Hacker News Reader

Because deep learning is expensive compared to simpler methods which are sufficient for business needs.

i.e. the newest or shiniest thing is not always the best choice for the business. Or it's not about the tech, it's about the value creation

For more examples, see blockchain

View on HN

Curious, what about deep learning is the expensive part? The developer time-cost for implementations? Time spent creating the model, or data collection + cleaning, or (owning/renting) the hardware? Because I have a sense that these have become a lot cheaper in the last few years (ok, maybe not the developer time-cost)

Yes, I think it has been lot cheaper and and more easier to train state of the art models such as fast ai.

Computationally cheaper, which leads to cost savings in many places for myself and my end users.

What framework are you using?

Reply on news.ycombinator.com