Thank you for the detailed response! I will keep these things in mind

The choice of algorithms usually comes down to your goals and your modeling assumptions. For example, if you think the data is drawn from a Gaussian distribution, then you would use mean squared error. If everything roughly lines up on a plane or different labels can be separated easily by a plane, then you could use a linear function and related tools for optimizing them.

We usually just make educated guesses by inspecting the data, but when you have messy real world data, that's hard to do correctly, so it makes more sense to just try a set of things to see what sticks. Note that this doesn't mean trying everything under the sun. You should still understand your problem well enough to restrict that set to a reasonable size.

(I started as a data science years ago, but quickly turned data engineer and have definitely lost touch, but I can help with some generic answers: )

"Brute forcing every algorithm" sounds a bit bad, but it's usually not that bad: in reality when you're "done" you probably want to retrain when new data comes in every so often, so trying out a few algorithms isn't too big of a waste.

That said, with experience comes some knowledge about the tradeoffs of each algorithm/model. For instance, neural nets are universal approximators, so they can in theory model anything, but they need a lot of data to make this happen. In addition, NNs are bad for explainability : when you want to tell e.g. a customer why you're recommending them xyz movie, "neuron 231 activated strongly" is not really useful.

While a linear regression can in theory work with just a handful of samples, but is only useful if your data actually roughly matches a line. This is really good for explainability however: if you see that particles with high energies are more likely to interact, then your linear regression will show that, and it's a clear correlation between those variables.

Apart from amount of training data needed and explainability there's lots of other variables that play a role like how fast the model decays (e.g. anything trained on news data may be outdated in days if not hours, while physics doesn't really change) and as such how often it needs retraining; how easy it is to update vs retrain from scratch each time; cost (in time usually) of training and inference; what you're optimising for (false positives, false negatives, accuracy, precision, etc).

It's not uncommon to just shotgun a bunch of models (and a bunch of variation of hyperparameters of those models) at a dataset and just pick whatever seems to perform best on a validation dataset.

This is actually one of the few good usecases of complex parallax js heavy website.

This is an art piece. It's not an actual product. The website is also a part of the piece.

This is an actually deep art piece.

Best thing I've seen today

Have you looked at the technical info? There's a write-up*, videos, photos, code and CAD files, plus another domain with the team behind, including author links to their LinkedIns which take about the company+project, plus a Google Scholar page which all confirm that they worked at Google Research. It's possible it's real people but a fake project, but that's not where Occam's razor falls for me.

* I think it's non-LLM, despite some em dashes)

I'm sure you'll get a break once all questions have been answered

I think I'd rather do anything but crank all day

At the bottom of the page: https://squeezlabs.github.io/handcrank/

Lists exactly how it's made

If they put these in datacenters, we would have a surplus of jobs. Just saying.

I’ll modify one to connect it to my concept2 rowERG. It’ll synergize with database prompts

Guy with one beefy arm can now use the excuse "I run a local AI chatbot" as an excuse

The Guinness Book of World Records credits Gareth Griffin of Harrogate U.K. with the record for growing the world's heaviest onion at 8.97kg or 19lb, 12.4Oz.

But it seems they need to update their record given the size of the onions people are eating in this thread.

I'd like to look into it more, but this site is frustrating dogshit and has no actual useful info.

I wanted to know:

Cost Release/availabilty effort to make and store tokens

Write it in fucking text. A forced slideshow and videos doesn't make me want the product more, it makes me frustrated with your company, and gives me a barrier to looking into it in the future.

The section with the three levels of generation make it seem like a joke, but it does appear to be real, with a lot of work gone into reducing power consumption. Unless it is actually a joke, and they've just committed to the bit with the technical info.

But their students use it and they freak out?

Slopademic research

Too long did not read;

But I am doing my masters in ML so I could understand a few concepts here and there. so I will provably check it out in a few months.

Right?

Wtf?

Accidentally deleted a production database from a staging environment

Then numerous people didn't do their fucking job.

No. This was an astonishing amount of fuck ups from those in charge when deploying this crap that translated into a ridiculously funny moment for everyone looking at the IA madness. From this article: https://www.tomshardware.com/tech-industry/artificial-intelligence/claude-powered-ai-coding-agent-deletes-entire-company-database-in-9-seconds-backups-zapped-after-cursor-tool-powered-by-anthropics-claude-goes-rogue

Briefly, the cloud provider's API allows for destructive action without confirmation,

Bruh?

it stores backups on the same volume as the source data, and “wiping a volume deletes all backups.”

Bruh!

Crane also points out that CLI tokens have blanket permissions across environments.

Bruh...

I mean, dude... are you blaming the knife for being too sharp when you use it to stab yourslef?

Makes sense as they are programmed to be affirmative and not to avoid interactions.

We put the wrong people in charge time and time again and enriched the wrong people with capitalism.

Humans as a whole aren't the problem. It's the systems we continue to prop up.

60s sci-fi "predicted" that we will settle Mars by 1980 and have interstellar colonisation program by 2000.