jordancampbell

Ha I actually think the complete opposite will be true - Here in New Zealand I got a degree in neuroscience without doing any papers outside of core science, whereas in the US I would probably have to do humanities based courses etc (which I think is a fairly good thing) - and it's not unlikely that these will come to include topics in ethics.

Also ethical review boards for other areas of science are very well established, and it's not unrealistic to imagine that extending towards machine learning as well.

This is completely untrue - machine learning is largely undergraduate mathematics - in fact a lot of the linear algebra is commonly taught in high school.

Definitely agree it's potentially narrow, but there's absolutely nothing wrong with that.

And just to clarfiy - it doesn't collect any data that you don't allow.

It's also really easy to delete any (and all) associated data after an account has been created.

Thanks for your feedback. You're told what 'begin' does when you click begin.

The website is simply a demonstration of how face login can be simple and unobtrusive - it's not actually part of the app and is going to be changed to reflect the rest of the application.

I write all my CLIs in Python.

I find that I always need various random things to work together and have always found Python to be the most robust at this. Plus it has pretty much anything you'd need already implemented.

This is cool - well done to you.

Even more than the money, just having someone tell them that their idea isn't terrible will be a great help.

There will be people for whom this literally decides whether or not they start / keep working on a side project - it's a great idea.

My gf has lots of problems with various foods and manages to keep on top of it fairly well - although she has very little concrete evidence about what it actually is that is causing the problems.

I could see her trialling a comprehensive ($30 a week) package for a few weeks, then dropping to a less intensive (~$10 a week) package that kept her on the right track.

I personally don't think she would pay $30 a week unless the service gave her a significant material difference in lifestyle (which I certainly imagine it could have the potential to do).

Deep learning is easy and popular - it's just as good a starting point as anything else. No one would argue that linear regression isn't a good starting point, but only because it's not fashionable to dismiss linear regression.

I'd highly recommend course.fast.ai - it's focussed on deep learning, but is designed for beginners to get to a production stage.

New Zealand is geographically quite large (about the length of France and Spain joined together, and a little bit less wide) and we have half the population of London!

It's literally a Ponzi scheme.

Where is the money coming from, other than speculation?

It barely even hides this fact, given that all its content is about how to generate more content and keep all the wheels turning.

It's inferring the depth of all the points on the surface, including those it can't see in the image. This isn't surface rendering, it's depth estimation.

No, inputs to each layer are concatenated before being passed through H.

This is crazy.

I agree - although I think there are three options:

1. Deep learning - therefore any deep learning framework 2. "Classical AI" - use OpenCV for most of it 3. APIs as you mentioned above

I second this - the two golden startup fallacies: 1) someone will steal my idea, and 2) I'm going to be overburdened with users.

Perhaps use a signup service that allows you to select which users can try your service.

I thought it was satire.

Words cannot describe how much I disagree with this article.

In general you don't actually need crazy precision to train the nets, and a small number of Taylor expansion terms tends to approximate functions fairly well anyway.

I was once told by a former Seattle based executive that the best way for a startup to price their product was to take what they thought it should be worth and then add a couple of zeros to the end.

There's a school of thought that says that you should only achieve a certain percentage of sales. So if you're getting 90% of deals then you might be pricing too low, and if you're not getting anything then obviously too high. But if you're making 10% of sales and your selling for 100x more than your guess at the price then you're likely doing well.

I'd also be really interested to learn how you generate synthetic data. Is it a 'simple' case of manually generating as much as possible and then using the statistics of this to bulk it out? Or are you using something more complex to augment it?

If you do go to University then most 'analytical' subjects (maths / computer science / engineering) will prepare you well for future work.

The most important thing to do is keep asking questions like you are now. Contact architects, engineers, city planners, entrepreneurs etc and ask them what problems keep them up at night.

Yep!

Learning algorithms typically find parameters that relate the input to some output.

Genetic algorithms are just a collection of methods that describe how to search through parameter space.

Supervised learning algorithms are ones for which we have some known labels on our inputs (known outputs), whereas in unsupervised learning we don't have any known outputs. Regardless, in both cases we need to learn the parameters that relate the input to the output.

Genetic algorithms can therefore be both supervised or unsupervised.

I'm doing a PhD in computer science and use a notebook religiously.

I couldn't imagine life without it.