The unveiling of Gemini also included the announcement of Google's newest custom chip, the TPU v5p, which will be made available to outside developers and companies for training their own AI. Google's move towards using its own custom chips highlights the trend among big tech companies to develop their own silicon to reduce dependency on chip giants. This development could potentially reshape the AI landscape, with Google releasing three versions of Gemini with varying levels of sophistication.
Key takeaways:
- Google has unveiled its new artificial intelligence model, Gemini, which was trained on its own custom chips, not the widely-used GPUs.
- Google's use of its own chips, Tensor Processing Units (TPUs), gives it a significant advantage over competitors and allows Gemini to run faster than previous models.
- Google is releasing three versions of Gemini, with the most powerful version, Gemini Ultra, set to be released early next year. The smallest version, Gemini Nano, will power features on Google’s Pixel 8 smartphone.
- Alongside the Gemini announcement, Google also unveiled the newest version of its custom chips, the TPU v5p, which will be available to outside developers and companies to train their own AI.