ArchRecord

I don’t know of any super high-quality ones that run well, but the Open Assistant project, (now archived) collected responses from voluntary participants (myself included) to build what is now considered a very high-quality dataset of chat conversation pairs, truly open source, and all voluntarily submitted instead of scraped.

The models are reasonable for fine-tuning, but aren’t very good compared to newer models from large companies.

I find those kinds of chatbots useful, but those aren’t the ones I encounter 90% of the time. Most of the time, it’s a chatbot that summarizes the help articles I just read, giving faulty interpretations of the source material, that then goes on to never direct me to a real person unless I tell it multiple times that the articles it’s paraphrasing aren’t helping. (and sometimes, they have no live support at all, and only an LLM + support articles)

TLDR; “weak people have to conform to social consensus otherwise they get hurt I guess?”

I, on the other hand, am a big, strong, high T alpha male, that isn’t worried about what anyone thinks! I am a free thinker, and I know my opinions are correct because I instantly based this entire opinion on a subjective, anecdotal view of the world that I then extrapolated meaning out of, the best evidence! /s

Oh yeah, it’s definitely useful for that!

Since LLMs are essentially just very complicated probabilistic links between words, it seems to be extremely good at picking the exact word or phrase that even a thesaurus couldn’t get me.

I primarily end up using LLMs through DuckDuckGo’s private frontend alongside a search, so if my current search doesn’t yield the correct answer to my question (i.e. I ask for something but those keywords only ever turn up search results on a different, but similar topic) then I go to the LLM and ask a more refined question, that otherwise doesn’t produce any relevant results in a traditional keyword search.

I also use integrated LLMs to format and distill my offhand notes, (and reformat arbitrary text based on specific criteria repeatedly for structured notes,) learn programming syntax more at my own pace and in my own way, and just generally get answers on more well-known topics a lot faster than I would scrolling past 5 pages of SEO-“optimized” garbage just designed to fill time for the ads to load before actually giving me a good answer.

I have never once found an “AI” feature integrated by a corporation useful.

I have only ever found “AI” useful when it’s unobtrusive, and something I chose to use manually. Sometimes an LLM is useful to use, but I don’t need it shilled to me inside a search bar or in a support chat that won’t solve my problem until I bypass the LLM.

There are people who think that “positive” or “negative” words have a magic-like effect on natural processes.

From what I’ve seen, this was originally popularized in 2004 by Masaru Emoto’s book “The Hidden Messages in Water,” where part of his claims were that snowflakes would develop differently in containers labeled with negative or positive emotions.

Naturally, this turned out to be a complete lie, but many people, such as those in the original post, still believe that words can somehow influence things like mold development on food.

I love this approach, balances user freedom & privacy with moderation & voting pattern analysis by the public.

ActivityPub might mean some data is slightly less private, but that doesn’t mean it has to be.

I feel like this says more about these students’ schools, rather than the students themselves.

How have I never heard of this before!? I’ve been looking for exactly this for ages now.

Already spinning up a docker container!

You will eat here, and you will be happy about it! 😡

It’s because they’ve conditioned their audience to believe a few key things:

1. “Voting” with your wallet is what matters over all other forms of action.
2. Giving money to “woke” companies harms you as an individual.
3. Buying products is the best way to signal your value as a person.

Let’s break that down.

They want money to be more important than voting, because they understand that their political demographic does not win the popular vote in most cases, and their policies are inherently not popular with the majority. But when they can get money instead, then use that to influence votes and policy, well, that just might get them the policies they want without substantial votes from the public.

They want people to fear giving money to “woke” corporations because it makes them seem like the only source of real truth, and objectivity. They’re the voices of reason in a world that’s all against you. Then, you’ll be willing to pay for their subscription streaming service, and their subscription streaming service for kids, and their merch, and their chocolate, and their razors, et cetera et cetera.

They want you to associate buying products as the way to define yourself, because when you so strongly identify with their politics, you’ll spend as much money as you can signaling to those around you that you don’t support the “woke” agenda through your wallet, you only support those who truly embody your cause. Giving them money becomes a symbol of your values.

And of course, they wouldn’t get any sales without this as a selling point. If they just start a razor brand, not affiliated with their political ventures, who’s gonna buy? Their razors are effectively the same price as Gilette’s, just without the likely higher standard of quality and availability in physical retail locations.

But when they combine all three of those tactics I mentioned to make their target demographic believe they need these razors to display their values, stop a perceived evil agenda, and make their voice heard… well, then you’ve got a good revenue stream.