A new chip called Taichi. It creates a large puzzle of similar microcircuits, creating a complex and powerful computing system. If scaled properly, it will be powerful enough to train and run AI in the future.

Detail

AGI is a yet hypothetically advanced form of artificial intelligence that would be as intelligent as humans in terms of their cognitive abilities. While AGI can be applied across many disciplines, the scope of today’s AI systems is quite narrow.

While some experts believe that such systems are still many years away, and the main obstacle is a bottleneck in computing power, others believe that we will have an artificial intelligence agent as early as 2027. The first steps for this have already been taken.

in recent years scientists began to reach the limits of the capabilities of traditional electronic componentsEspecially given the rise of artificial intelligence and the massive amounts of energy required to service these increasingly demanding systems. Graphics processing units (GPUs) have become essential components for training artificial intelligence systems because they are better than central processing units (CPUs) at performing parallel calculations. However, scientists say that as systems grow, the required level of power consumption becomes unsustainable.

What are photonic chips and how do they work?

In February, scientists unveiled a new type of photonic microchip design that uses photons, or particles of light, instead of electrons to operate transistors—tiny electrical switches that turn on and off when voltage is applied. In general, the more transistors on a chip, the more computing power it has and the more energy it needs to operate. LED-based chips consume much less energy and can perform calculations much faster than traditional chips because they can perform calculations in parallel.

Modern photonic chip architectures for AI models consist of hundreds and thousands of parameters, or training variables. This makes them quite powerful for basic tasks like pattern recognition, but large language models (LLMs) like ChatGPT are trained using billions, even trillions, of parameters. An AGI representative will probably need much more. Currently, there are no ready plans to build an AGI system.

In the new study, scientists designed Taichi to work in the same way as other light-based chips, but to scale much better than competing developments. That’s because it combines several advantages of existing photonic chips, including “optical diffraction and interference,” which are ways to manipulate light in a component.

To test the design, the researchers assembled several Taichi chips and compared their architecture with other light-based chips in key areas.

• Their architecture achieved a network scale of 13.96 million artificial neurons (compared to 1.47 million in the next largest competing design), with a power efficiency figure of 160.82 trillion operations per watt (TOPS/W).
• The next best result they highlighted in their paper came from a study published in 2022, where the photonic chip achieved 2.9 TOPS/W.
• Many traditional neural processors (NPUs) and other chips achieve well below 10 TOPS/W.

The researchers also claim that their Taichi-based architecture is twice as powerful as other photonic systems, but they do not provide direct test results. In tests, they used the Taichi distributed network to perform tasks including image categorization and classification, as well as graphical content generation, as a proof of concept rather than a performance comparison.

Taichi points to the great potential of photonic computing on a chip to perform a variety of complex tasks with large network models, enabling real-life applications of optical computing.
– engineers say.

They expect Taichi to “accelerate the development of more powerful optical solutions as a baseline model and a critical support for the new AGI era.”

If this technology turns out to be feasible, China could gain a huge advantage over Western AI developments.