Zos_Kia

Technical testing still leaves a lot of potential issues with business rules, UX etc…

If I had to guess, I’d say it’s not necessarily baked into the models, but rather part of a style guide in the system prompt

Good luck finding enough wood for that ! Energy was reaaaally expensive back then.

They had sewage and toilets since Roman times. It wasn’t affordable to many (and you couldn’t make it affordable) but they definitely knew how to make it.

I’d just like to interject that while traveling was rare in medieval times, it did happen. People usually didn’t get thrown in jail for it, even if they didn’t speak the local language.

Regular people didn’t really speak Latin beyond a few bits of prayer. The lingua franca was a mix of various coastal languages (think of the belter patois in the expanse), but even that was only known to traders.

You’d have a tough time for sure, but wouldn’t necessarily get in trouble.

They do math, just in a very weird (and obviously not super reliable) way. There is a recent paper by anthropic that explains it, I can track it down if you’d be interested.

Broadly speaking, the weights in a model will form sorts of “circuits” which can perform certain tasks. On something hard like factoring numbers the performance is probably abysmal but I’d guess the model is still trying to approximate the task somehow.

Minitel was a text only early internet that popped up in France in the 80s. You connected to it through a small terminal connected to the phone line, and had access to various commercial services such as phone book, train booking etc…

Most of those services have been shut down a decade or two ago but some hobbyists are operating new services on the network.

I think the real shocker was the step change between 3 and 4, and the hope that another step change was soon to come. It’s pretty telling that the latest batch of models was fine tuned for vibes and “empathy” rather than raw performance. They’re not getting the next a-ha moment and want to focus their customers on unquantifiables.

It seems logical that this would negatively impact performance and, well, looks like it did.

Nah the 500B$ is for building data centers. Well, it would be if that money existed but the truth is it was just an empty announcement.

The companies involved don’t have that kind of money, even pooled together.

I don’t know any AI artists (as in someone who prompts a model and then calls the result a work of art), although most traditional artists i know have come to incorporate AI one way or another in their process.

You don’t really hear about it because it’s all intermediate material used during the production phase. For example, as a hobbyist writer, one thing i struggle with is writing action scenes cause i don’t have visual memory and i tend to forget a lot about continuity and “spatial realism” (“this guy starts in this corner of the room so there’s no way he could grab that object at that point”, shit like that). With AI I can generate some kind of “story board” of my scene, which helps me write it much better. It’s just laid out visually in front of me and i catch a lot more details.

Sometimes when i’m toying with an idea i’ll also have a model generate a few variations on it, with different points of view, writing style, focus etc… Even if the writing is mediocre, it gives me a really good idea of how each version could pan out, and whether an angle works or not. I’ll then select the angle that works best and rewrite it entirely from scratch.

There’s nothing innovative about it, people have been using assistants to avoid tedious work forever. It’s just that before AI you had to, you know, be rich and able to actually pay for the labor.

Yeah he should be using real art like stock photos and shitty clip art

When you read that stuff on reddit there’s a parameter you need to keep in mind : these people are not really discussing Lemmy. They’re rationalizing and justifying why they are not on Lemmy. Totally different conversation.

Nobody wants to come out and say “I know mainstream platforms are shit and destroying the fabric of reality but I can’t bring myself to be on a platform except it is the Hip Place to Be”. So they’ll invent stuff that paints them in a good light.

You’ll still see people claiming that Mastodon is unusable because you have to select an instance - even though you don’t have to, you can just type Mastodon on Google, click the first link, and create an account in 2 clicks. It’s been ages. But the people still using Twitter need the excuse because otherwise what does it make them?

If you take into account the optimizations described in the paper, then the cost they announce is in line with the rest of the world’s research into sparse models.

Of course, the training cost is not the whole picture, which the DS paper readily acknowledges. Before arriving at 1 successful model you have to train and throw away n unsuccessful attempts. Of course that’s also true of any other LLM provider, the training cost is used to compare technical trade-offs that alter training efficiency, not business models.

just like for american models, the political bias is irrelevant. Realistically you are using the model for its reasoning capabilities, not its answer to “what happened in Tiananmen”.

Yeah it’s ridiculous. GPT-4 serves billions of tokens every day so if you take that into account the cost per token is very very low.

Care to source that statement? What’s the global consumption for AI compared to production by renewables?

deleted by creator

They have no ability to actually reason

I’m curious about this kind of statement. “Reasoning” is not a clearly defined scientific term, in that it has a myriad different meanings depending on context.

For example, there has been science showing that LLMs cannot use “formal reasoning”, which is a branch of mathematics dedicated to proving theorems. However, the majority of humans can’t use formal reasoning. This would make humans “unable to actually reason” and therefore not Generally Intelligent.

At the other end of the spectrum, if you take a more casual definition of reasoning, for example Aristotle’s discursive reasoning, then that’s an ability LLMs definitely have. They can produce sequential movements of thought, where one proposition leads logically to another, such as answering the classic : “if humans are mortal, and Socrates is a human, is Socrates mortal ?”. They demonstrate the ability to do it beyond their training data, meaning they do encode in their weights a “world model” which they use to solve new problems absent from their training data.

Whether or not this is categorically the same as human reasoning is immaterial in this discussion. The distinct quality of human thought is a metaphysical concept which cannot be proved or disproved using the scientific method.

Last year it was APIs

Hahaha the inane shit you can read on this website

They wanted you to go back to work, not in X months if and when a vaccine was available, but tomorrow, taking some bullshit drug if you needed reassurance. While they were happy working from home, triple vaxxed and double masked.

I don’t know what the intent was, only what it achieved. It is now political suicide to enforce public health mandates, so when the next pandemic hits, your boss will be able to force you to go to work in unsafe sanitary conditions and a lot of people will be cheering for it.