Alphabet, Google’s parent company, has entered the AI market with its own cutting-edge artificial intelligence model named Gemini, which is said to distil information from various sources such as video, audio, and text. Google also said it has better sophisticated reasoning inclusive of understanding nuance.
“This new era of models represents one of the biggest science and engineering efforts we’ve undertaken as a company,” Alphabet CEO Sundar Pichai said.
Google is launching its new AI offering behind several competitors, including Microsoft’s ChatGPT by OpenAI, which launched about 12 months ago.
With Gemini, Google decided to move forward with inserting some of the new model’s tech into its AI assistant Bard, with plans to release its most innovative version of Gemini through Bard in 2024.
According to Alphabet, in total, there will be three versions of Gemini, which will all employ a different amount of processing power, and the most powerful iteration will be run in data centres.
The lesser version will run effectively on mobile devices.
According to DeepMind Vice President of Product, Eli Collins, of all the projects designed by the Google DeepMind AI unit, Gemini is by far the biggest AI model. However, it is “significantly” less money to serve to consumers and clients.
“So it’s not just more capable, it’s also far more efficient,” Collins said.
There will be, however, a significant amount of computing power required to train Gemini. A problem Collins said they are already working on to ensure the process is improved.
In the announcement, Alphabet also revealed that it has developed a new generation of its custom-built AI chips, or tensor processing units (TPUs), of which the Cloud TPU v5p is made to train large AI models and is stitched together in pods of 8,960 chips.
