reference.

90 second overview of Gemini. I didn’t think it was especially useful. Not enough detail.
This is the marketing speak about how Gemini fits into the larger Google ecosystem. You can skip this one.
A series of practical examples demonstrating Gemini’s multimodal capabilities.
Mindblowing.
The implications for real-time, personalised learning experiences are staggering.
Useful if you think that generative AI is a text-based chatbot.
I wouldn’t consider a career in programming.
Short example of how well Gemini understands context.
Problem-solving, design, implementation, evaluation. These are probably the kinds of assessments we want our students doing.

Share this


Discover more from Michael Rowe

Subscribe to get the latest posts to your email.