One company that has already published early forecasts for the coming year is DAMO Academy, which is part of the Alibaba Group. It has published a detailed report outlining developments for the coming year.
Generative AI makes everyone an art curator
AI is increasingly moving towards creativity, enabling anyone with a smartphone to become a curator of artistic works. Instead of examining existing data, generative AI generates new and unique content based on machine learning algorithms trained on previous work. Developers can create images, text, program code, paintings and illustrations, and even videos and audio files. According to DAMO Academy, in the upcoming years, generative AI will be able to create content on par with human skill, making digital content creation easier.
Decision-making intelligence leads companies to success
Businesses strive to improve their decisions by using advanced technologies such as machine learning to be more successful in 2023. Fears of a pandemic and increasing risks from economic uncertainty are forcing companies to combine mathematical optimisation and AI to find the best solutions in an ever-changing and uncertain environment.
Ride-hailing platforms have already relied on decision intelligence to optimise driver dispatching, and factories are using it to schedule production to minimise downtime. According to the DAMO Academy, decision intelligence powered by machine learning will play a more significant role in carbon reduction programmes. One example is utilities that need help integrating renewable energy into the power grid.
By combining optimal mathematical models and AI algorithms, the dual-engine approach can help utilities make more accurate forecasts for power generation to reduce carbon emissions and ensure stable operations. An ecosystem and infrastructure based on generative AI will emerge to make it easier for people without a technical background to access the models and services.
Multimodal AI revolutionises the way we understand the world
Just as humans have developed different senses to explore the world, AI can now learn and act from other sources. Multimodular pre-training models combine different types of data, including image, text, language and numerical data, to better understand the world. In doing so, they have revolutionised the field of AI. Unlike earlier single-modal models, these AI models can simultaneously process and analyse many types of data.
According to the DAMO Academy, these multimodal pre-training models will be vital in increasing organisational productivity in the future by giving companies access to advanced models and data analytics.
CIPU: the future of cloud infrastructure for demanding applications
Data processing requirements have increased so much that traditional centralised processing units (CPUs) are no longer sufficient. Companies have therefore developed new server hardware to ensure processing more significant amounts of data with lower latency.
One solution is Alibaba Cloud’s Cloud Infrastructure Processing Unit (CIPU). It allows various virtualisation functions related to storage, networking, and security to be offloaded from the CPU to dedicated hardware. This will enable it to support more demanding applications such as Big Data and artificial intelligence.
According to DAMO Academy, cloud computing in the following years will include a new infrastructure system around the CIPU. It will create opportunities for core software development and the specialised chip industry.
Edge networks improve cloud data centre performance
Data centres support sophisticated tasks such as AI and machine learning. They provide large amounts of computing power, storage, and energy and allow customers to pay as they go.
However, cloud providers do not offer guaranteed network resources to their customers. This high variability in power provided by cloud networks often leads to unpredictable application performance. According to a report, a predictable network driven by edge network integration will become the norm in the coming years.
The future of image processing: AI and signal processes revolutionise smartphone photography
The DAMO Academy reports that the future of image processing will be revolutionised by using technologies such as artificial intelligence and signal processing. Smartphones will thus function like professional cameras, even without large sensors and robust lenses.
This computerised image processing will revolutionise applications from portrait lighting to vibration reduction, providing access to high-tech tools discounted in the price.
Chiplets: the future of chip design and computing power
It is generally accepted that computer chips can be made faster by putting more transistors on a piece of silicon and making them bigger. But given the growing volume of data, it would not be easy to make larger chips to deliver the computing power needed. So chip designers took an alternative approach by breaking the chips into smaller parts and putting them together. Each piece, called a chiplet, forms a larger integrated circuit.
Chiplets are less prone to manufacturing defects and cheaper to produce because they are smaller. Chipmakers can also add or remove chiplets on processors to customise the chips to their needs.
In 2022, the industry will also introduce a new standard for chiplet-based processor designs. According to the DAMO Academy, these small but powerful components will continue to transform chip research and development in the years to come.c