I have an iPhone 11 Pro. One of the lenses is broken and it makes a weird snaring sound while recording video. Battery life is still okay at about 80% capacity. It does everything I need it to do without fail.
If it weren’t for those defects I’d keep it at least for another year.
Real ironic was the day the LA (Los Angeles) DWP (Department of Water and Power) declared drought water restrictions on a day we got so much rain we almost had Lake San Fernando Valley. The Sepulveda Dam Recreation area did, indeed, flood. One of those "why do we have all these dams with no water in LA" moments.
Germany has no wilderness. "Nature kills you" is a concept that basically doesn't exist in Germany. As a healthy adult, "getting lost" is a minor annoyance, not genuine danger.
See for yourself: Open Google Maps, zoom in on a random spot, pick a random direction, and tell me how far you have to walk until you'll find a pub or a road that will lead you to one. Repeat as often as you want.
The first spot I got was 60 meters from the nearest forest path, the longest you could walk in a straight line without finding a forest path was 400 meters, and the nearest pub was 500 meters as the crow flies.
The second was much worse: 80 meters, 800 meters, 1.6 km for the same three metrics. Third: 10m, 230m, 1.2km. 4th: 140m, 800m, 600m 5th: 55m, 380m, 2.2km unless you count the Kebap place 1.8km away. You will be hearing the Autobahn the whole time.
The weather is also much less extreme. 41° C (106° F) seems to be the current heat record (https://en.wikipedia.org/wiki/List_of_extreme_temperatures_i...), but this is extreme, newsworthy heat, not common. Most of the time in most of Germany you can expect -10 to +35° C (from nighttime in Winter to daytime in Summer).
In summer, basically, t-shirt, shorts with 10 EUR in cash for the beer in a pocket, and sandals with socks will probably be all you need to survive. Everything else (water bottle, a way to navigate) are optional comfort items.
Somewhat related, it's hard to find things like Gatorade powder. "Rehydrating" on a serious mountain hike, the kind with 10,000ft peaks means stopping by at the nearest alpine hut, and drinking a liter or so of beer. I grew up (to age 13) in Germany but on adult visits, felt like a fish out of water, with my sandwiches and water bottles.
It doesn't get Death Valley hot (few places do), and Germany has no deserts. Combine that with the German fascination with the Old West and you've got a recipe for disaster. They see the terrain, figure they can handle it because they go hiking in mountains all the time. They can handle the terrain; what they can't handle is the environment.
That's a wasteland, but probably not a desert. Per the link, it gets 569 mm of rain a year; dry, certainly compared to most of Germany, but most definitions of "desert" run around 250 mm/yr (consensus, but even if the cutoff is off by a little bit, it's over twice that). By comparison, Denver, which is arid but not a desert by anyone's imagination, is around 368 mm/yr.
I'm making this a separate reply because I just realized that this is one of those language confusion things. If Germans think a desert is a place without plants, regardless of water resources, they will wander out into places that Americans and Australians call deserts, which are insanely dry, thinking that of course there must be water somewhere. In the US and Aus, if someone says "that's a desert", you should be carrying 10+ liters per day per person and some extra if you get into trouble. Do not assume that you will find water anywhere.
This is highly inappropriate for a job interview at least in my country. You can’t ask these type of personal questions. Who would feel comfortable sharing this with a complete stranger anyway?
Token embeddings typically only attend to past token embeddings, not future ones.
The reason is to enable significant parallelism during training: a large chunk of text goes through the transformer in a single pass, and its weights are optimized to make its output look like its input shifted by one token (ie. the transformer converts each input token to a predicted next token). However, if the attention weights attended to future tokens, they would strongly use the next token they are given, to predict that next token. So all future tokens are masked out.
