j0shm1lls
Member
Looks like another push of 2022.44.30.5 is happening, go out and burn your daily checks!
-
Want to remove ads? Register an account and login to see fewer ads, and become a Supporting Member to remove almost all ads.
FSD Beta 10.69
- Thread starter Buckminster
- Start date
FSDtester#1
Craves Electrolytes
j0shm1lls
Member
My pleasure!Thanks for the update, and thanks for getting Zombie stuck in my head!
Great song
I've poked fun at the release notes myself, but truthfully it's one of the things I love most about the FSD beta development. I would be very disappointed if they tried to dumb them down or not discuss things that they're working on, but that people might not easily grasp. There are plenty of follow-up YouTube videos for people who want an interpretation/explanation of what they mean.
I note that recent forum re-posts of release notes from Cruise are written in a very similar style. I'm not saying Cruise is taking a page from Tesla's style here; maybe this is just the way AV engineers, or lmore generally ML engineers write up their releases. But perhaps, just the act of publishing them is inspired by Tesla's FSDb practice
For all the nattering about Elon's descriptions of FSD progress, It's been a remarkably open and transparent process, far more so than any major competitor that I know of The release notes are one facet of that, and I'd be happy to see even more detailed and more frequent posting of such material from the Autopilot team. If you want to slam Tesla FSD, they're giving you plenty of weapons to do it; no need to whine about too-much or too-technical information in the published notes.
old pilot
Member
The notes are written by engineers for engineers which is of no use to the system operators. We have very little information as to what the system is supposed to do in any given situation and the notes do not add anything to our knowledge. Tailor the information to the audience who use the system.
Just got 2022.44.30.5 on my MYLR, and I do NOT have FSD beta enabled. (I'm also just on a one-month FSD subscription for a road trip, which is already set to cancel.) So this says to me that everyone is finally being merged onto one software release track.
What does this mean for non-beta FSD operation? (I realize highway operation is still the old stack for everyone until v11.) If I leave FSD beta disabled, is it still just using the legacy FSD-on-city-streets code? Or is it now actually using the new FSD beta code, but with some features disabled?
What does this mean for non-beta FSD operation? (I realize highway operation is still the old stack for everyone until v11.) If I leave FSD beta disabled, is it still just using the legacy FSD-on-city-streets code? Or is it now actually using the new FSD beta code, but with some features disabled?
n.one.one
Info w/o Drama
My guess is that you will will have this beta version until the next general (non beta) release. At that point you'll no longer have beta.
And the way they throw around percentages has the feel of a retail store marking up prices before the big 50% sell-off sale. A 20% chunk improvements in something like system delay should be noticeable but not so much especially when they add things like 30ft crawls to stop signs or eliminating operation in light rain.
scottf200
Well-Known Member
Looking at it another way tho ... is that it seems fantastic that they are using analytics and empirical values to monitor their test cases.
It seems people are thinking about FSD processes like old-school programming logic of IF-THEN instead of Machine Learning/Neural Networks.
I've got mixed feelings about your comment. It comes down to language; that is, the ability for people to communicate with each other.
As far as I can tell, for each and every specialty in human endeavor there appears to be a specialized language. Lawyers have their own, to the point where, in a bog-standard contract, the first couple of paragraphs are given over to definitions of what some of the words to follow mean. And that's just the words that aren't already in standard use by lawyers; from time to time, there are court decisions (usually by appeals courts and sometimes by SCOTUS) where they literally start hunting up word definitions in $RANDOM dictionaries.
Beyond that, take sailors: the language used for sailing ships has its own grammar and word definitions dating back centuries. Sheets, lines, ropes, left and right and port and starboard (both used, with specific grammar). For that matter, the people who run railroads also have their specialized language, to the point of not being understandable at all by mere mortals not involved in the business.
Why does this happen? Because people invent linguistic short-cuts so they don't have to try and explain which rope/lever/concept in excruciating detail every time they mention a topic. Once one has done that, then that short-cut gets used to explain other short-cuts, and the whole process snowballs. It can and does get to the point where word definitions are completely made up out of previous short-cuts. It may look (and be) ugly, but it means that quite complex ideas can be batted around faster than a shuttlecock in badminton.
I'm a EE; at one time I was part of a group that was building a honking big communications system, there were roughly a hundred of us, hardware and software, and we had to discuss the ins and outs of the system as we designed it. I joked at the time that, "If you're not inventing an acronym a day you're not working!" And, yeah, our conversations in the middle of all this got pretty obtuse. But the definitions of the words we were inventing, or the words that were in vaguely common use whose definitions we were changing on the fly, were pretty blamed exact, as grammar terms tend to be, and what was a noun, verb, adverb, and so on were pretty much set in stone. (By the by: Physicists and mathematicians are famous for this. The entire Greek alphabet, full of innocent little letters minding their own business, has been hijacked for quite complex ideas. Chi-analysis, anyone?)
The problem crops up when one tries to unroll all these complex terms and lingo into the common language. A single sentence unrolled into simple, non-jargon words, might take up a whole page; a page full of these terms couple easily take up a chapter, and so on.
I swear, at least a third of undergraduate EE is learning concepts, learning the lingo, and figuring out how to manipulate those concepts with the lingo so learned. People who think that engineers have lousy grammar and writing skills had better think again: It's not a nicety, it's a requirement.
It's pretty clear by now that the AI guys sweating the details of FSDb walked into the project with their own lingo, a lot of it probably learned in college, and probably a lot more learned on the job. And, to my eye at least, a lot of what they're doing over there appears to have a flavor of pure research: They're inventing concepts left, right, and center, with the lingo to manipulate those concepts as they go.
There's a quote I read, once, from some professor or other, who said, "If you can't explain your ideas in single or double-syllable words, then you don't understand what you're talking about." Ayup; good point. But that comes from somebody whose job is explaining concepts and ideas to the ignorant. (Note: Not stupid, just ignorant.) It takes a certain mindset and practice to take complicated ideas and distill them down to levels where the hoi polloi can follow. My father, who was actually a college professor and dean, was death on making durn sure that people who were researchers had to teach undergraduate courses for sure, and graduate courses as a maybe: His opinion that if you didn't teach and lecture, then you got worse at researching, because you got worse at the fundamentals and lost contact. But, again, he was riding herd on people paid to teach, not necessarily on those who had to invent at high speed.
Which brings us back to the people writing those release notes. There is a species of STEM called, "Tech writers" who are trained in that ability to break complex ideas down to simple ones. Dunno if Tesla employs those; probably, just so they can bring new people up to speed. But the in-house audience are people fresh out of college with their new-found lingo built in. Given the newness of the AI and tech built into FSDb, they're probably not writing for the public, especially as concepts and ideas change over time.
And that's probably the point. It would probably take serious time and effort to break what they're talking about in the release notes to the point where a person with a high-school vocabulary could follow what they're talking about. And what would you rather have? Understandable release notes, or FSDb as fast as possible?
Me, I'd rather have the FSDb. And the release notes are, well, an education in process. I've already been corrected by forum members a couple of times when I spoke about what looked like Standard English words, but had been redefined as FSDb AI Lingo.
Hang on for the ride
.
scottf200
Well-Known Member
old pilot
Member
Well said, and I understand your point but if correct why bother to even include the notes in the release. I graduated for college in the seventies when they were still using punch cards with computers and the computers that now fit in your pocket filled an entire room. I worked in a field that used more acronyms and special words than I care to remember and specialized computers. There were other specialized fields that supported me in this endeavor who's information was critical to my success in the operation and they were able to relay the information to me in a way I could understand and use. I think the question is what is it we are trying to accomplish? The engineers are trying to create a better FSD and we are trying to operate the FSD. If we the operators of the system are not important to the goal then the notes should be for the engineers if, however, we the operators are important to the success of FSD then the notes should reflect that. I do not know the answer to the question of how important the operators are to the success of FSD.I've got mixed feelings about your comment. It comes down to language; that is, the ability for people to communicate with each other.
As far as I can tell, for each and every specialty in human endeavor there appears to be a specialized language. Lawyers have their own, to the point where, in a bog-standard contract, the first couple of paragraphs are given over to definitions of what some of the words to follow mean. And that's just the words that aren't already in standard use by lawyers; from time to time, there are court decisions (usually by appeals courts and sometimes by SCOTUS) where they literally start hunting up word definitions in $RANDOM dictionaries.
Beyond that, take sailors: the language used for sailing ships has its own grammar and word definitions dating back centuries. Sheets, lines, ropes, left and right and port and starboard (both used, with specific grammar). For that matter, the people who run railroads also have their specialized language, to the point of not being understandable at all by mere mortals not involved in the business.
Why does this happen? Because people invent linguistic short-cuts so they don't have to try and explain which rope/lever/concept in excruciating detail every time they mention a topic. Once one has done that, then that short-cut gets used to explain other short-cuts, and the whole process snowballs. It can and does get to the point where word definitions are completely made up out of previous short-cuts. It may look (and be) ugly, but it means that quite complex ideas can be batted around faster than a shuttlecock in badminton.
I'm a EE; at one time I was part of a group that was building a honking big communications system, there were roughly a hundred of us, hardware and software, and we had to discuss the ins and outs of the system as we designed it. I joked at the time that, "If you're not inventing an acronym a day you're not working!" And, yeah, our conversations in the middle of all this got pretty obtuse. But the definitions of the words we were inventing, or the words that were in vaguely common use whose definitions we were changing on the fly, were pretty blamed exact, as grammar terms tend to be, and what was a noun, verb, adverb, and so on were pretty much set in stone. (By the by: Physicists and mathematicians are famous for this. The entire Greek alphabet, full of innocent little letters minding their own business, has been hijacked for quite complex ideas. Chi-analysis, anyone?)
The problem crops up when one tries to unroll all these complex terms and lingo into the common language. A single sentence unrolled into simple, non-jargon words, might take up a whole page; a page full of these terms couple easily take up a chapter, and so on.
I swear, at least a third of undergraduate EE is learning concepts, learning the lingo, and figuring out how to manipulate those concepts with the lingo so learned. People who think that engineers have lousy grammar and writing skills had better think again: It's not a nicety, it's a requirement.
It's pretty clear by now that the AI guys sweating the details of FSDb walked into the project with their own lingo, a lot of it probably learned in college, and probably a lot more learned on the job. And, to my eye at least, a lot of what they're doing over there appears to have a flavor of pure research: They're inventing concepts left, right, and center, with the lingo to manipulate those concepts as they go.
There's a quote I read, once, from some professor or other, who said, "If you can't explain your ideas in single or double-syllable words, then you don't understand what you're talking about." Ayup; good point. But that comes from somebody whose job is explaining concepts and ideas to the ignorant. (Note: Not stupid, just ignorant.) It takes a certain mindset and practice to take complicated ideas and distill them down to levels where the hoi polloi can follow. My father, who was actually a college professor and dean, was death on making durn sure that people who were researchers had to teach undergraduate courses for sure, and graduate courses as a maybe: His opinion that if you didn't teach and lecture, then you got worse at researching, because you got worse at the fundamentals and lost contact. But, again, he was riding herd on people paid to teach, not necessarily on those who had to invent at high speed.
Which brings us back to the people writing those release notes. There is a species of STEM called, "Tech writers" who are trained in that ability to break complex ideas down to simple ones. Dunno if Tesla employs those; probably, just so they can bring new people up to speed. But the in-house audience are people fresh out of college with their new-found lingo built in. Given the newness of the AI and tech built into FSDb, they're probably not writing for the public, especially as concepts and ideas change over time.
And that's probably the point. It would probably take serious time and effort to break what they're talking about in the release notes to the point where a person with a high-school vocabulary could follow what they're talking about. And what would you rather have? Understandable release notes, or FSDb as fast as possible?
Me, I'd rather have the FSDb. And the release notes are, well, an education in process. I've already been corrected by forum members a couple of times when I spoke about what looked like Standard English words, but had been redefined as FSDb AI Lingo.
Hang on for the ride
Checked software this morning and 10.69.25.1 downloaded. I guess Tesla has thrown in the towel for getting 11.x past the employee stage before the end of the year.
Sad. But, reflecting back over the year, FSDb has come a long way from jerking around simple turns and not knowing how to take it's turn at a stop sign. Still a long way to go, so may as well enjoy the journey.
Tomorrow's a new beginning. Maybe next month we'll see 11.x.
Sad. But, reflecting back over the year, FSDb has come a long way from jerking around simple turns and not knowing how to take it's turn at a stop sign. Still a long way to go, so may as well enjoy the journey.
Tomorrow's a new beginning. Maybe next month we'll see 11.x.
ok, but if you can't explain in English without saying bug fixes then you find a Bill Nye type of guy who can take science and explain in laymen's term. Most
people nowadays have a short attention span in everything from politics to medicine
I can't say that I've scoured hundreds of YouTube drives with FSD but most of them are on very uncomplicated routes. I'd do one around here but I'm tired of
fixing my rims
jebinc
Well-Known Member
I just got this too and on the FSD subscription model. I never have requested FSD Beta, but now it says I am on FSD Beta 10.69.25.1 (2022-44.30.5). So, does that mean I am now on FSD beta, even though I never requested it?
You don't have to...that's why most companies have press departments
Ever look at the release notes for any google app on an Android phone? "Bug fixes and performance improvements." Very rarely anything else. Many other app makers have followed suit.
The nice thing about Tesla's release notes is that they do have details. If you don't understand it, there are youtube videos that try to interpret them in layman's terms.
Similar threads
- Article
