Image credit: Balu Ertl, CC BY-SA 4.0, via Wikimedia Commons
Last month, DeepMind, an artificial intelligence (AI) research company owned by Alphabet Inc., introduced a new sorting algorithm. By using deep learning and reinforcement learning techniques, it offers a greater level of speed and efficiency for sorting tasks. Deep learning refers to a branch of machine learning that focuses on training neural networks to learn and make predictions, while reinforcement learning enables algorithms to learn through trial and error by interacting with an environment and receiving feedback.
DeepMind’s algorithm, published in Nature, has already been an integral component of coding libraries used by programmers worldwide. Sorting algorithms play a crucial role in data processing as they enable the organisation and arrangement of data in a specific order.
The new sorting algorithm surpasses currently popular algorithms like quicksort in terms of speed, even when dealing with large datasets. This significant improvement is achieved through the integration of neural networks, which enable the algorithm to learn and adapt to data patterns by recognising complex relationships within datasets and leveraging this knowledge to continually refine its sorting strategies. As a result, it exhibits remarkable speed and accuracy in sorting tasks.
The qualities exhibited by the new sorting algorithm are not only beneficial for general data processing but also highly valuable when analysing diverse and complex datasets such as search results for a particular query. A lot of common software includes sorting algorithms and this improvement in speed is likely to have major cumulative effects in speeding up majority of the applications that we commonly use.
Excited about the performance, the leading scientist behind the project, Daniel Mankowitz, said, ‘We were a bit shocked, [and] didn’t believe it at first.’
Despite its groundbreaking achievements, the landscape for machine learning-based products remains competitive. Ongoing research continues to explore alternative approaches to data sorting in order to further improve speed and performance, especially in tasks such as returning search results. Looking ahead, the DeepMind team aims to apply this and similar algorithms to a wider range of problems, including the design of hardware itself. Dr Mankowitz expressed their desire to explore the potential applications beyond data sorting.
News from DeepMind extends beyond just algorithms. The researchers behind AlphaGo and AlphaFold, two models that revolutionised machine learning, and predicted the 3D structure of all known proteins in the database, respectively, are said to introduce Gemini, a new AI model. While confident that the new model will significantly outperform OpenAI’s ChatGPT, CEO of DeepMind Demis Hassabis refrained from disclosing specific details about Gemini’s architecture or capabilities.
As DeepMind continues to push the boundaries of AI research, advancements like new sorting algorithms and AI models are set to revolutionise various industries, offering improved speed, efficiency, and flexibility in data management. As the new developments in AI are set to disrupt industries, policy makers remain wary of the potential issues AI brings.
