Philpax

They never said the word "only" or implied it.

Well, they already did [0]. It's not very fun to use; something that was less miserable would probably see more use. Retrofitting them to a language isn't impossible, either; there are a few…

You can align to general, broad principles that apply to all of humanity. Yes, of course, there are going to be exceptions and disagreements over what that looks like, but I would wager the vast major…

Might be a good idea to put the link to the book before the spoiler warning ;)

This is typically done by alignment teams (whose role is to ensure the AI behaves in accordance with us) and Red Teams" (whose role it is to intentionally find holes in the systems). An infamous …

1. Hard to count. Started as a youth on game modding, but in terms of serious projects, probably 5-7 years. 2. Rust. It's a Rust shop, and the only C++ we have is in dependencies that we wrap wit…

Sounds good to me!

Uhhh, are you sure about that? She wrote a paper that praised Anthropic’s approach to safety, but as far as I’m aware she’s not invested in them. Are you thinking of the CEO of Quora whose product was…

Really? I found this to be a very insightful article; it made it much clearer for me why the board acted as they did. It doesn't have all of the details, but it's much more than I had yest…

Yep, he started just under two weeks ago! https://twitter.com/ObbeVermeij/status/1723424031808839801 …

You're lucky - it's the same man ;) https://insiderockstarnorth.blogspot.com/2023/11/the-trouble... …

https://huggingface.co/microsoft/Orca-2-13b https://huggingface.co/microsoft/Orca-2-7b …

This is a very comprehensive timeline of what's happened so far with sources and relevant commentary. I think it's certainly worthy of its own link - it should help clarify what's happe…

It looks like it's OpenAI²: https://twitter.com/satyanadella/status/1726516824597258569 …

> Do not forget that in some scenarios using unwrap is totally fine if a panic is acceptable. Looking through the first few pages, most of these panics are easy to audit, and are infallible or in c…

It's so incredibly not-difficult that Boston Dynamics themselves already did it https://www.youtube.com/watch?v=djzOBZUFzTw …

> I really can't imagine a situation where having type hints in Python is worse than simply not having them. Can’t they lie / go out of sync with what’s actually happening? IMO, an unenfo…

There's still plenty of data out there, including in other languages and undigitised books - and that's before you get to data in other modalities, like speech and videos. Synthetic data can…

That's a very strong opinion about a project that is at 0.3 and is attending to many simultaneous problems, including building their own UI solution. It's certainly an issue, especially for …

What makes you say that? There are constant improvements in how they’re being trained and what they’re being trained with; there really isn’t any particular reason to believe we’re at a maxima. Especi…

I’d say SD works quite well at the macro scale, comparatively; while the artifacts the GP mentioned still occur with SD and to a lesser extent Midjourney and DALL-E, they’re much less of a common occu…

> AI cannot sustain itself trained on AI work. This isn’t true. You can train LLMs entirely on synthetic data and get strong results. [0] > If new languages, engines etc pop up it cannot synthes…

Here’s an article by the Financial Times demonstrating its real-world impact today: https://archive.is/hlcL2 More capable systems will have a larger impact still.…

I agree. I use an Apple ANSI Magic Keyboard, MacBook Pro with Swedish layout, and a Keyboardio Atreus on a regular basis, and I can switch between them without trouble. There’s definitely a bit of an …

Click through to the megabezel website - there’s a few examples there. I agree that it’s pretty compelling!

What are you on about

Samsung, but BOE may be dual-supplying it: https://twitter.com/SadlyItsBradley/status/17227592388066431... …