Hmmmmmmmmmmm. This begs the question, what even is true level?

Yes I agree. It’s relieving to see a scientific result be the similar to what one would intuit.

Interesting training strategy. Makes a lot of sense intuitively. Worried this makes the model even more susceptible to prompt injections. Feels like this method adds more attack vectors? It’s unfortunate they didn’t attempt to test the long term hardness and stability, though it’s probably beyond their scope.

I recently realized it’s a non-issue. The people doing this have already been looking for decades to find new ways to rot their minds. LLMs are just the latest in a long line of tools that help them tune out.

🇪🇺the🇪🇺land🇪🇺of🇪🇺the🇪🇺free🇪🇺

Because the public school funding comes from public taxpayer money. This means the school does not get to choose how to spend it since the money belongs to the people. The people have voted and greenlit several pre-approved items for schools to spend money on, but anything outside of that needs to be approved by a vote.

Getting people to vote on this item is a Heraclean effort to say the least. Education budget often is the least immediately impactful thing on the ballot, if it makes it that far. Especially in states with strong traditional religious areas. For example, Puritans don’t believe that sports are something that kids should take seriously cuz it’s a game (literally something along the lines of: Games can be pleasurable, seeking pleasure is sin you should only seek God, therefore games are sinful). They don’t want their taxes going towards such sinful programs so they will always vote against it. This perspective is rooted in zealous obedience and is not something other people are willing to fight against.

TL;Dr It’s easier for schools to just get private funding themselves and sidestep the public budget restrictions, than it is to get a majority in the voting pool to approve the vote and implement new school budget item.

The Germans are getting to me. Die DEI threw me for a loop. I was so confused man.

Holy moly proper use of the regular past tense of sneak?? I haven’t seen that on the internet ever. “I sneaked my way into the concert, but I also snuck several beers into the venue.”

“Your honor, although the prosecution has indeed depicted my client as the pathetic soy virgin in exhibit A, meme 4, please watch this 7 episode TV drama mini series that the prosecution wrote and produced for this very case before making your judgement.”

It is ultimately the same format which does not yield constructive conversation. The conversation is as though OP were speaking on a stage making their statement/announcement and then the audience yells back to them. It functions for some things, most notably sharing popular news - not necessarily journalism, just a glimpse through the eyes of the zeitgeist; dank memes are a form of good news by this definition. It does not function for discussion oriented conversation, since most people don’t read the whole thread. The implementation of up and down votes manipulate conversational integrity in weird ways Though that will occur with any type of rating system, I believe the simplicity of the reddit style system yields conversational benefits that are less valuable than how cheap it is to implement. A cheap hack of a system will yield a cheap shoddy output. This leads me to believe a better method exists, we just haven’t found it yet.

The emphasis on user control and instance freedom is novel and appreciated, but it has come with a reckless disregard for the dangers of the nowadays well understood echo chamber effect of current social media. There are zero safeguards to prevent Lemmy from shattering itself under any amount of external stress or internal corruption. Organized attacks are a major threat, and there is a nonzero chance of that happening.

Again, all of these are just like long term weaknesses in the structure itself. It doesn’t mean it IS going to fail, it doesn’t mean things can’t be mitigated, I just don’t trust that it will stand for a very long time nor will it reach the significance of reddit. Could be wrong, but it looks like Lemmy is capable only of moving around the problems with Reddit instead of being able to actively quell them.

No and not really. Reddit is just past its expiration date. I believe Lemmy suffers from a lot of the same systemic issues and that will prevent it from becoming an actual long term solution. It’s just the alpha version. It’s better than the prototype, but it’s still built on the same fundamentals and will likely suffer a similar type of failure.

I can’t say as to why things are how they are. It’s likely just the usual open source lagging behind; These technologies are fairly new and the corporate stronghold around them is strong.

Shaders are programs that run on graphics cards to act as shortcuts for common graphic scenarios, they are not programs that dictate how graphic cards function. That could clear up some confusion.

Don’t run away unless it makes you stronger because of it. Once something has been gained, never yield it without a purpose or a strategy - even if that ends up being a bad decision it was at least yours to make.

It is a difficult endeavor to gain new ground; it is nigh impossible to fully recover that which has been lost.

After you have ceded all your land and yielded every advantage, where will you seek shelter and from what will you find strength?

Left of rei. Likely her twin sister Ram is to the left of her, but iirc they have different color hairbands so it’s probably just two copies of Rem, so like [Rem, Rem, Rei]

Based on their response habits, it’s likely a poorly made AI or a 13 year old kid. Not worth interacting with it cuz either: it is incable or caring, or they really aren’t supposed to be here and we really shouldn’t welcome children into adult spaces by allowing them into the conversation. If they don’t wanna have discussion, then why would they contribute to conversation when spoken to?

Since you believe this stems from your personal perceptions and observations, the answer likely involves a paradigm shift that random people on the internet can’t realistically assist with beyond basic advice. Whenever I’m stuck on a problem and I don’t know where to go, a good place to start is collect more data and widen your research. Good luck o7.

eye got chu fam

My favorite part, is that ALL of the current AI companion products use subscriptions to LLM services instead of having built in models.

Which means ifwhen the company stops paying their bill, or if the AI company goes under, your waifus goes kaput. Or more likely just restructuring because of a buyout. It’s already happened with a few companies -> big tech buys out the company then stops supporting the product line, that then customers get mad. IIRC, coincidentally HP is being sue for doing just that with some recent buyout right now lmao.

The ante seems to get upped each time, and there exists at least one unstable lonely maniac that will not consider legal consequences to their actions when they are suddenly rugpulled out of their unhealthy addictive relationship, so there’s still room for the bubble to burst in an even more epic crash.

Oh man I’ll try, but I can’t make any promises …

Modern particle physics breaks particles down into two groups: Dice that are weighted (bosons) and Dice that aren’t weighted but also aren’t fair (fermions).

Bosons always roll the same number, because they’re weighted.

Fermions always roll numbers, but we have no clue how many sides they have, or what numbers they can even roll because they change each time we roll them.

Classical Computers ignore this problem. They just count the number of dice they have, and are really really good at rolling precise amounts of dice and putting them into specific labelled jars. Their math works by carefully keeping these jars organized, and are limited by how quickly and accurately the CPU can organize amounts of dice.

It turns out if you roll a set of dice enough times, no matter what set of dice you use as long as they are random, you eventually wind up with a similar looking “standard distribution” of probabilities. Quantum computers let us zero out the dice to a fixed starting position, kind of like zeroing out a scale, and then we can use that to make calculations. This process is very sensitive and difficult and has a lot of scaling issues.

Enter Anti-Dice. Anti-Dice are the polar opposites of existing Dice. They are just like all the other particles but they have their numbers printed upside down, and their shapes are inverted.

A Majorana particle is a particle that takes this metaphor even Further BEYOND!!! It is a type of Fermion (dice that we can roll and will give us random numbers instead of the same number each time), but whenever we roll a Majorana particle it turns into its own Anti-Dice. This is a really cool concept that Microsoft is using here as a proof of concept to make a quantum computer that is easier to scale up, because now if we roll say a bunch of 6s and a bunch of -6s, we know it’s actually supposed to be the same number because of how Majorana particles are defined, and we can theoretically use this cheaper and easier method to scale up a quantum chip.

Yes, they could remain private but they are attempting to go public. This action is a legal speedbump that they must now work really hard to maneuver around.

Mistakes can still occur. Just because it is designed to be resilient does not mean that it’s infallible.