GPT Chat, from the U.S.-based research lab OpenAI, launched in November 2022 and already surpassed 1 million users in its first week. Naturally, with the growing importance of AI, more companies want to grab a piece of the pie: Google Gemini AI is expected to be a serious competitor to GPT Chat.
What is Google Gemini AI?
Gemini is a set of large language models (LLMs) that use a number of training techniques, including searching through the so-called “language models. decision tree, through which AI is supposed to give me the most accurate answers. Google’s plans are ambitious: Gemini is set to become the dominant generative system (that is, one that generates new data based on user input) in the world. This information came only a few months after Google merged its AI Brain and DeepMind labs, creating a new research team called Google DeepMind.
Given the researchers’ prediction that the generative AI market is expected to be worth $1.3 trillion by 2032, it is clear that Google is investing significantly in this area to position itself as a leader in the development of artificial intelligence.
Everything we know so far about Google AI Gemini
Although many expect the Google Gemini to be released in the fall of 2023, little is still known about the model’s capabilities.
According to Sundar Pichai, CEO of Google and Alphabet, Google Gemini was designed from the ground up to be multimodal, meaning it can handle different types of data, such as text, images and other types of information, simultaneously. Thus, it can enable more natural and versatile conversational abilities.
Pichai also hinted at future capabilities, such as memory and planning, that could enable tasks that require reasoning.
Google’s Jeffrey Dean also noted that Gemini will use Pathways, Google’s new AI infrastructure, to enable scaled training on diverse data sets. This suggests that Gemini could potentially be the largest language model ever created, possibly exceeding the size of GPT-3 with more than 175 billion parameters.
Demis Hassabis, CEO of DeepMind, provided additional information: he said Gemini will use techniques such as reinforcement learning and tree search that can give it reasoning and problem-solving abilities.
Reinforcement learning involves the model being rewarded for taking certain actions, allowing it to learn which of those actions lead to the desired results. This is a technique often used in machine learning so that models can improve their behavior in dynamic environments.
Tree searching is a technique that involves exploring different possible sequences of actions or decisions to find the best solution. This is particularly useful in problems that require analysis of multiple possibilities, which can range from strategy games to more general decision-making tasks.
Hassabis said Gemini is a series of models that will be made available in different sizes and capabilities.
He also mentioned that Gemini can use memory, fact-checking from sources such as Google Search and enhanced reinforcement learning to improve accuracy and reduce unsafe content.
Will Gemini take the crown from ChatGPT?
One of the most important topics in the context of Gemini’s launch is whether the mysterious language model has what it takes to sideline ChatGPT, which surpassed 100 million monthly active users this year.
Initially, Google used Gemini’s ability to
generate text
and images to differentiate it from GPT-4, but on September 25, 2023, OpenAI announced that users would be able to input voice and image queries into ChatGPT.
Now that OpenAI is experimenting with a multimodal model, perhaps the most serious difference between the two is Google’s massive training database. Google Gemini may process data from services such as Google Search, YouTube, Google Books and Google Scholar. Using this data to train Gemini models can bring an advantage in terms of the sophistication of the inferences and conclusions that can be drawn from the data set.
Will the solution from Google allow for better content generation? This will naturally turn out when it enters the market. However, it is to be expected that we will see rapid development of AI in the coming years. And how to use it to generate texts? I wrote a bit more about it:
here
.
