In May 2023, at the Google I/O developer conference, the CEO of Google Sundar Pichai announced the upcoming artificial intelligence (AI) system Google Gemini. The large language model (LLM) is being formulated by the Google Deep Mind team. Google Gemini AI may compete with other AI systems like Chat GPT from open AI and even go on to outperform them.
Though the details available are scarce, there is ample evidence that you can gather about Google Gemini from the latest interviews and announcements.
Google Gemini is expected to be multimodal
Pichai states that Google Gemini AI combines the strength of DeepMind’s Alpha Go system which is known for mastering the complex game Go. It is available with extensive language modelling capabilities. From the ground, it is designed to be multimodal integrating images, text and other data types. This may set the tone for natural conversational abilities. Pichai also explored future abilities like planning or memory that could enable tasks that require reasoning.
Gemini may be using tools and APIs
In his update to his professional bio over the summer, the chief scientist of Google Jeffrey Dean said that understanding of what is Google Gemini suggests that it is the next-generation multi-nodal. He was co-leading the module.
He stated that it will be using Pathways which is the new Google AI structure to make it possible to scale up training on many datasets. This suggests that Gemini may be the biggest language model developed to date.. It goes on to exceed the size of GPT-3 with over 175 billion parameters.
It is available in various sizes and capabilities
A few additional details were obtained from Demis Habbis, CEO of Deep Mind. In June he told Wired that techniques from Alpha Go like tree search and reinforcement learning may provide Google Gemini AI with new capabilities like problem-solving and reasoning. He went on record saying that Gemini is a series of models that will be made available in various sizes or capabilities.
Google Gemini may utilize memory, fast checking against sources like Google Search and improved reinforcement. The learning will enhance accuracy and trim down hallucinated content.
The early results of Google Gemini are promising
In a September Time interview, Hassabis stated that Google Gemini AI aims to combine scale and innovation. He was of the view that incorporating planning and memory was necessary in an early exploratory stage.
Hassabis also stated that Google AI may go on to employ retrieval methods to output entire blocks of information rather than a word-by-word generation to improve factual consistency. He disclosed that Gemini builds on DeepMind’s multimodal work, including the Flamingo picture captioning system..
In the overall context, Google Gemini is showcasing very promising early results.
Advanced chatbots in the form of universal personal assistants
In an interview with Wired, published a few years later, Pichai gives early indications of how Gemini fits into the product map of Google. He claimed that while conversational AI systems like Bard are a step in the right direction, they are not the goal in itself.
Pichai stated that Google Gemini along with future iterations are likely to become incredible future assistants. They are likely to be integrated into the daily lives of the people like work, travel and entertainment.. Gemini predicts that in a few years, chatbots will seem insignificant in contrast since it will mix the power of words and pictures.
A few selected companies have early access to Gemini
There are more clues about the progress of Gemini this present week. According to early information available Google provided a small group of developers outside Google early access to Gemini.
However, the general feeling is that Gemini may soon be ready for a Meta release and integration into services like Google Cloud vertex AI.
Meta is working on LLM to compete with open AI
The early news about Gemini is promising till this juncture, but Google is not the only company to launch a new LLM to compete with open AI. According to the Wall Street Journal, Meta is working on an AI model that will compete with the GPT model that powers the ChatGPT.
A testimony to this fact is that Meta recently went on to announce the release of Llama 2 which is an open-source AI model in partnership with Microsoft. The business seems committed to safely making AI more widely available.
Google Gemini and an eye on the future.
What is Google Gemini we have a fair idea by now but it is expected to represent a significant advancement in natural language processing. The future of Deep Mind’s latest AI research with Google’s vast computational resources could make the potential impact challenging and difficult to overstate.
If Google lives up to its expectations, it could bring about a revolution in interactive AI. This may align with the ambitions of Google to bring AI responsibly ways to billions of people.
The latest news from Google and Meta comes after a few days of the AI insight forum. It was at this juncture, that tech CEOs privately met with a portion of the United States Senate to discuss the future of AI.
For more such blogs, Connect with GTECH.
Publications, Insights & News from GTECH