lustig

I started out as a Software Engineer, then switched to Data Scientist. However I have decided to switch back to Software Engineer again now, simply because I enjoy the work more. So I would advise to try get into the nitty gritty work of both, preferably through internships, and then decide which route you prefer.

I would also suggest to apply for a lot of them anyway, there's not enough skilled and experienced people to fill all positions right now, so some of them will recruit more junior people than they might be looking for at first.

Totally agree and I would argue that there even is financial gain for society in letting people take such risks. Maybe I was kicking at an open door, but my point was that there, in my opinion, does not exist a huge incentive to misuse the welfare system voluntarily.

Not sure I'm reading you right, but here are my two cents.

Also living in Scandinavia and I would disagree with this advice. To be dependent on the welfare system is not exactly being free and have security.

Like the Swedish Prime Minister Göran Persson said: "One who is in debt is not free".

And living off of the welfare system is in my view to be indebted to society. I would not be able to shake that feeling if I was, at least not if it was voluntarily. Add to that the social stigma, even from close friends and family. This would limit your freedom of having an agency in social interactions.

Also I would not consider it secure, since the rules of the welfare system changes quite a lot over time. Which you have no way of impacting, meaning you are dependent and not free.

EDIT: Just realized the question was about "security+free time". I guess you would have free time, which is not exactly the same as being free in any meaningful sense.

EDIT: Not sure why I'm being downvoted. But for clarification I can add that I think this applies if you voluntarily would live off of the welfare system, thus leeching from the ones who really needs it. It's of course a totally different story if you are involuntarily need to get welfare to survive and live a decent life.

This was funny! I recognize these conflicting views. The first is the view that math is a fundamental force governing nature. The second is that math is simply a way to describe the forces of nature. I tend to agree with the second view and are often dragged into arguments which has the roots in this difference of viewpoints.

Thanks! I think 30-70 visits per day without paying for them is quite a lot after six weeks. I'm in a similar situation, working a day job while trying to get something off the ground. Right now I'm just experimenting with ideas and trying to learn as much as possible.

I think paid ads is a pretty good approach to validate the product and copy, even if it's not a viable strategy in the long run.

Making it free for academic use makes sense in more than that sense. If you're lucky you might get a .edu backlink from it, which is valuable for seo.

How are you marketing the product?

Ah I see! Looks good!

Also, I would consider increasing the price quite a lot. If this really increases sales people are willing to pay a lot more. That would also mean that you can take the effort to do direct sales, whereas now the ROI for manually aquiring a customer is too low. Good luck!

Cool! I think this might be a great product idea!

I think the messaging is very clear, but would have wanted to see an example of how the button actually works/looks like on the front page.

The landing page looks great! I've seen a few similar ones recently (with the lines that move etc). How did you create it?

So they are basically creating that creepy life recording product from S01E03 of black mirror?

Hi David! Thanks for a great summary! I will keep an eye on your blog for notes from upcoming conferences.

This is true! Overfitting is definitely one of the biggest problems with deep learning. Some techniques to avoid it have been developed, such as dropout (introducing noise) and early stopping. But in general this is why deep learning requires huge amount of data, a deep learning model will overfit if not given enough data. That is also why (at this time) it only performs well for certain problems where the ratio between available data and problem complexity is high enough.

Technically yes, most often it's about stacking more layers in neural networks, making them "deep". However, there is some merit to the new hype since stacking more layers worked way better than anyone previously working with neural networks and ML thought it would. But in theory you could generalize deep learning to other methods than neural networks, it's basically about creating way more complex models than those used in previous research and feeding them lots of data. Thereby assuming less about the problem and letting the model figure it out.

We moved from slack to hipchat at my old job because it was cheaper. Hipchat is just a terrible product, both on iOS and in the browser. Don't even know where to begin, it was just a pain to use and every employee I talked to hated it. I'll just give a few examples:

- Unreliable notifications. Sometimes it notified you about messages sometimes it didn't, no idea why.

- You could get a push-notification about a message, then open the app and have no clue who had messaged you or in which discussion/group.

- Emojis had really unintuitive shortcuts. I remember people used to send some kind of skeleton dancing all the time, because it's shortcut was a (Y) (which is commonly some affirmative/"yes" emoji like a thumbs up).

- Also it looks horrible.

I agree, the numerical results are poorly presented.

What was measured was wether their tests can find a dominant eye in the subject or not as well as the difference in Maxwell's centroids between both eyes.

Eye dominance results:

They had two types of test for this; the sighting test and the after-image test.

For the control group they found a dominant eye in 28 of the sighting tests and 30 in the after-image tests. In the 28 cases where they found a dominant eye with both tests, the tests corresponded perfectly, i.e. if the sighting test indicated the right eye was dominant the after-imagetest also did.

For the dyslexic group the sighting test found a dominant eye for 14 subjects, and the after-image test for 3. In the 3 cases where they found a dominant eye with both tests, the results corresponded perfectly.

Maxwell's centroids result:

For the control group they showed that in 29 cases the asymmetry between the centroids was at least above 0.3 (as far as I can tell from the figures), where 0.3 means "weak asymmetry" and 0.6 means "strong asymmetry". One case was slightly above 0.2.

For the dyslexic group 27 cases had ~0.0 in the asymmetry measure. 2 above 0.3 and one slightly below 0.3.

I have generalized a bit and labeled results ranging from 0.3 to 0.5 as "above 0.3".