As Intel continues to weather some of the worst news in its history over security issues related to its chips, the company today also announced some developments related to what it hopes will be the next generation of super-fast processors.
The company announced that Loihi, its first “neuromorphic” chip — designed to mimic the way that a human brain learns and understands — is now “fully functioning” and will be shared with research partners this year.
And the company’s 49-qubit Tangle Lake chip — designed for quantum computing processes — is also being shipped with a research partner, QuTech, for further development.
First announced in September, Loihi is Intel’s effort in the fast-developing world of AI chips, an area where Nvidia and startups like Graphcore are also attempting to stake a claim. The idea is that the processes involved in AI will be more complex and require more computing power, and some of that can be moved to the chip to increase the efficiency of the system.
As with other AI systems, it “learns” over time and gets smarter, but unlike other AI systems, it doesn’t require masses of training data to learn a process. Initial applications are likely to be in robots and self-driving cars. Currently it functionality is limited to simple object recognition, CEO Brian Krzanich said today.
The chip itself is conceived as modelled on the human brain — or at least how we know it to work — with pulses and spikes based around synapses, with different parts of the chip taking on different learning functions.
While the Tangle Lake chip is solving something completely different — the quantum computing challenge of working in extremely stable environments — it’s also addressing what has remained the main objective for Intel when it comes to chipmaking: processing speed, and in the case of quantum computing, parallel processing speed, a “fundamental, scientific engineering challenge.”
Intel has said that the 49-qubit chip is about the size of a quarter and has improved thermal performance and reduced radio frequency interference, scalable interconnect for more signals to pass in and out of the chip and advanced material and design to scale for quantum integrated circuitry.
In both cases, it will be interesting to see how the recent revelations about Meltdown and Spectre vulnerabilities will impact the pace of development of these chips, and what position speed will take longer term relative to security.