Not sure why you're saying that systolic arrays never went anywhere. They're widely used for matrix operations in many linear algebra accelerators and tensor units (yes, largely in research), but they are literally the core of Google's TPU [1] and AWS EC2 Inf1 instances [2].
As someone with a very midwestern US accent, I got a much better UK score when I used my terrible Cockney accent, about 93% UK, than I did when I used my default accent, at roughly 80% US.
Interestingly, it looks like the model has quite the UK bias for words like "queen", "tea", "chips", and "pub." All words that are used pretty frequently in US English but have a clear cultural affinity to the UK. When I use sentences with these words in my casual midwestern accent, they almost always come up as 80+% UK.
[1] https://cloud.google.com/blog/products/ai-machine-learning/a...
[2] https://aws.amazon.com/blogs/aws/amazon-ecs-now-supports-ec2...