I would also call the news plus, sports, and audio tabs in the News app advertising as they cannot be disabled. Likewise it defaults to the Apple TV+ tab in the TV app on launch and adds Apple TV+ spam to the Home tab even with no subscription. The Stocks app has News+ spam starting this year. App Store is full of unrelated “promoted” apps. Accidentally hitting the “play” button on a computer prompts for Music subscription.
If Apple had any self respect at all these would be one time per year at best prompts or something which could be disabled with a subscription. I would no longer recommend Apple as the go to for respecting their customers.
Yes but that's really the same as this just in a separate box. Makes total sense to bring it under the same naming tree.
I'd call it "Google TV Box" though. Streaming is too contrived and not everyone knows what it means. Xiaomi use the Box naming too and that seems to go down well.
which makes sense. you can pretty easily imagine the problem of "selecting the next token" as a tree of states, with actions transitioning from one to another, just like a game. And you already have naive scores for each of the states (the logits for the tokens).
It's not hard to imagine applying well-known tree searching strategies, like monte-carlo tree search, minimax, etc. Or, in the case of Q*, maybe creating another (smaller) action/value model that guides the progress of the LLM.
Absolutely, maximizing conditional probabilities is easily modeled as a Markov decision process, which is why you can use RL to train Transformers so well (hence RLHF, I've also been experimenting with RL based training for Transformers for other applications - it's promising!). Using a transformer as a model for RL to try to choose tokens to maximize overall likelihood given immediate conditional likelihood estimation is something that I imagine many people experimented with, but I can see it being tricky enough for OpenAI to be the only ones to pull it off.
IME that's been common or the norm in plenty of circles for more than 25 years (so probably longer). It became less common when people started phone-posting and when phones started auto-capitalizing by default.
it is inconsistent in language usage to write differently than to speak. we don’t speak big sounds, that’s why we don't write them either. and: doesn’t one say the same thing with one alphabet as with two alphabets? why does one merge two alphabets of completely different characters into one word or sentence and thereby make the written image inharmonic? either large or small. the large alphabet is illegible in the typesetting. therefore the small alphabet. and: if we think of the typewriter, the limitation to lower case characters means great relief and is time saving. and if we think further, it would be simplified by switching off upper case characters.
To be fair, we're on a geek website. Not many have default settings on any of their devices.
It think the above is consequence of a "I can afford to write all lowercase/do unconventional thing X". And by "afford" I think here more like "I don't have bosses, nor do I have to please anyone doing conventional things".
There was an article or a discussion here a while ago how in a organizational pyramid the people on the botton usually write as normal/nice as possible, while going upwards; people can afford themselves to write however the like, including being super rude if they choose so.
This guy independently tested a Flip 3, getting authentic humans to perform the folds rather than an automated test rig, and the hinge was the first component to give up after 418,500 cycles with the display itself surviving.
The 200,000 figure isn't super meaningful unless they report average folds to failure and standard deviation.
From that article, it could be that they test to 200,000 folds because at that point 99.6% have failed, and the 50% failure rate might be at 100,000. Or 20,000. Or 190,000.
I don't think they're saying that 100% of phones will last to 200,000 folds. That would be a bold claim indeed.
If it's 0% failure at 200k folds, that means the 200k is about six standard deviations away from the mean, so I think (statistics gurus please correct) that says the average device will last for 1.7m folds.
Product lifetime is generally modeled using the Weibull distribution[1]. Depending on the parameters, a normal distribution is a reasonable approximation. Without data on the parameters, and just discussing whether "tested to 200,000 folds" means that every device will survive 200k folds, I think it's fair to use a normal distribution.
Agreed that they probably expect some percentage of RMAs. In fact, I'd argue that "tested to 200k folds" means that 200k gets them enough failures to model the lifetime distribution, so the average lifetime is probably considerably less than 200k.
That's the same Fold that everyone had to RMA repeatedly due to display failures. The story visited the HN front page at least once, and then Samsung did absolutely nothing to fix it.
I'm sorry, but I don't trust anyone claiming they've solved it until they've proven after 2 or 3 gens that there are no elevated failure rates.
Maybe this Google phablet will be the first that solved it, but I can't press X any harder for doubt.
Yes that's one of the key services you get in exchange for 30%. Small companies really don't want to be handling sales tax requirements for every jurisdiction on the entire globe, it's a nightmare. Not to mention fraud, etc.
It's 15% for developers making less then $1m/year.
And as Apple demonstrated in Netherlands with the dating apps it is actually cheaper to use them that try to run your own payment system. Especially if you're trying to target a global audience.
1ns of error equates to ~30cm of error (from speed of light), so:
You might get a couple of ns of error from the initial s and p waves. (Assuming on the order of 50cm displacement)
If you’re on the fault and there is a major displacement (say order of 1-2 meters) there might be several more ns of error.
Practically, this seems to suggest that by the time that displacement from an earthquake would cause a significant loss of clock accuracy (say > 10ns), you most likely have bigger problems to worry about..
