psv1

> The big tech companies’ only advantage is that they can manufacture more complex tech.

Only advantage? What about the already established user base, decades of experience, brand recognition, and giant piles of money? And I'm sure that I'm leaving out other important factors.

> Their goal is to move what users want towards things that are hard to build

Is there any evidence for this? Obviously there will be some incentive for this but that's not the same as being their goal.

Flexible and foldable products are so underwhelming. At this point they seem like something being developed just for the sake of demonstrating that it's possible. If the goal is to make attractive products, they're failing spectacularly.

You should really compare it directly to your other options. I recently moved to London from the north of England and was also somewhat reluctant at first exactly because of the higher cost and possibly lower quality of life.

> I’d wait until the dust has settled from Brexit

Even in the best case scenario it will be years before you can consider Brexit settled in any way.

> some standardised method, assessing working memory, verbal fluidity

> (I'm not talking about IQ obviously.)

Why not IQ? It's the most well-researched, standardised and validated test for cognition that we have. It's trendy to criticise IQ and IQ tests but any alternative will be supported by even less evidence, far less in fact.

And even if they accomplished it, they wouldn't come even remotely close to being as financially successful with the same product.

That's one of those things that make for a nice Medium article, but in practice work horribly or not at all.

I agree, he's really difficult to listen to. It's surprising how someone with communication skills as bad as his managed to get quite a few great guests.

If you're:

- working interactively (i.e. your code isn't part of a larger application)

- working with relatively small datasets that fit into memory

- don't need any deep learning libraries

then both R and Python can do a great job and choosing one over the other is simply a matter of preference. I might even lean slightly towards R because its data frames are a bit easier to use than pandas and RStudio's REPL is the best.

But if you need to deploy your code somewhere, or high performance, or the latest deep learning libraries, then Python absolutely crushes R. And it's not even close.

The elastic library on CRAN got a major update a few months ago which made it much more powerful and easier to use. Makes uploading easy too, especially if there's a direct relationship between the structure of your dataframe and the Elastic json. Uploading nested structures is still a bit of a pain though.

By data it hasn't seen before I mean a sample which doesn't exist in the training set. Maybe I'm not understanding something in the parent's argument.

For example, you train an image recognition model to tell cats and dogs apart using images from the internet, then you take out your phone, snap an image of your dog and give it to your algorithm to determine the species. When preprocessed this picture is equivalent to one row in a table with thousands of columns and this specific combination of pixel values doesn't exist anywhere else. Where is the hash function looking and for what in that case?

> You can always verify why it made the choices it did on old data.

This makes no sense. Your model typically isn't asked to make predictions on "old data" because it was trained on it. And understanding why it made certain prediction isn't any different on "new" vs old data.

You can look into bias-variance tradeoff, train-test splitting and cross validation to get a better picture of this.

> purpose of black-box models is to be impossible to understand

This is just outright wrong. And the argument about moderation is completely unrelated.

The whole point is that your models should perform well on data that they haven't seen before. You can't create a hash function for that. Let alone one that works for 100s or 1000s of dimensions.

> non-standard interpreters like pypy and cpython

Isn't CPython the default and most standard interpreter?

This problem is coming from the way providers choose to deliver content, the way they design their data products and the metrics they optimise for. NLP is simply the group of algorithms that power this process, not the underlying cause.

> apps being built in collaboration with AI

Oh please no, that's just so unnecessary.

None. That person had plenty of available advice that he didn't listen to. What he needed was better role models and people to emulate, not advice.

You can try this for yourself with zero downsides - leave it at home for a day and see how it goes. If it's not that bad, leave it at home the next day too and then don't use in the evenings. Or try to delete some apps and keep the phone with you.

You don't need a drastic "I'm deleting all my accounts and ditching my phone" intervention. And in my experience, in the long run big sweeping changes are far less effective than incremental changes to your habits.

> If you like books and you want to deeply understand ML techniques I'd suggest jumping straight into "Introduction to Statistical Learning" and only learning calculus/stats/matrix methods (linear algebra) as you need them (you really don't need much from them in practice).

This doesn't work. ISL is good, but it aims to be accessible by excluding most of the math. So if you go over it, you'll neither "deeply understand ML techniques", nor will you encounter enough math that you can learn along the way as you suggest.

You're right about ESL, that's why I started the list with some more fundamental material. Also, +1 for Aurelien's book, it's really good; I didn't know he had a revised edition for TensorFlow 2.

Good free resources:

- MIT: Big Picture of Calculus

- Harvard: Stats 110

- MIT: Matrix Methods in Data Analysis, Signal Processing, and Machine Learning

If any of these seem too difficult - Khan Academy Precalculus (they also have Linear Algebra and Calculus material).

This gives you a math foundation. Some books more specific to ML:

- Foundations of Data Science - Blum et al.

- Elements of Statistical Learning - Hastie et al. The simpler version of this book - Introduction to Statistical Learning - also has a free companion course on Stanford's website.

- Machine Learning: A Probabilistic Perspective - Murphy

That's a lot of material to cover. And at some point you should start experimenting and building things yourself of course. If you'are already familiar with Python, the Data Science Handbook (Jake Vanderplas) is a good guide through the ecosystem of libraries that you would commonly use.

Things I don't recommend - Fast.ai, Goodfellow's Deep Learning Book, Bishop's Pattern Recognition and ML book, Andrew Ng's ML course, Coursera, Udacity, Udemy, Kaggle.

Yes, I have, and my experience has been the exact opposite. Their courses on calculus and linear algebra are some of the best that I've come across.

You have one of the best universities in the world sharing their material for free. That's what's valuable, not the forums that can be attached to the platform or the certificate that no employer cares about.

Are you switching because you want to or are you forced to switch and take a pay-cut? Either way you shouldn't put up with it but knowing that detail can paint a very different picture.

Their ad model is what gets discussed every time Brave comes up and it's irrelevant to me - I block all ads myself regardless of the browser.

My point is that if you switch ads off in Brave, you end up with a really good browsing experience, better than Chrome + adblock or Firefox + adblock.

For people who are considering switching away from Chrome - give Brave a try, it's really good.

I was a bit reluctant for a while because their branding is weird and there's too much focus on their crypto ad model. But if you switch the crypto stuff off, you have a fork of Chrome that performs just as well and also has a few additional privacy features and isn't linked to your Google account. Highly recommend.

Just realised that I don't know or use any languages that can be considered non-mainstream. Although Racket has been on my to-do list for way too long now.

Federer says yellow - https://www.youtube.com/watch?v=1at7aQqjhC8

The advice here is good but it only confirms for me that I would much rather commute to and from an office every day than work from home.