[selestify]

> It feels like cooking now requires commercial quality equipment only available to the Michelin star restaurants.

Why does AI make you feel this way? It feels the opposite to me — like meals that formerly required a master chef to make, but soon anyone can make for themselves at an acceptable level of quality with meal prep kits

[speff]

A few months ago there was a post on HN about catching up to the current state of LLM dev and learning how to use it. In it there were recommendations for hardware - the lowest tier being a 3090/4090. When looking at the decision tree for even cheaper options, it basically said to find another hobby.

Not sure if that's changed but ever since seeing that line, I've been put off of that world. I still occasionally click on HN links advertising new methods which can be run on "consumer" cards and every time it's just a 3090/4090...

I don't have that much money.

[ParetoOptimal]

For training or inference?

The p40 was on the inference side of the tree and you can get one on eBay for $200 or less.

[kube-system]

If we're still entertaining the above analogy, a home kitchen costs more than a PC with a 3090/4090.

But when I got into software in the 90s, it was about $4000 for just a run of the mill desktop PC, which is about $8000 in today's money. And it didn't even have 3d acceleration.

In the grand scheme of things, a 3090/4090 is not expensive.

[yurishimo]

A gas camping stove is like $50. A toaster oven is another $50-100. The best reviewed chefs knife on the market for commercial kitchens is $30.

You don’t need thousands of dollars to start cooking.

I have friends in the who cook on this stuff at home all the time. We in the west have elevated our kitchens into something luxurious. Nothing wrong with that, but it’s overkill for many people.

[kube-system]

You can get started with cooking on a gas camping stove like you can get started with AI projects on a Coral Edge TPU accelerator.

[apwell23]

My feeling is that only big tech can now create any good apps since they all require massive compute now.

Yes writing a todo list app is now faster with copilot and stuff but thats now what i meant.

[anymouse123456]

Please don't be discouraged by whatever "Big Tech" is doing.

Products that we build today represent many millions of decisions and trade-offs.

If you believe something should exist, please build it and don't worry about what anyone else is doing or saying.

I was working on an application in the early 2000's and learned that Gigantic Inc. launched something to solve the same problem. I immediately bowed out, and they immediately let that first-launched beta languish for more than 10 years with less staff than we had.

In the intervening years, another startup built a similar competitor and sold for hundreds of millions of dollars.

As another example, I eventually worked at Gigantic Inc, and the org that I was part of had been failing to deliver a useful product to the public for about 7 years, and continued to fail for many more. This was an organization with hundreds of people and many hundreds of millions of dollars of budget, and they were being absolutely clobbered by a combination of their own unbelievable incompetence and the brutality of the market around them.

One of the biggest lessons I've learned in my career, is to never assume that just because Big Tech has some budget, that they also have attention or competence.

[JohnFen]

> since they all require massive compute now.

They do? How so? I can only think of a small subset of applications that require large amounts of compute.

[apitman]

Interesting, I actually feel the opposite. The scale at which big tech apps have to operate leads to a lot of bad UX and more bugs due to complexity. For example you can typically get much lower latency hosting a web app at or near your home than Google can provide.