danielbln @ danielbln @lemmy.world

Posts

9

Comments

171

Joined

2 yr. ago

Moderating

---

You sound like a sour lemon. I'd rather hang out with a hippie that thinks the universe is conscious than someone who slaps a paper tiger license to their low originality threshold posts. Just sayin.

Microsoft's Phi model was largely trained on synthetic data derived from GPT-4.

If their soft deletes (so instead of actually deleting, it's just a flag on the comment that hides it) then no, it won't make a difference at all.

Windows is the one where I need an account to install and that spies on me and throws ads in my face, that one?

Not nearly as bad as hearing "click...click....click" from the HDD. The PTSD is still strong...

Are you asking why stock of a single company is different from "stock" of the richest country and only superpower on earth?

Also, money is liquid, can be spent immediately. Stock is not liquid, it has to be traded, vested, etc. and given enough stock will tank yje value if too much of it is liquified at once.

I'm German you "hit" a decision.

Länder:innen

Clearly, because chairs are obviously male (German). Anything else is just silly.

No Elephants in South America, where the Avocado evolved.

https://www.smithsonianmag.com/arts-culture/why-the-avocado-should-have-gone-the-way-of-the-dodo-4976527/

Is evolutionary anachronism really a disputed theory when it comes to avocados?

Fun fact: some fruit evolved to be eaten by mega fauna, giant animals, that no longer exist because uh we ate them. One of those is Avocado. A pit that is way too big to be eaten by fauna of current day size.

They run all of Gamepass as well as all of Sony's PS+ on Azure, I think they'll be fine.

https://app.suno.ai/song/6f459049-6bfc-4d7d-929f-59db42766105

Eh, that's not quite true. There is a general alignment tax, meaning aligning the LLM during RLHF lobotomizes it some, but we're talking about usecase specific bots, e.g. for customer support for specific properties/brands/websites. In those cases, locking them down to specific conversations and topics still gives them a lot of leeway, and their understanding of what the user wants and the ways it can respond are still very good.

Depends on the model/provider. If you're running this in Azure you can use their content filtering which includes jailbreak and prompt exfiltration protection. Otherwise you can strap some heuristics in front or utilize a smaller specialized model that looks at the incoming prompts.

With stronger models like GPT4 that will adhere to every instruction of the system prompt you can harden it pretty well with instructions alone, GPT3.5 not so much.

I've implemented a few of these and that's about the most lazy implementation possible. That system prompt must be 4 words and a crayon drawing. No jailbreak protection, no conversation alignment, no blocking of conversation atypical requests? Amateur hour, but I bet someone got paid.

MRGA!!

Or maybe an eggy Mona Lisa? https://i.imgur.com/WQHOf1r.jpg

This is made with a Stable Diffusion LoRA called FriedEgg. Here, I used it to generate a wider version: