It depends, sometimes the private API models are better at a specific task to the point it wouldn’t be worth using one that is worse. (Really it’s the harnesses around them tbh)

Overall sovereign AI is where it’s at. If you have good enough tooling in place

Im surprised that the EU doesnt have a secuirty requirement for IT infrastructure in the food supply. Seems like an oversight

Id be done for a “it just works” day to celebrate all of invisable labour that makes the world work

First find a way to track your costs. Measure how much it costs to give it away (to your self and people that support you starting it as a buisness). Do market analysis based on those number. How much do you need to support both your operatiinal costs of hosting and devoloper time paid with a fair wage, and what does compare to how much consumers are currently paying. If that looks good invest in an accountant and lawyer to see what the buisness plan viability might look like and set up a LLC. From there look for partners that can help you create a sustainable buisness and fill in skill gaps you dont have, or hire out the rest.

Thats it.

Even better make it a community coop so that the skilled labor and community can have actual ownership in the org to help balance out thr risk and encourage litteral buyin.

https://www.scientificamerican.com/article/openai-unleashes-hundreds-more-math-results-upon-a-field-already-in-shock/

This article seems was a little easier to get through for me

Its a real the boy who cried AGI problem. They are doing constant bullshit hype to keep investors so when they announce anything the assumption is bullshit hype. It doesn’t help that their product is really really good at writing bullshit hype.

Is there really a market for mass production of low quality software?

Maybe it’s just never been needed in any of the market segments I work, but it always seems like it’s better to spend 40 hours today to refactor or clean up code then to spend 10000 hours bug hunting spaghetti code. Like sure your perl script worked but for the love of all let’s actually make a project with a maintainer and a code base you can reasonably attach an SLA to plz

Home and family: Has working nodes

Hell yeah, I’ll take that assumption

Me too

We are saying the samethings but you are adding no to it.

Right, during inference random numbers are generated, that plus the numbers from input are added to the weight values and the matrix multiplication happens. If you used the same random numbers and inputs it is deterministic. For regular use you don’t do that because you want a stochastic output, if you wanting to do forensics and trace what led to an output you would benifit from that determinism.

This to me is one of my critics of wilderness areas (I love them otherwise), without having a place for people to be apart and worth with the rest of the natural ecology of an area it’s too easy to see avoiding active harm as an optional charity instead of a part of maintaining a living biosphere

Honestly for physical therapy if these are built right a huge amount of people could benifit from this. For people with no treatment options too that just need assistive tech.

No that should be discoverable with the models weights, input and random numbers added to the weight at the time.

Say for example you find that a collection of outputs behave oddly or in an undesired way, you could use this to find what simularties they share with each other but delta with other and naively prune the nodes or simply decrease their weights. Those you could also try to corralate that to certain input tokens to engineer better prompts or try to trace it back to initial training data.

It could also be a failed tool call adding garbage data in, or malicious. A trace could catch that as well.

EVs and electricity are both easier to do DIY at home than ICE and homemade ethanol (and no one even pretends home mode petrol production is remotaly feasible). The only reason EVs are seeing the subscription model harder than ICE is: a they are newer and it’s new and b. They are HARDER to lock people into bullshit. No sweetheart deal with a dealership for biannual fleasing by lube techs they are ripping off, no guarantee that in 3-5 years the rube Goldberg explosion machine and the rest of the car are going to be so shitty that a new car makes sense.

Slate, Edison, and Mutany Motors are all examples of what new world of EV could be like instead of the bull shit that ICE and oil industry and kept us in for nearly a century of enslaving humanity.

Also also, e-motos and e-bikes because not everyone needs a ton of metal to move around cities if they are built for people and not just to extract wealth from suckers.

It’s not really coin flips it and random adjustments of weighted values that decides which nodes are activated which, plus random adjustments, influences the next node activation until out

First agreed. Honestly I haven’t an LLM do anything close to an engineer. You can get lucky for a code snippet or two and maybe a few ADRs but even those are fraught with the chance of slop clean up work. I am not arguing that.

The node is only probalistic because of the random number added. After the fact the number is known.

You can try to reason about it which is an arguable improvement from the former engineer you can’t get ahold of anymore. So you can could see that it triggered a node associated with c++ when it generated this string of tokens or compare it other generations to map it to a type of call.

Where it doesn’t compare to a human engineer at the moment is quality or novelty.

What I mean is for post hoc analysis since you know the coin flips and the inputs you can step through the layers and nodes calculation deterministically.

Neural nets are dumber than you thought for sure. What I mean is the coin flip results are known after the fact.

I mean probabilities in matrices are nodes in a neural net, right?

The only thing that makes it non-deterministic is the tempature value which is known after the fact from my understanding, so that should be able to deterministic.