Comment by jasode | Hacker News Reader

Comment by jasode | Hacker News Reader

>Almost every recent A.I. advance has come from one tiny corner of the field, machine learning. Machine learning exposes a set of connected nodes, known as neural nets, to mass amounts of labeled real-world data in an attempt to give those nodes tacit knowledge. The breakthrough example was software that was able to identify cat pictures.

>So far, these neural nets have given us some great demos but mostly niche real-world applications. We don't have self-driving cars quite yet!

What's the author's threshold for "real-world applications"?

- Google's Youtube algorithm for recommendations uses neural nets[1]. So ~2 billion viewers being affected by it doesn't seem like a "niche" application.

- Google language translation uses neural net[2]

- Apple Siri voice recognition uses neural net

It doesn't seem like neural nets are analogous to the joke that "graphene is the wonder material that can do everything except escape the research lab".

In contrast, deep learning neural nets have escaped the research lab and are widely used in production systems today.

The author's blog post is recently dated August 2021 so it seems like he's not kept up-to-date on this topic since the experimental neural net winning ImageNet in 2012. Yes, that was an artificial contest but things have progressed quickly and there is real-world commercial deployment of NN trained models.

[1] https://static.googleusercontent.com/media/research.google.c...

[2] https://smerity.com/articles/2016/google_nmt_arch.html

The author didn't really mean "real world" here. That is a misnomer. The article is about explaining why productivity gains have slowed even in the face of better software techniques. Machine learning has possibly made YouTube more addictive than it otherwise would have been, but this isn't increasing aggregate economic output for the world. It's just concentrating ad revenue in Google accounts, where it used to be spread between many more content distribution platforms.

> but this isn't increasing aggregate economic output for the world

How is it not? Demand is increased as people want to consume more content, and supply has risen to match it. We have more creative & interesting content by more creators than ever, that's certainly a huge increase in overall economic output by any measure.

If you want to make some separate point about how "this isn't good for society" then make it, the economic productivity benefits of YT are huge regardless.

I think you'd have to consider the productivity benefits of youtube relative to the productivity benefits of anything else. For instance, a customer goes on youtube for three hours and sees some advertisements or whatever. Pays nothing. Another customer goes to a local restaurant with some friends for three hours. Spends $60 on food and drinks. Of the two, what is better for the economy? I would wager spending disposable income within the local economy is better than watching advertisements for companies registered in Dublin or wherever has the lowest taxes.

I think the author wording is a little bit confusing. I think he want to point out, that narrow AI is used successfully only in (niche) applications with narrow purpose. Niche here is not about total market size.

Not to mention Google search uses neural networks for language parsing [1], social networks use neural networks for ranking feeds, etc. I'd venture to guess the author interacts with the results of neural networks dozens if not hundreds of times per day but simply doesn't realize it.

[1] https://searchengineland.com/google-bert-used-on-almost-ever...

Reply on news.ycombinator.com