It turns out that they took 640lbs (!) of weight to mars to be tossed off at various points during EDL. The video is worth watching if you'd like some more of the nitty-gritty behind the process: https://www.youtube.com/watch?v=W0NakShgbHY
Great video! I am also surprised by the fact that they bring that much mass just to jettison. In theory they could mounted some of the useful mass on a slider/rail system to achieve the necessary adjustment to the mass distribution without dropping mass overall, but apparently it wasn't worth the complexity/volume cost.
I'm sort of surprised we don't yet have ML powered "de-accent-ization". His french accent isn't hard to understand at normal speed, but when I set it to 1.5x or 2x speed it becomes hard to decipher in a way native speakers usually are not. If there was just a button (for him or me) to hit to tweak the sounds a bit to reduce the accent, I bet this problem would go away.
I’m amazed that people still say things like “de-accent” as if there was such a thing as “no accent”. You are asking for a button that makes his French accent more like your own. It’s a separate thing from native vs. non-native speakers - there are plenty of native English speakers with accents that you would also find challenging.
You are reading something into my comment that wasn't there in order to pick a boring fight. There is of course no such thing as no accent a priori, but there is such a thing as "accents understood by (vastly) more people" and "accents closer to the mean accent of native speakers". When I learn Russian, my English accent is not on the same footing as a Muscovite's; the intended notion of "de-accenting" the English accent on my Russian is obvious.
Consider responding to the substance of the comment instead.
> accents closer to the mean accent of native speakers
Notice that, for the case of English, most speakers are not native, by a huge margin! Native English speakers are a biased minority, and with a lot of variation within. Not sure that an "average native" accent is a useful concept at all. I, for one, tend to find most non-native English speakers vastly easier to understand than many native speakers.
> Notice that, for the case of English, most speakers are not native, by a huge margin!
I'm well aware, and this does not rebut any of my points.
> I, for one, tend to find most non-native English speakers vastly easier to understand than many native speakers.
That "many" native speakers in a language of hundreds of millions of speakers are hard to understand does not challenge the claim that a non-native accent brought closer to any native accent, much less the mean native accent, will for the large majority if listeners be easier rather than harder to understand.
I, for one, would be uncomfortable with AI removing my accent. I understand it's for other people to understand me better - and I am fine with AI-generated subtitles - but altering the way I speak would reduce the amount of "me" in ways I'm not fully ok with.
What's special about speech that makes this argument apply to speech alteration but not to subtitles? It's a tool to make you easier to understand, not to erase your person.
Speech is more than articulating words. It is also about rhythm and melody and idealy also body language.
The way someone speaks is very unique ... and it is actually very, very important how you speak to bring your point across. Or ... to convince people.
A robot voice might present the best arguments, but it will very likely loose to a good speaker who can (literally) tune in to his audience.
Speech is a complex pattern of sound waves, containing much more information, than binary encoded words.
So if there was a ML tool to make people with strong accent more understandable, why not. But you can also numble without any accent.
And I can enjoy and understand certain people with strong accents much better than natives, because they are just good speakers.
And having subtitles is one thing, but changing their voice .. would require consent I believe. (unless you run the tool for yourself, but I believe parents point was, he speaks and then automatically a tool enhances his voice, I would not like that, too)
No need to have a fancy all-new ML algorithm - stick a text-to-speech output on the auto-generated video subtitles and you can set it to whatever language you like.
If the speech recognition / subtitling algorithm can't understand the nuances of the language, that's going to be a problem anyway... accented pronunciation is so multidimensional, you're pretty much going to have to transcribe syllables/phonemes first...
I was hoping someone would link this video, it describes the various phases with a ton of details that.
Some of the other videos on this channel are just as in-depth: the ones about the plumes/exhaust of rocket engines as well as star occlusions are incredibly detailed.