After testing via api for a few days, this model is genuinely impressive for its size/price.
Feels a little like running opus 4.6 on medium think. Quick enough, clean enough. Still makes some minor errors once in a while, but it’s fast enough to be useful and cheap enough that you can use it for everything and still pay very little.
Upgrade that to 3.8 as soon as your hardware allows (and your use case makes sense). 3.8 is quite a bit more rational.
A new malaphor minted in the wild!
The two expressions you merged were:
The idea that “public companies only exist to maximize value for the shareholders” is toxic af, and only introduced because some famous economist made it popular [source needed, I have some vague recollection but I’m talking out of my ass]. Once it became the de facto law of capitalism, “investors” stopped allowing decisions that didn’t benefit their bottom line, and could sue to ensure decisions that they disagree with (especially if they cost more than they earn)
If we took away the “legal obligation to the shareholders”, which is an ultra-simplistic optimization that often results in “greedy algorithm” decision making for next quarters earnings, and made the obligation to 50% employees and 50% shareholders, along with better ongoing taxation of negative externalities, being on the board would be a real job instead of a few hours a week position in some cases. You’d have to negotiate big picture things, handle employee concerns seriously, while still being pushed to generate revenue by the shareholders.
Capitalism would probably still work, mostly. I think if leadership weren’t driven by “shareholder value”, adding plans to mitigate externalities that aren’t required by law would be easier. Currently, I don’t think any company builds a coal power plant with smoke or co2 capture that isn’t required to by law.
Stock markets would still be able to pick and choose investments, but if a company chose a 10 year plan that loses money the first year, they couldn’t be sued for replacement so long as the employees and one board member agreed on it.
Idk. It’s a poorly thought out fantasy of mine to fix how corporations work. I’m sure I’m missing something other than big piles of money that would never let it happen. There are a lot of other systemic legal and structural problems we’d need to fix, but reallocating boards to be 50% voting employee representatives would be a starting point.
The learning curve for quadlets is quite harsh in my opinion. I started with podman compose 3 years ago for my homelab, because it allowed user containers.
I tried to migrate to quadlets unsuccessfully, several times over the years. it was only recently that my self-hosted Qwen was capable enough to figure out where I was messing up and automate the process a bit.
I probably wasn’t sufficiently motivated. It felt like podman compose is basically docker compose, but quadlets are a quite a bit different in form and function, so I was never able to grok them:
deleted by creator
Not that I believe anything the government says anymore, but a shitty reading of Sun Tzu is “pretend to be weak when you are strong, pretend to be strong when you are weak.”
Wait till you start calling her “my first wife” when you introduce her to people. That will really get her jimmies in a rustle.
Not an MD, but that sounds like a problem you might want a doctor to look at. My partner of many years had terrible gallstones that were fairly debilitating. When it got to its worst and she started turning yellow, a few days before she had an episode like you described.
Talk to your doctor about weird painful shits.
As an expert in image processing:
.png supports pretty extreme compression, while a jpeg can also be lossless. The extensions tell you nothing except which family of algorithms was used to encode/compress/store them.
Webp though, webp is only used for the internet. I mean, you could use it other places, I guess.
These look like early ai-generated dog images. Some of them make sense for how the panorama failed, but others just look like a 2 year old drew a dog.
Congrats! Mine also came yesterday. I grabbed the 2 TB with Steam controller. I think you and I were in the same batch of emails!
This was actually my superpower. When I would review my friend’s papers during grad school, as soon as my eyes glazed over or my attention wandered, I would circle the paragraph and tell them to clean it up.
Brain going everywhere is usually a pretty clear sign of convoluted writing, and if you can tell someone to fix it, do.
This was a wild ride.
That 8 hour 20 minute spec for returning to baseline could probably be sped up a bit to improve the inter-frame timing
Modern monitors measure gray-to-gray time as the response, rather than black-to-black. So if she computed the 50% luminescence threshold time on the decay side, and then started injecting the next round at that time, could likely cut the decay time to closer to 200 minutes minutes for the first frame, and then probably double that for the inter-frame times, depending on the GFP decay rate.
Who knows, maybe she already took that into account. She seems to be well rounded.
From the charts in the video (above), it looks to be symmetrical near the peak, but the 50% concentration on decay is around 200 minutes minutes of decay followed by 70 minutes ramping up would put the inter-frame time to be about 3.5 hours.
I realize concentration of fluorescing enzyme may not be 1:1 with lumens emitted.
If this worked the way I imagine, This cuts the total time to play from 600 years to about 200 years. Rounded to 1 significant figure, since this is all ballpark math anyway.
I have my PhD now, and have published 3 conference papers and 2 journal articles. I’m not in academia now.
About half of my academic citations came from a poster I presented at a conference from before I started grad school. A poster that was never uploaded online, has no paper with it, and as far as I know, 15 people have seen in real life.
They were citing the online abstract, which has no real meat to it, just a single picture of my choosing and a paragraph of text or so.
Holy shit man. This project was barely science at this point. The methodology was soooo bad, and only made sense because I barely knew what I was doing and the researcher advising me was a bit out of their depth.
A fellow connoisseur, I see!
Trying to etch models into a chip is a dead end until we reach “peak” quality.
However, unless they include some kind of LoRA (low-rank adaptation) adapter onto the silicon, it severely limits the utility of whatever model or architecture they choose. Being able to modify the weights is way more useful.
Honestly, diffusion decoders are probably where we’ll end up some day. Not end there, but that’s probably the next logical step in the throughput chain.
General purpose compute is infinitely more valuable during times of great software improvements than highly specialized compute.
Things like Tensor Processing Units (TPUs) still aren’t ubiquitous yet, even though they’ve been around for 10+ years. They’re Too specialized to allow for reasonable flexibility on testing.
Did this one achieve some intense introspection just now?
From the way the article describes it, he’s basically trying to assert that the collective bargaining of the union is not the city council’s job, and is the sole responsibility of the mayor.
I guess the argument is that City council shouldn’t be trying to pass laws about compensation of union employees, as it violates the established collective bargaining rules.
Dunno if there’s more to it than that.