Greenpants

I still haven't found "that one visualisation" that makes the attention concept in Transformers as easily understood as these CNNs.

If someone here on HN has a link to a page that has helped them get to the Eureka-point of fully grasping attention layers, feel free to share!

Much like the author, I'm also really loving the Steam Deck as someone who casually plays a game or two. It really is like a more powerful Switch with the option for much more, and cheaper, games.

What I'm missing in many reviews that is gladly mentioned here is the ability to use the right trackpad as a mouse. I've also played many older games in handheld, like Impossible Creatures, which are pretty much impossible if you aren't emulating a mouse. The Steam Deck delivers in this sense, which very few competing manufacturers seem to realise.

As happy as the author is with the sleep & wake functionality however, as someone who only wakes the Steam Deck once every three days or so, it is unfortunately pretty common to wake my Steam Deck to an almost empty battery.

I, for one, often glance at the comments first to check if an article is valid. If not, there's probably a highly rated comment explaining what's wrong with an article's facts or reasoning.

Saves me not just a click, but lots of time reading lesser quality articles too.

I recently distributed laptops to politicians and the first step in the setup process is choosing the keyboard layout. Nearly half of the people were already selecting "Dutch" before I could tell them we don't actually use that for the keyboard layout. We Dutch people are all using the US (international with dead keys) keyboard layout nowadays. There's no need for any special Dutch keys.

This is indeed a good point to keep in mind. I do tend to forget sometimes. As a perfectionist I'm always planning code refactors in my mind, but rarely do they see the light of day; they're often minimal and don't add anything to the solution. There's very little added value.

Know your audience – know what they need, and you can develop accordingly.

Google is checking through Opinion Rewards whether users can watch a certain advertisement video. On Android I'm using the Adguard DNS and their ad video wouldn't load twice. This occurred last week and fits this article to a tee.

I'm sure I'll get a new advertisement video soon which will load despite my Adguard DNS. That's how Google can confirm whether their ad-blocking-blocking works.

Opinion Rewards is great not just for being able to get apps for 'free', but also to be one of the first to see what Google is researching.

The dream-like appearance of AI generations is really interesting. Humans, when creating art, use our representations for objects to control muscles to move a brush or pencil across (digital) paper. What we see from AI is what you get when you remove the "muscle module", and directly apply the representations onto the paper. There's no considering of how to fill in a pixel; there's just a filling of the pixel directly from the latent space.

It's intriguing. Also makes me wonder if we need to add a module in between the representational output and the pixel output. Something that mimics how we actually use a brush. How we consider what to paint where.

I used to give Apple the ole' eye roll for that as well. Then I realised, as I got a MacBook myself and dove into running Machine Learning models on it, the RAM setup is pretty unique.

Essentially, the RAM is so close to the CPU and GPU that it can effectively be used as VRAM, at least for the M1 and up chipsets as far as I'm aware. That means a 32GB RAM MacBook would be able to run incredibly large (e.g. LLM) networks on-device. Nvidia GPUs with that much VRAM (although they are clearly better at GPU tasks) can cost as much as an expensive MacBook already.

There's a lot of research done with drones but the world isn't quite ready yet. Realistically, drones delivering packages will get shot down in the US, simply for being unmanned and flying over people's neighbourhoods. People easily feel threatened.

Perhaps more importantly, there's too much that could go wrong. What about legislation where they may fly and how high? What if a drone crashes into someone or something? What if someone's package gets stolen?

As much as the technology enthusiasts in us enjoy the concept of delivery drones, most of us humans still prefer a fellow human in the process of delivering packages to handle edge cases where things might go wrong.