What I would really want is for this to be a hardware button that, without requiring input on the phone itself, starts and stops a recording on my iphone which uses my airpods as a mic.
This would actually be super helpful in the lab, dictating notes on a protocol ("I did something weird in this step") without needing to stop to write things down (sometimes protocol is quite time-sensitive).
Not really. If you know about uv, you know how to use "uv tool run", so you know how to use any formatter of your choice (which you can find easily on Google, arguably easier than reading the documentation and learning about uv format).
Well, it is arguably worse to run an unknown, not version pinned, unconfigured formatter over your code and expect it to improve things, unless the code is an utter catastrophe in terms of formatting.
_You_ may find it irrelevant, but speak for yourself. I don't want dependencies, that are not version-pinned and checksummed running over my code. It is certainly not irrelevant to me.
for doing laboratory work in my PhD, I've found no better app than OmniFocus. It's particularly valuable in its ability to create tasks via a templating system. This is crucial, for example, for managing 10+ genetic crosses at a time. Each cross takes weeks to move to the next step, but when that next step occurs, I need to be on top of the cross 2x / day. Juggling different crosses at different stages would be impossible for my brain without a system I can rely on. Other lab work follows similar workflows.
Instead of writing the counting tool he could have used the Multi-Point Tool in ImageJ [1] [2]. I used it just this morning for counting some embryos I collected.
It sounds like this may have been one of the pieces of software the author intentionally chose not to use:
> There are some clunky old Windows programs, niche scientific tools, and image analysis software that assumes you’re trying to count cells under a microscope...
C. elegans is nice for this since you can freeze stocks in glycerol. Labs routinely go and thaw out the main wild-type reference stock if the lab stock has been around for too long.
Now I'm in a fly lab and no one's really figured a good way to freeze a fly stock down for long-term storage. So we're left to just accept some degree of background mutation and generally assume that it's not impacting our experiments too much...
It's worth noting that we've found genetic differences between the N2 wild type strains used by different labs as well, so this is still a problem for C. elegans.
Before these companies like plasmidsaurus that do whole-plasmid sequencing for relatively cheap with nanopore, people generally only sequenced a region of interest using sanger sequencing. The rest of the plasmid was assumed to be mostly correct, as long as it grows on a bacterial resistance. As noted in the article, the rise of nanopore-based whole-plasmid sequencing has reduced a lot of these types of errors.
Would it be possible to consider them separately though? Like maybe it will turn out that say 10% of them are beneficial, 65% of them are neutral (either they do nothing at all or a mixture of benefit and harm), and 25% are slightly bad for us (can't be too harmful or we would have already known ig).
Delivering gene therapies into brain cells is a non-trivial task. Also, there's alternatives to cutting the original sequence out; you can also dampen the transcribed RNA with downstream therapies.
'Bad' is notoriously hard to figure out. It might be good for the group to have a few people with major psychiatric disorders even if it's not ideal for that individual or the people who have to directly interact with them.
From my experience, the thing that makes using AI image gen hard to use is nailing specificity. I often find myself having to resort to generating all of the elements I want out of an image separately and then comp them together with photoshop. This isn't a bad workflow, but it is tedious (I often equate it to putting coins in a slot machine, hoping it 'hits').
Generating good images is easy but generating good images with very specific instructions is not. For example, try getting midjourney to generate a shot of a road from the side (ie standing on the shoulder of a road taking a photo of the shoulder on the other side with the road crossing frame from left to right)...you'll find midjourney only wants to generate images of roads coming at the "camera" from the vanishing point. I even tried feeding an example image with the correct framing for midjourney to analyze to help inform what prompts to use, but this still did not result in the expected output. This is obviously not the only framing + subject combination that model(s) struggle with.
For people who use image generation as a tool within a larger project's workflow, this hurdle makes the tool swing back and forth from "game changing technology" to "major time sink".
If this example prompt/output is an honest demonstration of SD3's attention to specificity, especially as it pertains to framing and composition of objects + subjects, then I think its definitely impressive.
For context, I've used SD (via comfyUI), midjourney, and Dalle. All of these models + UIs have shared this issue in varying degrees.
It's very difficult to improve text-to-image generation to do better than this because you need extremely detailed text training data, but I think a better approach would be to give up on it.
> I often find myself having to resort to generating all of the elements I want out of an image separately and then comp them together with photoshop. This isn't a bad workflow, but it is tedious
The models should be developed to accelerate this then.
ie you should be able to say layer one is this text prompt plus this camera angle, layer two is some mountains you cheaply modeled in Blender, layer three is a sketch you drew of today's anime girl.
Totally agree. I am blown away by that image. Midjourney is so bad at anything specific.
On the other hand, SD has just not been on the level of the quality of images I get from Midjourney. The people who counter this I don't think know what they are talking about.
previous systems could not compose objects within the scene correctly, not to this degree. what changed to allow for this? could this be a heavily cherrypicked example? guess we will have to wait for the paper and model to find out
We introduce Diffusion Transformers (DiTs), a simple transformer-based backbone for diffusion models that outperforms prior U-Net models and inherits the excellent scaling properties of the transformer model class. Given the promising scaling results in this paper, future work should continue to scale DiTs to larger models and token counts. DiT could also be explored as a drop-in backbone for text-to-image models like DALL E 2 and Stable Diffusion.
Afaict the answer is that combining transformers with diffusers in this way means that the models can (feasibly) operate in a much larger, more linguistically-complex space. So it’s better at spatial relationships simply because it has more computational “time” or “energy” or “attention” to focus on them.
This would actually be super helpful in the lab, dictating notes on a protocol ("I did something weird in this step") without needing to stop to write things down (sometimes protocol is quite time-sensitive).