if you think about the speeds involved, a single additional car in front of you on the freeway (or even any additional cars) adds pretty miniscule time to the total commute.
Let's compare a few situations. In the baseline you're tailing the car in front of you with a focus on not letting anyone cheat and get in front of you, let's say 50 feet away. Your commute is 30 miles, and in this frictionless sphere of traffic you're going 60mph the whole time. You get to work in 30 minutes flat.
In the second scenario you're following the 3-second rule[0]. This would put you ~285 feet behind the car in front of you. Let's say over the course of your commute 20 cars move in front of you. If the average car length is 15 feet, and they all are 50 feet away from each other, when all 20 cars are in place you're a net -(20 * 65) feet away from the original car, or 1300 feet total. At 60 mph that adds ~15 seconds to your total commute time.
Well worth having an easier time avoiding a potential crash IMO! Also has the benefit of helping prevent traffic to begin with[1]
It is not about efficiency or trying not to be slow. I am not bothered how many people cut in line or "cheat".[1]
Someone will always keep cutting into the space. It is impossible to maintain 3 second to the car ahead. First one car will cut in, you make room for them and add 3 second gap the next car will cut in. Maintaining even a 1 second gap is sometimes very hard close to exits.
[1] Personally my driving pattern changed once I switched to driving a hydrogen fuel cell Mirai, slower is better on the mileage and fuel cell owners are very range conscious.
In the Midwest here and drive the same way, from rural to busy city and back. Never had any issues at all with it minus the car or five I let in front of me. I usually keep about 5 car lengths, so a fairly reasonable gap without being annoying to drivers that happen to be behind me. Also, I have zero panic braking incidents, which seem to be a big cause of crashes and slowdowns in heavy traffic due to the slinky phenomenon that tends to happen.
Not only can I see everything beyond the car in front of me, I tend to "soak" up the braking energy when there is a panic braking incident in front of me. I have no data to back this up, but I'm almost positive I've kept traffic moving much better behind me and prevented rear ending incidents using this tactic. Also, I wouldn't downplay the amount of fuel that is being saved that comes with not having to almost stop, then start over again from the slinky effect.
GPT-4 is a fine-tuned model (likely first fine-tuned for code, then for chat on top of that like gpt-3.5-turbo was[0]), while PaLM2 as reported is a foundational model without any additional fine-tuning applied yet. I would expect its performance to improve on this if it were fine-tuned, though I don't have a great sense of what the cap would be.
> We develop a large multimodal model (LMM), by connecting the open-set visual encoder of CLIP [36] with the language decoder LLaMA, and fine-tuning them end-to-end on our generated instructional vision-language data
afaik sentence embeddings via sbert are still considered a pretty viable path. This may be what you were already looking at, but there's more info here: https://www.sbert.net/index.html
Part of its contents come from the "USPTO Backgrounds" dataset. From The Pile's paper:
> USPTO Backgrounds is a dataset of background sections from patents granted by the United States Patent and Trademark Office, derived from its published bulk archives. A typical patent background lays out the general context of the invention, gives an overview of the technical field, and sets up the
framing of the problem space. We included USPTO Backgrounds because it contains a large volume of technical writing on applied subjects, aimed at a
non-technical audience.
For twitter, here are the actual number for the bay area, according to state filings[0]
* SF: 784
* San Jose: 106
Still high numbers IMO but SF for example has ~800k residents, so this only represents ~0.1% of the total population. Not super convinced yet that it's going to have any immediate impact on housing in the city unless this triggers a much larger avalanche of firings that don't have an associated swarm of smaller companies waiting in the wings to pick good employees up at a discount.
Housing is a very "tipping point" thing - if there are 100 houses for sale and 101 buyers, prices will rise until one buyer drops out. If there are 100 houses and 99 buyers, prices will drop until another buyer appears or a seller takes a house off the market.
It's not exact, but small changes can have large effects.
However SF likely has a huge waiting list of people who would like to move in, so it's not likely to take effect immediately.
> Not super convinced yet that it's going to have any immediate impact on housing in the city
It's not "Twitter layoffs cause issues" - it's more that if you can identify 5, 10, 15+ well known companies doing layoffs concentrated in that area there is probably also a very long tail of smaller companies doing layoffs also. Cumulative effect of it all can be quite large.
> ~40% of the workforce
This is an enormous exaggeration. However, even a 4-5% reduction in jobs in the area can have an enormous effect on the economy and housing market.
wow - how is this not the default? It's always the behavior that I expect to happen, and I'm pretty much always surprised and disappointed when it does a global search instead.
This is just speculation, but based on a conversation with my partner, there are two types of computer users. The first group makes use of hierarchical storage, consistent naming conventions, and other organizational tricks to give them a rough idea of where any file might be. The second group has never heard the term "file systems" and just stores everything with an arbitrary name in whatever location the originating application uses by default.
The first group would prefer to search the given directory, because the supplied context (of which folder to start the search in) drastically improves/speeds search results. The second group prefers to search the entire disk, because supplying that additional context is impossible - any file might appear anywhere.
The set intersection between the first group and "people who change their default settings" is much higher than it is with the second group. Consequently, the whole disk search is enabled by default.
Additionally, given the addition of an "All My Files" view in Finder (a feature which the first group would probably find baffling), Apple may also believe that the latter group outnumbers the former.
Very much an intentional goal of the OS when it came out, the hope being that the concepts of files and folders might never be needed. A bold move to get rid of a mental overhead for using a computer that was never natural or intuitive to non-tech users. They eventually had to back off a bit, but it’s an amazing accomplishment that most users can still use these devices as if files weren’t a thing.
Yeah this is true, friends that are professors say they now have to have a basic class early in the semester that explains folders, files, and other basics to students.
Funny, I'm in the first group but I'd much rather have it search everything by default: because I organize my files, if I'm searching for one that means that I don't know where it is, so I need to look everywhere.
To me this is definitely it. Using non-technical people's computers is eye-opening. Thousands of files scattered around on the Desktop, Documents, or wherever they just happened to go. Finding something in a specific directory would be a completely foreign concept.
Well, count me in the third group: I make use of hierarchical storage, consistent naming conventions, and other organizational tricks to give me a rough idea of where any file might be. I also fail miserably at it (apart from dev/programming stuff, which all is in `~/dev/<project name>`) and rely heavily on search to find anything again.
As search is fast enough nowadays with indexing, I'd rather have it search the whole disk every time than the directory I'm in just to realize I've put it someplace else.
Windows also defaults to full text file searches instead of filenames, which is super annoying when you have thousands upon thousands of documents. It's often faster to open a terminal window, type 'dir search string and open the file from the terminal.
I also have a 10g card and saw the same thing - to get it working I had to pass in "--n_samples 1" to the command, which limits the number of generated images to 2 in any given run. This has been working fine for me
Let's compare a few situations. In the baseline you're tailing the car in front of you with a focus on not letting anyone cheat and get in front of you, let's say 50 feet away. Your commute is 30 miles, and in this frictionless sphere of traffic you're going 60mph the whole time. You get to work in 30 minutes flat.
In the second scenario you're following the 3-second rule[0]. This would put you ~285 feet behind the car in front of you. Let's say over the course of your commute 20 cars move in front of you. If the average car length is 15 feet, and they all are 50 feet away from each other, when all 20 cars are in place you're a net -(20 * 65) feet away from the original car, or 1300 feet total. At 60 mph that adds ~15 seconds to your total commute time.
Well worth having an easier time avoiding a potential crash IMO! Also has the benefit of helping prevent traffic to begin with[1]
0: https://driversed.com/trending/what-safe-following-distance.
1: https://www.youtube.com/watch?v=iHzzSao6ypE