How to Compete With the Industry in the Natural Language Processing Space

I think definitely there's a lot of people wondering in academia how to compete with the current industry efforts. So what is your take? I mean, are you going to go to industry to train larger models? Do you think there are things to do in academia that are even more exciting? I don't know. Probably I will fail, but I'd rather try than going with trendy bandwag on. And so long as they implemented the algorithm correctly, it does tend to work right away.