JBorrow

I’m not so sure. For me a reliable way to find a “good” dentist is to find one that’s attached to a (big) medical school or university. Of course this is easier to do in cities than in rural areas though. I’d always take a waiting list and, say, a dentist from the UC system than one that’s a regular practice.

Thankfully there are a few amazing Diamond/Platinum open access journals popping up (that are often ‘arXiv overlay’, meaning they simply provide peer review services to arXiv-hosted papers). These journals are free to publish and free to read but still provide the useful categorisation/review/cataloguing services of traditional publications. Notably this includes a post-review DOI.

Relevant for the HN crowd is the Journal of Open Source Software: joss.theoj.org.

[I am an editor at JOSS]

Calling PostgreSQL a 'legacy C codebase' is funny.

I don't think the only utility of a depth model is to provide synthetic blurring of backgrounds. There are many things you'd like to use them for, including feeding into object detection pipelines.

They use a distributed data management tool called RUCIO: https://rucio.cern.ch to distribute data on the grid.

At some level this is an almost uniquely American problem; see similar statistics here from Great Britain: https://www.gov.uk/government/statistics/reported-road-casua..., Australia: https://en.wikipedia.org/wiki/List_of_motor_vehicle_deaths_i... which have been dropping rapidly since the 1970s. So, relative to other very similar countries, the U.S. is doing much worse.

Models like AlphaFold are very different beasts. There's definitely a place for tools that suggest verifiable, specific, products. Overarching models like 'The AI Scientist' that try to do 'end-to-end' science, especially when your end product is a paper, are significantly less useful.

As someone 'in academia', I worry that tools like this fundamentally discard significant fractions of both the scientific process and why the process is structured that way.

The reason that we do research is not simply so that we can produce papers and hence amass knowledge in an abstract sense. A huge part of the academic world is training and building up hands-on institutional knowledge within the population so that we can expand the discovery space.

If I went back to cavemen and handed them a copy of _University Physics_, they wouldn't know what to do with it. Hell, if I went back to Isaac Newton, he would struggle. Never mind your average physicist in the 1600s! Both the community as a whole, and the people within it, don't learn by simply reading papers. We learn by building things, running our own experiments, figuring out how other context fits in, and discussing with colleagues. This is why it takes ~1/8th of a lifetime to go from the 'world standard' of knowledge (~high school education) to being a PhD.

I suppose the claim here is that, well, we can just replace all of those humans with AI (or 'augment' them), but there are two problems:

a) the current suite of models is nowhere near sophisticated enough to do that, and their architecture makes extracting novel ideas either very difficult or impossible, depending on who you ask, and;

b) every use-case of 'AI' in science that I have seen also removes that hands-on training and experience (e.g. Copilot, in my experience, leads to lower levels of understanding. If I can just tab-complete my N-body code, did I really gain the knowledge of building it?)

This is all without mentioning the fact that the papers that the model seems to have generated are garbage. As an editor of a journal, I would likely desk-reject them. As a reviewer, I would reject them. They contain very limited novel knowledge and, as expected, extremely limited citation to associated works.

This project is cool on its face, but I must be missing something here as I don't really see the point in it.

> Over time, as both Spawn and the underlying models improve, it will be able to build more complex software.[^Citation needed]

It seems to me that with the transformer model world we now live in, utilities like these are excellent at generating things that appear in many tutorials/posts/etc. However, there's little-to-no evidence that they are able to generate anything truly custom. For instance, any application requiring any significant level of domain-specific knowledge (your last point), and given the model's architectures there's no reason to believe that will occur. What is it that makes you so confident that will become a possibility?

I would be interested to hear what fraction of this script and README were generated by large language models. At first glance, the code contains a number of repetitive anti-patterns that 'feel like' they are Copilot-isms (e.g. large stacks of elif statements instead of using appropriate data structures), and the README is very verbose and includes a high fraction of filler words.

Buses are only slow and inconsistent because of cars.

This release coincides with the SciPy 2024 conference and a number of other things. I would suggest reading the article to learn more.

Sure, but you may want to use different databases as your application may be expected to be deployed in different environments. I like to keep some of my apps db-agnostic, so I test with postgres in a testcontainer and sqlite outside of one to ensure consistency.

Yes, I use this kind of setup in a number of places. I am also a fan of running the full test suite using SQLite (if you are using a compatible ORM) to make sure where you are (or not) running into DB-specific behavior.

What are 'people like me'? It's not like I know nothing about large language models, I just think using them for civil engineering is a bad idea...

Boy, I can’t wait for the foundation of my house to disappear because the LLM mis-classified a drilling report as non-hazardous.

What’s the deal here with liability and accountability? That’s a serious problem when considering using these for anything other than toy problems.

Well, that is only true (and helpful) if it is a dataset that benefits from being in a tabular format.

Consider, for instance, particle-based simulation data (loading that into any database is a waste of time...), or a set of images that you need to perform non-standard individualized processing on (again, a waste of time).

Stuffing stuff into postgres or clickhouse is great for your typical 'data science' workflows where data is consistent and the problem is 'oh no our transaction volume has increased 100x'. But in other (some would claim more interesting) cases, using database systems is unhelpful.

There would potentially be some flexibility, also in start date. We are generally used to the academic hiring cycle where acceptance of candidates happens 6 months before start dates.

University of Pennsylvania | Full-time | (Senior) Software Developer, Front-end | In-Person | $55-$95k (3 year term)

We are an academic department hiring a primarily front-end developer to work on data visualization, analysis, and sharing tools for the Simons Observatory.

The Simons Observatory (SO) is a millimeter-wave survey experiment currently being built for observations from the Atacama Desert in Chile, with a variety of scientific goals from probing the early universe to detecting millimeter-wave transients.

SO is a medium-size academic collaboration with around 500 members. The Data Delivery group is a small team composed of around 5 people working on tooling and deployment.

https://wd1.myworkdaysite.com/recruiting/upenn/careers-at-pe...

This is a correctness nightmare.

Whether you like it or not the government does tell you how much your house is worth through monetary policy.

This isn't a benefit of go, but rather a drawback of the counter-example of typescript... All tools generally designed to work for creating small utilities ({ba,z,...}sh, python, perl, go, swift, ...) have this feature.

An ORM is not too much pre optimisation…

I really loved my scanwatch but there was two major problems:

The hardware (at least gen 1) has around a 1 year lifespan due to the quality of the seals. You can only get a warranty replacement once. Mine literally fell apart once when I put it on, then the other one was shorting out in the case.

They don’t allow heart rate sharing over Bluetooth. Kind of a niche feature but this stops you from using it as a heart rate monitor with a bike computer or other exercise equipment.

In the end, after the support refused to service my watch for a warranty repair, I just switched to a Garmin Venu 3S, which I like, but I do miss that semi-analogue design.

Well the 235 is nearly a decade old now, there has been much improvement in heart rate tracking since then. The 235 is as old as the original Apple Watch.

On a Mac you can do this rebinding entirely natively within the system keyboard settings.

It seems to be a competitor to Amazon FBA

Did you read TFA? That is a core part of it.

All the 10x engineers don't exist for the next couple of hours.

It could be just me but my Instagram Reels are all of like random thirst traps for Indian dudes with 4 likes on each post...

(Yes, I am aware that the 'algorithm shows you what you engage with' but no matter how many times I click 'not interested' I seem to be in Indian thirst trap hell)