Hacker News Reader

Comment by RcouF1uZ4gsC | Hacker News Reader

> Desai, through the spokesperson, also said of the company’s work with patient data: “We use a great deal of artificial intelligence and machine learning to automate this process as much as possible, which is the only way a task like this is even possible.”

I wonder if we are heading for an AI winter soon. AI has been heavily hyped, but is having a lot of real world failures and is becoming associated with a bunch of fraud. At this time, I am less inclined and I believe a company if they invoke “AI” or “machine learning”.

Sure, businesses will still use machine learning for problems where it makes sense, but the hype of AI will die down and funding will dry up.

View on HN

You can't just "use AI" to solve record linkage and dataset cleanup; we deal with just data from the US and even then it's a messy problem; cannot imagine how much messier it can be for multinational datasets. There are many companies (including where I work now) which have a lot of data in this regard but you don't hear too much from them because doing any analyses with these large-scale messy datasets is a hard unsolved problem.

Interestingly, there is this article in the same publication: "Core progress in AI has stalled in some fields"

https://science.sciencemag.org/content/368/6494/927

>. AI has been heavily hyped, but is having a lot of real world failures and is becoming associated with a bunch of fraud

AI, specifically deep learning is far more than self driving cars. The industry is in its infancy and we are building ML solutions for a multitude of industries. You don't hear about any of it because none of it is sexy (or even accessable) to the layman - but rest assured, the industry is growing exponentially (like the research) and we are finding much more in the way of success than the high profile failures of self driving.

Some portion of AI companies are likely a bunch of lawyers and bullshit. Whenever someone talks about all the companies jumping on some software hype train, I think about this post written on HN some time back [0], and I imagine that at least half of the buisnesses on that hype train fit that description.

0. https://news.ycombinator.com/item?id=19169151

I think AI/Machine learning fits much better at the big big companies like google.

You only need one or 2-3 competiting Neuronal Networks for cancer detection and if you look how expensive it is to train that stuff, not many can actually do it anyway.

But what tripples down are pre learned networks you can easily reuse for your small use cases which are specific and have plenty of data and only need to outperform the existing solution.

>You only need one or 2-3 competiting Neuronal Networks for cancer detection and if you look how expensive it is to train that stuff, not many can actually do it anyway.

This is just wrong. You can train a cancer discriminator on a low end gpu on your desktop. The hard part is getting quality [often annotated] data - and thats why data brokers like Google are positioned to dominate.

So why are those papers with good results are not running on 'low end gpus'?

I don't mean running them, i mean training them.

Which papers? There are thousands of papers being published.

There are a couple potential reasons. Powerful GPUs accelerate research and iteration. Some state of the art problems have hit the limits of current theory and make up the deficit by building massive nets - but even there we already have multiple automatic pruning/optimization algorithms to shrink those nets so that they work with smaller resources.

Make no mistake, the field is advancing exponentially. The state of the art googlenet/inception that arguably kicked off the whole craze with image recognition are laughably obsolete now and easily outperformed by simpler nets.

MNIST was the gold standard for recognition problems just a couple years ago, and now it's considered a solved toy problem.

If i google for it specificly, the paper in nature states:

"This study had some limitations. Mammograms were downsized to fit the available GPU (8 GB). As more GPU memory becomes available, future studies will be able to train models using larger image sizes, or retain the original image resolution without the need for downsizing. Retaining the full resolution of modern digital mammography images will provide finer details of the ROIs and likely improve performance." https://www.nature.com/articles/s41598-019-48995-4

Here they use a Nvidia V100 https://www.researchgate.net/publication/336339974_Deep_Neur...

Which yeah okay is more reasonable than i thought. But the advantage will still be at who ever has the hardware and thats just cheap for google.

You wouldn't need a market for models, you would just use whatever research delivers from whoever has the most accurate data & hardware.

I expect "Blame it on AI" to become a thing, before AI winter happens.

Reply on news.ycombinator.com