Using a prototype chip that looks like a brain, IBM, one of the world's most respected technology giants, says it may be possible to increase the efficiency of artificial intelligence (AI) by enhancing energy efficiency. The advancement addresses the challenges related to the emissions emission associated with artificial intelligence systems that require expansive buildings to store electricity.
In the realm of artificial intelligence (AI), IBM claims that its prototype chip, which makes AI more efficient, is likely to revolutionize things. There are several components found in IBM's chip, including ones that are designed in a similar way to the connections found in the brain, which results in a more energy-efficient chip and a shorter battery life for Artificial Intelligence systems.
According to Thanos Vasilopoulos, who is a scientist stationed at IBM's research lab in Zurich, Switzerland, the brain's ability to carry out intense tasks at a low energy consumption is partly responsible for the exceptional energy efficiency of computer systems.
According to Apple, this technological breakthrough could lead to the development of more efficient smartphone AI chips. By doing so, large and complex workloads can be executed in environments that are low-power or battery-constrained, such as cars, mobile phones, and video cameras, using AI technology.
Several components in IBM's chip differ from digital chips popular in the past, with which information is stored as 0s and 1s, as opposed to memristors which are analog components that can store a wide range of numbers in an analog format. By using memristors, the chip is capable of mimicking the way synapses in the brain work, which allows it to "remember" how it got its electricity from year to year. A brain-like technology could provide the building blocks for the development of networks that are similar to biological brains by using this type of technology.
Analog to Digital Conversion
A major flaw of most chips is that they are primarily digital, meaning they store data as 0s and 1s, whereas the new chips are analog, meaning they can keep a range of numbers using components called memristors.
A digital switch is an electrical switch that can be compared to an analog switch, which is an electrical switch that sees a different light when you flip the switch. The nature of the human brain is analog, and the structure of memristors is similar to that of synapses in the brain, which are analogous to each other.
Ferrante Neri, a professor of physics at the University of Surrey, explains that the use of memristors falls into the special category of what could be called nature-inspired computing, as it mimics the functions of the human brain. Memory cells play an important role in storing information about the electrical history of a biological system, in the same way, that a synapse in a biological system can store information about the electrical history of that system.
The memristor, similar to a synapse in a biological system, comes with the ability to "remember" its electrical history within the circuit board. Essentially, he said, there could be a system of memristors that would look like a biological brain if the devices were interconnected.
Despite this, he cautioned that developing a computer with memristor technology is not a simple proposition and that there will be several challenges ahead before memristor technology becomes widely adopted, such as dealing with manufacturing difficulties and rising costs of materials.
Improved Energy Efficiency
Using these components makes it possible for the new chip to run more efficiently and be more energy efficient while also having some digital components. It makes the chip easier to integrate into a system that already uses artificial intelligence. Nowadays, many phones come with onboard AI chips for them to be able to perform tasks such as processing photos. Taking the iPhone for instance, it has a chip with a neural engine that makes it make intelligent decisions.
IBM hopes to improve the efficiency of the chips in phones and cars so they can have a longer battery life and be capable of supporting new applications in the future. Eventually, it is possible that chips such as IBM's prototype could save a great deal of electricity if they were replaced with chips that are currently being used in the banks of computers that operate powerful artificial intelligence programs.
James Davenport, an IT professor at the University of Bath, has said the findings from IBM are "potentially interesting"; however, he cautions that the chip is not immediately effective as a solution, but rather only acts as a "possible first step" in solving the problem.
In a similar way to how the brain stores information on synapses in a wide range of peripheral nerve cells, this analog marvel uses memristors to store an immense amount of data. This chip due to its low-power and analog nature is not only more energy efficient than other chips on the market, but it also makes it possible for AI to be integrated into low-power environments such as mobile phones and vehicles.
It is important to note that while there are still challenges ahead, researchers have marked a significant step forward toward a more efficient and greener future of artificial intelligence.
Despite not being a solution to the problem of AI energy consumption, it is a vital first step that could be taken to address the ever-evolving challenges associated with it.
In the future, users will be interested to see how this 'Brain-Like' chip will impact AI ecosystems and sustainability, as it is fascinating to see how it unfolds even at this early stage of development.
