Project like this make me stop and truly appreciate the time someone took to make this... and make me wish I had more time in my day to consume this beauty in its entirety. So much knowledge retained here.
My dad passed away from squamous cell carcinoma in 2010. In retrospect, through my casual research into the space and tools like this one, it occurs to me that the entire event was likely preventable and occurred merely because we did not react quickly enough to the cancer’s presence.
Thanks for your comment - my practice is focused on skin cancer and I see so many patients that bring a photo from many months earlier that shows an obvious skin cancer that could have been treated more easily at an earlier stage. Patient education should enable these to be picked up sooner.
My 3yo loves chatting with Gemini, more than just watching some show. It give her undivided attention and always asks follow up questions to her silly comments.
I do wonder if this isn’t an improvement on watching TV though. Is AI time better than watching a show? The comment on a lack of friction is true in both cases but the active experience has to be an improvement on the passive one?
Knowing the MTA, I assume "outfit" means "provide a turnkey solution for the buses in the program, including software and staff" not just "slap a camera on a bus."
This is backed up by an earlier report on ABLE[1], which was the earlier program that this one (ACE) is built on: hardware was 1/3rd of the net cost to the MTA in that program.
also keep in mind this is automotive grade stuff in relatively harsh conditions (the front of a bus); it needs to be more rugged than your average camera
Surely it's just a dash cam with a button and maybe some kind of comms module. There are many mass-market dashcams with a 'record' button that saves a time period before and after pressing the button, it's pretty standard usage. The only new thing is sending the video on, and making sure the time/location are correct (many cameras do this via GPS already - standard usage). I'm not seeing what is so expensive or unique.
* How many years of service are included (both for the data plan and just repairs when someone smashes a bus camera). Employing people to fix/replace cameras could easily cost hundreds of thousands per year
* Where does the video go? Redundant storage in geographically disparate areas?
* How long is video retained?
* R&D costs?
* Does the price include manual human review?
* There was also some drama where the 2 vendors sued each other, idk if that has any impact on the prices
* The tickets during the pilot made millions of dollars per year, do the cameras pay for themselves?
* How much efficiency are they gaining if the program results in fewer blocked busses?
A bus goes through many more miles and hours of operation than the vast majority of cars. Most cars are parked for the vast majority of the day, and sometimes in covered garages.
And how long do those cameras last? The feds only fund bus replacements at twelve years or 500,000 miles.
Unlike most other parts of the bus, this part isn't mission-critical and presumably is easy to replace. It doesn't need to be made to the same durability standard as, say, the engine.
I have doubts this is actually that profitable. More likely the company is rolling their labor overhead into the per-camera pricing. It's a small order too, only 1000 cameras.
Since this is reported on by "Insurance Journal" it makes wonder how much of this fine was paid for by their liability insurer. Does anyone know? Presumably they have no coverage for illegal activities, but they never agreed they did anything illegal.
>I apologize, GPT-4, for mistakenly accusing you of making mistakes.
I am testing large language models against a ground truth data set we created internally. Quite often when there is a mismatch, I realize the ground truth dataset is wrong, and I feel exactly like the author did.
Apologizing to a program seems rather silly though. Do you apologize to your compiler when you have a typo in your code, and have to make it do all that work again?
I still remember paying 40 euros like 16-17 years ago for the 512mb micro-sd card (it was called trans-flash at the time, don't know why) to put in my phone (lg u8360) and saying nothing to my parent because it was a crazy expensive thing at the time.
> it was called trans-flash at the time, don't know why
It was the original name, without the hyphen, though:
>> MicroSD form-factor memory cards were introduced in 2004 by SanDisk at CeBIT[16] and originally called T-Flash,[17] and later TransFlash,[18] which was abbreviated to "T-Flash" or "TF". T-Flash was renamed microSD in 2005 when it was adopted by the SDA.[19]
I just started getting a few texts from friends saying 'your post is on HN' and was confused. Prob been 10 years since anything I wrote has been on here so was an unexpected surprise especially as the comments are almost all quite nice (didn't expect that :)
They claim 50% with fine-tuning and/or(?) RAG (unclear marketing phrasing imo), and claim their method achieves 5% on the same task which is apparently a text-to-sql task set.
