Just over a year ago I wrote a blog about the impact of machine learning (ML) algorithms to boost Formal Verification performance . The data for that blog was firsthand experience on a set of complex benchmarks. The data was amazing and quite convincing, but when I wrote that blog the reader needed to “trust” me as the data could not be shared publicly.
You may be concerned about the widely published Spectre and Meltdown vulnerabilities affecting most processors, and if your phone and computer are OK. Or more importantly, if you are designing or verifying SoCs, do you have a specter in your design? Let’s first look at what these two vulnerabilities are and how they are affecting your system.
It is commonly believed that Formal property verification is the realm of PhDs and experts with many years of experience and have the magical solution and intellect to solve complex verification problems. I see the application of Formal verification solutions solving design bugs very much like the way supercomputers and artificial intelligence have evolved to beat human intellect and eventually become mainstream. In 1997, Gary Kasparov the chess Grand Master lost to IBM’s Deep Blue supercomputer. Kasparov and other chess masters blamed the defeat on a single move made by the IBM machine, where the computer made a sacrifice that seemed to hint at its long-term strategy. Kasparov thought the move was too sophisticated for a computer and got side tracked. Indeed, fifteen years later, the designers of Deep Blue acknowledged and attributed Gary’s loss to a software bug in Deep Blue that misled him. Either way, this was a lesson that as humans, we tend to blow things way out of proportion and could sometimes make wrong assumptions.
Posted in Automation
Needless to say that Machine Learning is a very hot topic nowadays. From the software giants such as Google and Facebook, to hot companies like Snap, Waze and Uber, to traditional businesses such as IBM, ExxonMobil and Toyota to name a few, it seems like all companies need to be talking about Machine Learning. I would not be surprised if every company in the S&P 100 has a list of active Machine Learning projects…or so they say.
Posted in Automation