HN user
https://tristan.bruge.re
Hell sometimes the ai may look at values in .env or other uncommited stuff. Its execution path may depend on which commands are available in the environment. The specific model and settings. On top of the inherent randomness if you havent set model heat to 0. i suspect it is very hard to get consistent reproducible ai runs
But then who's responsible for reviewing and correcting the ai-written code ?
I used to put swap on zram when my laptop had one of those early ssds, that people would tell you not to put swap on for fear of wearing them out
Setup was tedious
Indeed ?
The EU in its current form is mostly about markets. It routinely pushes for the sacrifice of government monopolies to the altar of the free market (see for a recent example the french train network). Most of its regulations are to ensure a level field for a balanced market.
Hell it pushes for free markets even when it makes very little sense (the eu electricity market and its weird idiosyncrasies are an artifact of that)
It basically bans member governments from printing money and imposes very strict limits of 3% GDP on government deficits. For reference the US deficit was 5.9% gdp this year, Almost twice as much. this greatly limits government control over the economy.
Socialist is a very weird term to use here. The eu is the epitome of neoliberalism, even more so than the us
As someone who worked on transformer-based diffusion models before (not for language though), i can say one thing: they're hard.
Denoising diffusion models benefited a lot from the u-net, which is a pretty simple network (compared to a transformer) and very well-adapted to the denoising task. Plus diffusion on images is great to research because it's very easy to visualize, and therefore to wrap your head around
Doing diffusion on text is a great idea, but my intuition is it will prove more challenging, and probably take a while before we get something working
In between the beds?? Does that mean the shower was right in the middle of the room ? So that it would be impossible to place a double bed ? This is the weirdest part to me
I tried that concept whith my ergodox when i had an arm in a splint, but i couldnt quite get my brain to wrap around it. I could type on the right key, but not press the mirror/switch key at the right moment.
What would have made it easier is if it could infer the right key like an autocorrect
You are slightly wrong.
Python does allow you to put anything in annotations. ( pep 3107 that defines type annotations says that explicitly [1]).
But it also defines a type checking annotations, which is a specific convention for using pep 3107 annotations. Type annotations were introduced in PEP 484 and updated in a lot of subsequent peps. The python typing system is fully specified in [3].
It does have several implementations, although the reference implementation is mypy
[1] https://peps.python.org/pep-3107/#rejected-proposals
Python has static typechecking which, while not perfect, works pretty well as long as you're not actually trying to circumvent it (or manipulating things "too dynamically")
This feels like something that should be opt-in, not opt-out. It feels trivial to have all clients that support it send a header stating they do, and it is ridiculous that the default is to allow sending reacts to clients that don't support them
"And hence invertible" <- does every output embedding combination have an associated input ? Are they able to construct it or is this just an existence result ?
They are looking at the continuous embeddings, not at the discrete words inferred from them
That would be purely statistic and not based on any algorithmic insight.
This is machine learning research ?
As little as one experience is worth,
When I lived in socal, almost every person under 35 i knew there was living in a 2-4bd with roommates.
1bd and studios were very scace, and almost often prohibitively expensive.
So I second that creating a real offer for 1bd and studios would definitely free up family housing.
On top of that 1 and 2-person households still need housing. Building some for them is a good thing in my book
Whether good or bad, it's important to realize this is not true in California, with regard to these laws. They apply everywhere, not only in urban centers.
According to the linked article, the only areas affected by this bills are major transit centers and high throughput public transit stops, which tends to exclude small towns far from cities. If you look at the linked map, affected areas are all concentrated in big cities
Are you commenting on the bill being discussed here or on something else entirely?
Of course but configuration for software is exactly what hidden files are usually used for
If that is the case then why call it Agents.md instead of integrating it with already existing documentation files or calling it something like "Summary.md"?
It is files that are meant to be read by software, not humans. From my point of view this seems like a prime candidate for a hidden directory?
On the subject of artificial monopoly, it's interesting to compare video streaming to music streaming.
Music streaming platforms (Spotify, deeper, apple music, tidal etc.) Generally work a lot better than movie/series streaming. It seems that competition between them works quite well, prices are reasonable, and more importantly, any subscription gives you access to pretty much all of mainstream music. There's hardly any content exclusive to one platform, so you can essentially get any of them and be done with it
Contrast that with video streaming, where content is pretty much exclusively tied to one platform. As a consequence, people routinely have several subscriptions instead of one, and platforms compete on library more than on price or quality of service. Overall experience is much worse
I wonder why this difference came to be, although these are very similar services (with basically the same copyright mechanism)
From my noobish standpoint, it feels like most code shounldn't care what the page size is? Why does it need te be recompiled?
What typically tends to break when changing it?
103 years old. Pretty recent at the scale of Mathematics, although that does make it older than Bourbaki. And than category theory
My very personal opinion is maybe it shouldn't be ok for humans either. Maybe the benefits of cars absolutely don't outweigh the contant risk of deaths and injuries they pose (especially in metropolitan aeras where they are easily replaced by other modes of transportation)
Reality is "because humans can take responsibility for their actions". And it seems the way our society is built, we'd rather have individuals take responsibility than question the system
If Python is too slow or otherwise not good enough to write a good Python package manager in, why use Python altogether?
To write things that are not package managers? Are you suggesting not being the right tool for this job implies not being the right tool for any job?
I mean, they might not have given thought to that particular corner case, they probably wrote something like
memcpy(void* ptr1, void* ptr2, int n)
Copy n bytes from ptr1 to ptr2. UNDEFINED if ptr1 is NULL or ptr2 is NULL
‐------
It might also have come from a "explicit better than implicit" opinion, as in "it is better to have developers explicitly handle cases where the null pointer is involved
The infamous bane of French high-school students
it's more like saying something like "english is more readable than french to a spanish speaker"
On the contrary French is much more readable than English to a Spanish speaker. Because French is much more similar to Spanish than English is.
Same with your JS example, I would guess it is much more similar to what you are used to
If I were a security engineer/pentester, I might post unsigned "updates" that automatically alert us if they ever get installed
RL absolutely considers that the environment is stateful and that said state is affected by actions of the agent yes