Google is embarking on a new endeavour, code-named "Magi," to address the growing competition from AI-driven search engines, such as Microsoft's Bing. Currently, Google dominates over 90% of the search market whereas Microsoft's Bing search has experienced a 25% surge in monthly page visits due to the incorporation of ChatGPT and GPT-4, which improves user prompt requests, model effectiveness, user experience, and search outcomes. In response to this competitive landscape, Google is creating an AI-based search engine designed to offer users a personalised experience by anticipating their needs.
As reported by The New York Times, Google plans to unveil its new AI-powered search features next month, with additional functionalities slated for release in the fall. Initially, these features will be accessible exclusively to a maximum of one million users in the United States. The exact offerings of the new tools remain to be seen, but they are expected to expand upon the conversational basis of Google's experimental Bard chatbot. Developed under the code name "Magi," these new search tools represent Google's strategy to counter competition from emerging systems like Microsoft's Bing chatbot and OpenAI's ChatGPT.
Let's take a deep dive into Google’s past with AI.
DeepMind: Pushing the Boundaries of AI Research
Two key players that have significantly contributed to the advancements in AI are DeepMind, an AI research company under Alphabet, and TensorFlow, an open-source machine learning library developed by Google. DeepMind Technologies Limited was founded in 2010 by Demis Hassabis, Shane Legg, and Mustafa Suleyman became a trailblazer in AI research, particularly in reinforcement learning and deep learning (Hassabis et al., 2010). Acquired by Google in 2014, DeepMind operates as a subsidiary of Alphabet, with its primary mission to develop artificial general intelligence (AGI) that can perform any intellectual task that a human can accomplish.
One of DeepMind's groundbreaking achievements was AlphaGo, a computer program that made history in 2016 by defeating the world champion Go player Lee Sedol in a five-game match (Silver et al., 2016). This triumph marked a significant milestone in AI development, given the complexity of the ancient board game and the vast number of possible moves.
Another notable accomplishment was AlphaZero, an evolution of AlphaGo that employs a more generalised approach to reinforcement learning, enabling it to learn multiple games such as chess, and shogi, and Go from scratch. It was found that in a short amount of time, AlphaZero outperformed specialised programs and human world champions in each respective game. DeepMind's AlphaFold was yet another groundbreaking AI system that predicted protein structures with remarkable accuracy. This innovation had the potential to revolutionise fields such as biology, medicine, and drug discovery by providing a better understanding of protein functions and facilitating the development of new drugs and therapies. Furthermore, DeepMind collaborated with the UK's National Health Service (NHS) on several projects, including the development of AI systems to detect eye diseases and predict the progression of kidney disease (De Fauw et al., 2018).
TensorFlow: Fueling the Growth of AI Applications
TensorFlow, an open-source machine learning library developed by Google, has become an integral tool in the AI landscape since its release in 2015 (Abadi et al., 2016). The platform provides tools for building and training a wide range of machine learning models, such as deep learning, reinforcement learning, and unsupervised learning algorithms.
TensorFlow's scalability allows it to work efficiently on various hardware platforms, from smartphones to powerful servers and cloud infrastructure. This versatility has facilitated the integration of AI into diverse applications, including image and speech recognition, natural language processing, recommendation systems, and autonomous vehicles.
As AI continues to progress rapidly, the next five years will likely see significant advancements in both research and applications. With companies like Alphabet, OpenAI, and IBM rushing to win this race.