Except… it costs a huuuuge sum of money to transform that tranche of inert information into a totally different useful form (an LLM).
That process of transformation, the expertise required to enable it, and the cost of then making it available to users is what is being charged for, no?
> The issue comes when taking information you haven't paid for, and then locking it into a machine you charge for.
This is two distinct complaints:
1) Using information they've not paid for
2) Charging for the outputs
You also wrote:
> It would be an entirely different story if OpenAI was indeed open and the resulting model was available for all.
This implies that you don't have a problem with #1 (using information without paying) but you do have a problem with #2 (charging for the outputs).
So my point wasn't that #1 is okay (as much as I dislike copyright law as it stands, corporations should follow the law, just as individuals are forced to) but rather than I find #2 to be okay, in isolation.
And how much did it cost for the authors to produce the books in the first place? Everyone wants to believe that piracy is okay because the creators costs don't matter.
> And if the value is correcting misunderstandings, I don’t see it. The very fact that the communication failed and there was a misunderstanding is a huge no hire red flag. Communicating clearly matters as much as doing the job
What if the miscommunication is 50:50, rather than just the candidate’s fault?
Working in an international organisation I became aware that people from Europe vs. the US speak a different corporate language. At its most extreme, smart experienced colleagues from the US would say things and I literally wouldn’t understand their meaning.
However, colleagues better versed in that particular corporate dialect would absolutely understand their meaning, and might think what they were saying had great value.
And the flipside is that I must assume that at times, my (clear direct, I think) European communication could also be misunderstood.
Which is to say: a miscommunication in an interview might be as much your fault as theirs; and you may reject excellent candidates based on this.
Your prefilters only lasting two weeks, plus two household members developing asthma, might make me think very seriously about moving somewhere less polluted, and quickly.
I was using Windows on my personal-but-for-work laptop; it got to the point where I hated it too much so tried Linux. Most stuff was great (even Teams was 90% fine!) but I couldn’t function professionally without installed versions of Office, and I couldn’t find an acceptable workaround.
I then switched back to MacOS and am mostly happy hanging on to Sequoia… but I’m dreading the point at which (for some reason) Apple make it unavoidable to upgrade.
> I then switched back to MacOS and am mostly happy hanging on to Sequoia… but I’m dreading the point at which (for some reason) Apple make it unavoidable to upgrade.
MacOS 27 is a lot less shit than 26, thankfully. I upgraded from 15 to 27 directly (on a M-series Mac).
IIRC in Linux you can run the windows version of office via wine or one of the successor/forks. Not sure if AD integration works nicely. And maybe it is just a group policy that somehow blocks it. But running the software should be possible these days.
This is the billion dollar question for corps though. Windows and MacOS are first class citizens of the corporate IT universe and all the weirdness of corporate IT works. MDM, identity/auth, Office, LOB apps, video, document signing, data labeling, everything.
Linux has taken huge, heroic strides, and has caught up to about 2005. Corporate IT simply will not sanction Linux desktops unless they are provably as well versed in the entire ecosystem as Win/Mac, barring a "political" decision made at the top to make it happen.
To change this would require a profoundly deep, expensive, and continuing investment in all aspects of the 'Modern Work' experience, plus genuine commitment from Microsoft to make it happen (which would likely require lawsuits and enforcement).
The situation today is that a few orgs are playing catchup on some features, but that will never be enough to match a platform that has been dedicated to corporate life for 40+ years, and from a company infamous for extending their advantages to extinguish the competition.
AFAIK, Wine/equivalent only works for older versions (up to 2016?) - newer versions +/- AD require a VM of some variety. (This was originally where I noped out and switched to MacOS instead.)
Small bug, Firefox/MacOS: if I search for a location, use the suggestions that appear to select it, remove it by clicking on the x button, and then try to search for a different location, the suggestions no longer appear. I had to refresh the page to get them back,
I've played with diffusion models on and off since the first release of Stable Diffusion - just for amusement, without a particular goal.
Recently, I've been helping a friend's wife with some basic vector images for her sewing hobby (she has what is essentially a CNC sewing machine) and have been super-impressed with FLUX.1-Kontext, which I've been running on my Macbook Pro with mflux. Its ability to (for example) take a photo of a human or an animal and return a line drawing which is recognisably them (rather than just a generic similarish image as I've experienced with other models) is excellent.
It's an older model now, but (AIUI) has the text-handling features baked in, and in my various testing is very reliable at giving me the outputs that I want, without the randomness I've experienced previously. It's big and relatively slow (~3 mins per 512x512 image edit on my M1 Max Mac) but excellent to work with. It's also very straightforward to set up, without the harness complexity of e.g. comfyui.
Off the shelf - it’s a Brother. It prints via a proprietary file format (.PES) but there’s an extension for Inkscape that supports creation and export.
This is currently the weakest part of the workflow. I'm tracing the output images with VTracer [0] and then finishing them manually.
Through a combination of careful initial prompting (e.g. requesting line drawings, requesting limited colours ["posterization" works] if you need colour at all, requesting background removal) and then using the settings in VTracer to limit the colours used for the resulting SVG to the minimum needed, you can end up with an SVG that doesn't take too long to clean up in Inkscape. It usually takes a few minutes of deletions and path unions to produce something usable.
Diffusion models outputting SVGs would obviously be very interesting for this, but this is probably a small niche compared to everything that diffusion models can do with raster images, and I've not found anything usable online so far. LLMs (understandably, given their core technology) are still pretty bad at creating SVG art.
Claude and I have spent most of the summer on exactly this problem for Skechworks (skechworks.com), a free little Mac design app I made to help my own workflow that is very similar to yours. I use it for my custom coin shop The Achieve Mint (theachievemint.com).
It runs VTracer under the hood too. Most of the cleanup you're doing in Inkscape could possibly be pushed to before and after the trace?
Skechworks picks the palette before the trace and forces it again after. I quantize to 8 colors with pngquant, trace, then snap every fill back to the nearest palette color. VTracer fills each region with its own average color, otherwise you get hundreds of nearly identical grays. One design went from 279 fill colors to 3.
I upscale 2x before I quantize. The anti-aliasing is what tells the tracer where an edge really sits. Quantize at 1x and you throw that away, and small text comes out lumpy. Related gotcha: filter_speckle is an area, so don't double it when you double the image.
I also flatten transparency onto white first. VTracer counts every alpha value as a different color. I had a 5 color picture that it saw as 688 colors, and the trace took 6 GB of memory. Flattened, the same picture took 300 MB.
For the background I sample the whole edge of the picture. If 85% of it is one color, I delete every path in that color that touches the edge. White inside the drawing never touches the edge, so it stays.
For black and white I threshold with Otsu, then trace in color mode with hierarchical stacked instead of bw mode. bw mode hands you one giant black path full of holes. Stacked gives you every region as its own shape, which is much easier to edit. You have to pad the image with the background color first, or VTracer decides your subject is the canvas.
Your FLUX step is the same trick I use for photos and messy art. I have an image model redraw the picture as flat art, then trace the redraw. That came out about a quarter the file size of the best direct trace I could get. The catch is that the model moves and resizes the subject a little every run. I measure the box around the ink in both pictures and fit the trace back onto the original.
What I haven't solved is curve fitting. vectorizer.ai gives you real arcs and straight lines. VTracer gives you splines, so circles are never quite circles.
I agree to a degree - it feels manipulative on the part of the writer, trying to exaggerate any sympathy a reader might feel.
That said, I also feel that the level of negativity the lady suffered was a little out of proportion to her 'crime' - essentially amounting to bullying.
This must be a personal preference based on ear anatomy.
The first-gen Airpods fit my ears perfectly, and were secure. When I lost them and upgraded to the 2nd gen, I was initially upset - aside from the body being fatter (and so less comfortable in my ears) they were very insecure - I couldn't even walk around without them loosening and dislodging.
I was on the point of returning them when I saw the advice online to 'twist' them into a different position. I now use them with the stalks horizontal, and they're secure. Still a touch less comfortable than the v1s, but easily tolerable.
reply