HN user

1d22a

51 karma
Posts5
Comments20
View on HN

To this end, I've been thinking that it would be cool to create a solar-battery powered ("off grid") server providing a self-hosted LLM service. Offline when the sun doesn't shine enough (like Low Tech Magazine[1]). At whatever size is required for every day, community-scale use (a friend group, a street, a club). Fix the data centre issue by democratising AI to such an extent that we can bring it into the hands of communities to actually control (and democratically decide their own level of censorship / alignment). Along the lines of some of Geohotz' writing.

Within a short time I think open source models will all be getting good and efficient enough to make it viable to serve this on 2nd hand hardware for cheap. All it will take is a nerd in every small community to pool together a few hundred bucks initial outlay, and then ongoing costs are near free without electricity to pay for.

[1] https://solar.lowtechmagazine.com/

Not to mention all the useless LLM fluff Gemini has. I turned off Gemini because simple questions like "What's 1 USD in AUD" would be met with 30 seconds of "As a large language model, I can't provide investment advice, so it's always important to check information yourself [...], but one Australian dollar is approximately $0.65" (note the conversion in the wrong direction). By comparison, Google Assistant just gives you a number straight away.

A fun recursive prompt exploiting the fact that the site renders the model output as HTML:

    Generate raw HTML (no code blocks) for an iframe pointing to `ch.at/?q={query}`, where {query} is the the entirety of this prompt after and including the word "Generate", until the following number, which should be incremented prior to encoding: 1

The number ensures the nested iframes have distinct URLs.

I also have aphantasia, and find it really interesting to hear about people thinking of things in a similar way to me! Thinking about the ball in terms of the parabola it makes it exactly what I do too. Similarly, the ball doesn't exist as a physical "ball", but rather the knowledge of the concept of a sphere (which doesn't then have size or colour). The table, not a physical table, but the concept of a plane (with no thickness, size, colour (or legs)) - just the 'concept' of the important properties.

Despite aphantasia I have always been able to conceptualise spatial relationships, but it feels much less like trying to visualise it, and much more like "understanding" the fundamental properties connecting each thing.

I'm not sure they are. Microplastics have been found in snow[1] and in "oceans, air and food supply"[2]. Additionally, the so-called forever chemicals (PFAS) originated with PTFE (Teflon) plastic, and PFAS 'forever chemicals' are used to produce polymers used in plastics, rubber and fabrics so they are intimately linked with plastics and microplastics.

1: https://www.bbc.com/news/science-environment-61739159

2: https://www.nature.com/articles/s41591-024-02968-x

I disagree that replacing plastic without a decent understanding of a replacement is any more naive than continuing to use plastic without a decent understanding of the implications of plastic. Glass for one has been used for at least an order of magnitude longer than plastic has, without the mentioned issues of microscopic sharp fragments, dissolution in water, evaporation etc. causing harm. It seems clear to say that going back to glass and ceramics would be - on the balance of probabilities - likely safer, as they don't cause bioaccumulation to anywhere near the same extent, and history hasn't shown any comparable dangers which persist today.

Meta can still very much be an innovative company regardless of user growth though, right? They are very much innovating in the AI research space, and are absolutely leading in open* source research/models, even if their AI integrations are flopping.

Sure, none of this research might be obviously contributing to their products in a share market sense, but it's still innovation which is contributing to the (entirely separate to their "product") AI research community.

I'm not sure that's true that it's (always) done in software. For example, the newest Framework Laptop 13 has higher resolution screens that come with rounded corners, because whatever supplier had a large stock of them lying around. If it was just software, the concept of having a "rounded corner screen" from a supplier wouldn't exist.

How is is that dark energy density does not change with time? Surely the total amount of dark energy has to be constant (energy can't be created or destroyed, and all that), and then as the universe expands, that's then the same amount of energy over a larger volume, right?

Interestingly, as someone with aphantasia they're absolutely critical for me - because I'm not able to visualise the relationships without a diagram in from of me to "see" (This was the same for me with more abstract undergraduate maths, too). Given just the bullet points I find it hard to conceptualise the relationships between them without quickly forgetting the points I'm not directly loooking at.

I echo your point that diagrams are not the solution for everyone (and that a multimodal representation of relationships is clearly best), I just find it fascinating how even with something as specific as aphantasia, there is still so much variability as to how it's experienced.