Google Assistant features are to be powerfully enhanced with the company’s new AI tools. The company wants to combine a well-known application with new generative AI techniques. In this way, it has the form of a “turbocharged” Assistant with much greater possibilities.
Google plans to significantly expand the Assistant app. Although the program is currently a combination of helpful tools for users and the foundation of the company’s smart home system, currently it cannot be compared to solutions such as ChatGPT or Bard AI. However, that is about to change soon.
Turbocharged Assistant with generative AI support
Google VP of Assistant Peeyush Ranjan and CTO Duke Dukellis recently shared their vision for the future for the division responsible for developing the Assistant services. They present quite rich plans for a slightly tarnished application.
“As a team, we need to focus on delivering a high-quality product experience for users. We are aware of the huge potential of generative artificial intelligence, which can even change people’s daily lives. We see a huge opportunity in exploring what a turbocharged Assistant could look like, backed by the latest Major Language Models (LLMs),” managers write in an internal email to employees.
Importantly, in the same message, Ranjan and Dukellis reveal that Google has already taken the first steps to integrating AI LLM models into the Assistant on mobile devices. However, further details are not known, although a combination of the Assistant with Google’s Bard AI is strongly expected.
Google is preparing Gemini AI
As we wrote in Wprost.pl, another branch of Google is to independently prepare a completely different model of artificial intelligence.
DeepMind CEO Demis Hassabis boldly talks about the large language model (LLM) that the company calls Gemini. This is a new foundation that can be used to create more AI tools. For example – ChatGPT uses the GPT-4 model and it is this model that allows the tool to create correct answers to numerous user questions.
Why Google’s “Gemini” should be better than the competition? The manager says that it will be a huge advantage for him to include another system in the model, which was previously developed by DeepMind. It’s about AlphaGo – an artificial intelligence that years ago became famous for beating the best Go players.
The new component is to bring to the developed artificial intelligence system the ability to plan and solve very complex tasks. AlphaGo is based on the technique of learning by reinforcement. This means that after each failure, the machine takes feedback about the result and gradually changes its actions, all the time improving them. This is supposed to be the key to gradually surpassing ChatGPT.