Does this mean that when you disable, all labels are deleted, and when you turn it back on it has to re-scan all of your photos? Could this be a cost-saving measure?
Out of curiosity, have you ever attempted to create something from scratch and bring it to market? Not as an employee, but as someone trying to figure it out?
Your work history will impact the way you view this issue IMO.
Yes. 28 years ago I founded a small a software company (4 employees) that made test software for a Tier 1 automotive vendor's ECUs. I sold the IP to them in 2007. Next question? Or do you think my background makes my opinion invalid for another reason?
It was amazing just being able to build things back then. These days you need to cough up $1500 for a PDF of ISO 26262 to even think about doing something commercially with automotive ECUs. Depending on what you're trying to do, there are dozens of additional standards with thousands of pages. Not even to mention CARB compliance.
That sounds really cool. The background does provide a lot of perspective on your views. It makes a lot of sense. Thank you for keeping us safe on the road!
It's confusing how someone who has tried to build anything of substance, in say California, couldn't have run into a regulation or ten stifling innovation.
I for one have seen mid-5 figures spent on a dumpster enclosure, because of building codes.
The company would have more cash in the bank when he exercises those options.
They also wouldn't have to pay to create a new pay package to replace the voided one. This process is surprisingly expensive to create and implement a CEO pay package.
There were more reasons that shareholders voted for the package in the revote, but omitting others as I don't think they are related to financials
This is a great point. I often ask LLMs which coding language they are most proficient in, and Python is the most typical response. This has makes me want to write more Python.
Keep in mind, LLMs can't introspect. They integrate your prompt with their training data.
If Python is an extremely popular programming language people like to write about using (it is), you'll expect a lot of people to state that Python is their most proficient language and for that to make its way into the training data. When you ask the LLM, you'll get Python as an answer some fraction of the time and other languages some other fraction.
Those answers are orthogonal to how good the LLM actually is though. In the training data, you'd like a sufficiently robust sample of code (which Python should have) without most of the code being terrible (which Python doesn't have in the wild).
Mind you, IME they do better with Python than very new languages, and they do better with Python tooling than whatever maven build garbage you might have to deal with, but that's faint praise.
It seems to understand the way you want your unit tests written. So if you have a particular style, it's best to write one or two tests in your style, then it will use that same style when it starts writing tests.
You can do prompts like "I need unit tests for the exceptions that are raised in the SuchAndSuch.function()", and it will do it -- particularly if you have a unit test already written similar to what you'd like.
I test-drove a Model S about that time - it was hair-raising because the system would correct the cars position about a half-second after I would have. I hovered my hands by the wheel the whole time, waiting for it to careen us into a bridge abutment.
While SuperCruise is a highway only system, riding in a friend's car with it was pretty much a non-event.
Realistically, how many chances do you think the average person is going to give the alg to drive them into an object at 70mph?
Has there been a significant customer side advancement in Tesla autopilot in 10 years? From what I can tell, all Elon has accomplished is becoming the new Prius- a sprinkling of Teslas blocking the left lane of our highways (ask any Autopilot aficionado, that’s the safest place to use it) while the human inside watches TikTok.
> Has there been a significant customer side advancement in Tesla autopilot in 10 years?
Not super interested in whether you're "impressed" or not. I think the video answers your question. The part of the video you chose to point out (poor behavior in parking lots at the end of a drive caused by bad cell signal) says a lot on its own, imo.
Autopilot of 10 years ago was lane keeping and changing on highways only. Now the car does a drive from point A to point B with some erroneous behavior in the end in the parking lot, and you're like "huh, what's the difference?"
Clearly you are not coming into this with an open mind.
The video does answer my question. It shows a Tesla vehicle that (still) requires human oversight and intervention to navigate point to point. At around 6m20s in the person behind the wheel says "OK it's going the wrong way into that one way entrance again" as he puts his hands on the wheel and takes over from the system.
Tesla engineers may have made engineering progress towards their goal. Maybe they make it N miles further before requiring intervention. Maybe they can successfully navigate an additional M% of some region without issue. That is progress! However that progress does not necessarily translate into something that customers will consider an advancement or valuable. Certainly not all customers.
There is no value to me in a semi-autonomous driving platform that requires my attention and feedback. So I would not consider the performance captured in the video to be a "significant customer side advancement". The fact that others might disagree and find value in the video is OK with me. We need only to be honest with ourselves and others when describing the system performance.
There is another company actively running an autonomous fleet every day. That is impressive to me.
> Clearly you are not coming into this with an open mind.
I think it's great that you want to cheerlead for Tesla. "FSD" and "Autopilot" just don't live up to their names yet. I'm sure eventually they'll ship something worthy of the trademark and make all your evangelism worth it.
Come experience it between my house and walmart before you go telling people they need to experience it. Good luck having good weather during your drive, here in louisiana.
Self driving probably works great on Interstates with numbers like 5, 10, and the feeders. Based on my experiences with subaru self driving and watching videos and watching tesla drivers around here, a self driving car would give up after maybe 3 minutes. There were more teslas and other expensive "self-driving" style cars a couple years ago, but now i rarely see them. I wonder why?
I can consistently get my wife's subaru self driving to swerve into another lane without warning, without jerking the wheel, with very minor control inputs. Subaru keeps updating the firmware, but they haven't fixed the suicide merge!
Are you from Houston originally, by chance, or from a part of western Louisiana? Anytime I hear the word “feeder” it’s a dead giveaway the person either grew up near Houston area or hung around there long enough to use this word in their lexicon
Nope! I'm from the state where i had to consciously not type "the 5, the 10" - California. I've lived in Louisiana for 12 years, total, now, though. I couldn't think of a better name for like the 405 or 710 than "feeders" - maybe tributary; i know there is a term of art for those freeways, though.
“Frontage” is the nomenclature I usually see in lieu of “feeder”. Though I agree that feeder is more descriptive of what the road actually does when we consider a road as being as a network or graph with flow.
Though the roads you reference would probably not be called “feeders” in the same way. I take feeder specifically in the Houstonian meaning to refer to a frontage road that is used for local access that runs parallel to a limited access highway. Notably it must run parallel to the highway and exist for the primary purpose of providing local ingress and egress while preserving the limited access nature of the highway.
Yeah 10 years ago it was a bit much. It was pretty much just lane keeping and distance keeping, but it didn't read signs or lights yet. So if you're coming up on a signal it would just keep going!
> Chrysler was the first manufacturer to implement the device in 1958. They called it “Auto-Pilot” and it appeared in their luxury model as an upgraded option. Soon after, General Motors installed it in their Cadillac vehicles, naming it “Cruise Control” which has stuck to this day.
I love the tesla service. Bought my car without talking to a single sales person, they come to my home to rotate my tires on my schedule. Only complaint is cosmetic service tends to have a long appointment wait time, but things like a cracked windshield were immediately serviced.
I don't know how I will switch to another car brand after this.
I met julian touring UCSB as perspective grad students. We sat together at dinner and he was really smart, kind, and outgoing. Great to see him presenting this work!
TLDR They drove down narrow 1-lane streets and had to disengage once. While this sounds quite impressive, the article spends a lot of time focused on fear of FSD. There's even a hypothetical comment that he will surely be hit by an FSD enabled car.
reply