I was impressed by Jev, please explain why I shouldn't be
I'm not a computer science person, but I tried Jev and found it impressive for classifying thousands of participant entries by arbitrary themes at near-instant speed and negligible cost, which could help non-technical researchers. On Hacker News, many more qualified people say Jev is a con or a step backwards. I want to know what software did this before Jev, and I need it to classify thousands of entries with arbitrary themes in 0-3 seconds at sub-0.1 cents.
My understanding is that the underlying tech is pretty old and well-established. So if you understand those older models, this is old news. A lot of the negativity is basically "yeah, this is old stuff." But what Jev actually has done that's new and interesting is that it allows you to define a new semantic decision at runtime in natural language, without training a model for that task, and get a fast, cheap, bounded, probability-bearing result designed to be consumed directly by software. In other words, you don't have to fine tune a model. You should if super high quality is really important to you -- but you should also do that with other things that people reach for LLMs for as well. But for the rest of us for whom good enough is, well, good enough, and who aren't interested in training or fine-tuning models? Jev unlocks a lot off-the-shelf. You should be impressed. It's doing something more novel than the haters suggest.
A lot of the hate is in the same spirit of the infamous hacker news comment discounting Dropbox because you can just setup an FTP server to do the same thing. In this case, its people who have built BERT classifiers for years, pointing out you can just fine-tune some model to classify. Of course that's true, and still have value, but its not exactly turnkey. And Jev is surprisingly high quality at what its built for.
BeRT, BART and FLAN-T5, for text classification. All of those models are ~half a decade old now, are based on the transformer model, and small enough to run locally. They're not perfectly SOTA, but perform quite well in embedded applications. Fine for low-stakes stuff, I think. Image classifiers are a hugely competitive space. You've got BeiT, DeiT, MobileOne, ConvNext, FastViT and several more models that all support image classification pipelines.
I frequently have found myself saying, "People pay for that?! But xyz already exists," in the past, but overtime I've learned that (1) marketing is important and (2) if something reduces friction enough to get the wheels turning on something you're trying to achieve, there's always someone (actually a lot of people) out there willing to pay for it even if it's measurably worse than a self-scaffolded solution or even a really good solution that requires a lot of tinkering to get things rolling.
As William Gibson said, the future is already here, it’s just not evenly distributed. So people get excited about existing technologies because they are new to them. Wrapping an existing technology in a new package is a common way to introduce it to new people…there are many wrappers around ffmpeg, html2pdf, etc. that are marketed as products. And they seem like magic…hell cassette tapes feel like magic to me even though I had a cassette player in 1973. Often but not always, the best tool is the one in your hand or the one you know of because then you can just get to the work you are trying to do. First make it work, then make it good. Good luck.