I use self-driving every single day in Boston and I haven’t needed to intervene in about 8 months. Most interventions are due to me wanting to go a different route.
Based on the rate of progress alone I would expect functional vision-only self-driving to be very close. I expect people will continue to say LIDAR is required right up until the moment that Tesla is shipping level 4/5 self-driving.
Same experience in a mix of city/suburban/rural driving, on a HW3 car. Seeing my car drive itself through complex scenarios without intervention, and then reading smart people saying it can’t without hardware it doesn’t have, gives major mental whiplash.
I would like to get my experience more in line with yours. I can go a few miles without intervention, but that's about it, before it does something that will result in damage if I don't take over. I'm envious that some people can go months when I can't go a full day.
Where are you driving?! If the person you're replying to has gone 8 months in Boston without having to intervene, I'm impressed. Boston is the craziest place to drive that I've ever driven.
Pro tip if you get stuck in a warren of tiny little back streets in the area. Latch on to the back of a cab; they're generally on their way to a major road to get their fare where they're going and they usually know a good way to get to one. I've pulled this trick multiple times around city hall, Government Center, the old state house, etc.
Or when. Driving during peak commute hours really makes you a sardine in a box and it's harder for there to be intervene-worthy events just by nature of dense traffic.
> Based on the rate of progress alone I would expect functional vision-only self-driving to be very close.
So close yet so far, which is ironically the problem vision based self-driving has. No concrete information just a guess based on the simplest surface data.
On a scale from "student driver" to "safelite guy (or any other professional who drives around as part of their job) running late" how does it handle storrow and similiar?
Like does it get naively caught in stopped traffic for turns it could lane change out or does it fucking send it?
I don't drive in Boston, but there is some impatience factor and it will make human-like moves out of correct-but-stopped lanes into moving ones. It'll merge into gaps that feel very small when it doesn't have other options.
Is it fair for the company to have received three months of payments from a customer but the salesperson doesn’t get commission at all? How will you retain sales staff when word gets out? What’s the period length over which if the deal dies the salesperson doesn’t get their commission? Do you roll back commission payments later when the customer stops paying?
These are all great questions which people have answered and it’s the standard solution to the problem of misaligned incentives between the company receiving recurring revenue and the sales person receiving an upfront commission.
You're not allowed to say that it's not reasoning without distinguishing what is reasoning. Absent a strict definition that the models fail and that some other reasoner passes, it is entirely philosophical.
I think it’s perfectly fine to discuss whether it’s reasoning without fully committing to any foundational theory of reasoning. There are practical things we expect from reasoning that we can operationalise.
If it’s truly reasoning, then it wouldn’t be able to deceive or to rationalize a leaked answer in a backwards fashion. Asking and answering those questions can help us understand how the research agendas for improving reasoning and improving alignment should be modified.
This reads like the author is mad about imprecision in the discourse which is real but to be quite frank more rampant amongst detractors than promoters, who often have to deal with the flaws and limitations on a day to day basis.
The conclusion that everything around LLMs is magical thinking seems to be fairly hubristic to me given that in the last 5 years a set of previously borderline intractable problems have become completely or near completely solved, translation, transcription, and code generation (up to some scale), for instance.
> but to be quite frank more rampant amongst detractors than promoters, who often have to deal with the flaws and limitations on a day to day basis.
"detractors" usually point to actual flaws. "promoters" usually uncritically hail LLMs as miracles capable of solving any problem in one go, without giving any specific details.
Google Translate just spits out nonsense for distant language pairs (English<->Korean etc) and doesn't compare to Sota LLMs, Whisper is a Transformer (Architecture used for LLMs) and Code Generators have nothing on LLMs.
Very impressive the top comment on this nonsense hit piece is someone vaguely implying extrajudicial ‘retribution’ is the appropriate solution to a problem which doesn’t actually exist.
The razor to use to determine whether something is actually evidenced based under uncertainty is whether you would follow the same policy if it was your own child.
There are many things that are simply uncertain and “untrue until proven otherwise” isn’t an exclusively optimal policy.
> The razor to use to determine whether something is actually evidenced based under uncertainty is whether you would follow the same policy if it was your own child.
What? This makes no sense. How do you explain anti-vaxxer parents with this perspective? Parents may feel they know best, but feeling and fact have nothing to do with each other.
It's ok, the strongest defenders of EBM are never going to discover anything worthwhile as they get caught in a loop of "no evidence enough to test" and "no evidence for this because nobody tests it"
The opposite approach exposes people to a lot of unnecessary and dangerous medical treatment. The evidence based approach has uncovered that stenting doesn't work[1], yet a lot of do something proponents are still installing them at great risk to patients and at great cost to medical systems.
Counterpoints: the detractors of this purported loop would likely neither fund the vast amounts of research they’d demand be done nor believe the results if they conflicted with their anecdata. I have yet to see a good faith argument against evidence based method that provides an effective and realistic alternative. Because that would take evidence.
I think this is actually not as obvious as it seems as equity is also power-law distributed. An executive founder may have 10-50x the equity of a founding junior employee, who themselves might have 10-20x the equity of a key early employee.
The power laws actually cut both ways. I think the optimal path is not entirely obvious without some particular understanding of whether or not you are a stronger player as a leader or a follower.
Frankly, I think Tesla is going to win this. The new self-driving is remarkably good. Based on this alone, I estimate (actual) level 5 self driving in 2-3 years. I'm convinced that the lidar sensors are essentially unnecessary and the vision-only strategy is basically going to work and be much cheaper in the process.
How about we wait till they have given a single public ride before crowning them winners? At the moment Tesla's effort is nothing more than a marketing campaign.
Tesla FSD version 13 (new version) videos are starting to trickle out on YouTube and while they could be edited, it does seem to handle some crazy stuff fairly well.
If Tesla can do level 5 in 2-3 years as you say (and that might be a pretty big “if”), that places them 5+ years behind Waymo.
What leads to the win here, then? Waymo constrained by the cost of LIDAR? Is it truly such a massive % of build cost that they can’t succeed? Is it that Tesla is vertically integrated?
Apparently they have been surprised at how few photons are required to see for these sensors. They are skipping the image computer vision step and going from photons to car control in as few layers as possible.
It's not an event camera, so it's very much taking images, which are then being processed by computer vision algorithms.
Event cameras seem more viable than CMOS sensors for autonomous vehicle applications in the absence of LIDAR. CMOS dynamic range and response isn't as good as the human eye. LIDAR+CMOS is considerably better in many ways.
Next time you’re facing blinding direct sunlight, pull out your iPhone and take a picture/video. It’s a piece of cake for it. And it has to do far more post processing to make a compelling jpeg/heic for humans. Tesla can just dump the data from the sensor from short&long exposures straight into the neural net.
Humans can also decide they want to get a better look at something and move their head (or block out the sun with their hands) which cameras generally don't do.
Based on the rate of progress alone I would expect functional vision-only self-driving to be very close. I expect people will continue to say LIDAR is required right up until the moment that Tesla is shipping level 4/5 self-driving.
reply