lowki jammin

beats, jams, and contracts. Internet money bitch

[lowki jammin @0xlowki]

Has anyone else had literally 0 emotional reaction to market tank

[lowki jammin @0xlowki]

@random_walker .....?

[lowki jammin @0xlowki]

@abacaj if nothing else we'll just continue to amass quality datasets, which will hold value as learning algos continue to improve

[lowki jammin @0xlowki]

@rezkhere the only people who see bard as competition are still using gpt3.5, or don't actually use AI at all

[lowki jammin @0xlowki]

@tszzl unfortunately we'd probably just think the models are hallucinating

[lowki jammin @0xlowki]

@simonw Anyone who puts Bard close to GPT4 in performance is out of their mind

[lowki jammin @0xlowki]

@_Dave__White_ @OpenAI ex: the best compression we have for 3D models is overtraining a model on one data sample, and the weights serve as your compression vector

[lowki jammin @0xlowki]

@_Dave__White_ @OpenAI embedding vectors are a compression of sorts, dependent on model architecture. There are plenty of examples though where the state of the art on compression is not through an embedding

[lowki jammin @0xlowki]

@iamgingertrash also if you match f(x) cheaper than the original creation, you could potentially RLHF to f(x) + 1 with a lower total cost (seems like shaky ground but I wouldn't outright dismiss it)

[lowki jammin @0xlowki]

@iamgingertrash the goal isn't for f(x) + 1 but simply to match f(x) though, no?

[lowki jammin @0xlowki]

@krishnanrohit the point is that solving those error manifolds speaks roughly nothing to interpretability / explainability. Regardless on your views of Yudkowsky, this point stands. And there is no clear path as of yet towards alignment (if anything, deeper models are harder to constrain)

[lowki jammin @0xlowki]

@lsukernik so like many things, depends on the reasoning. Quality of team and execution universally applies tho

[lowki jammin @0xlowki]

@lsukernik will add the caveat though that outsiders often have naive perspectives on companies, industries etc. You can even be in the *same industry* and have a very poor understanding of what customer needs and company solutions *actually* look like

[lowki jammin @0xlowki]

these GPT rate limits are gona be the boomer things we reminisce about

[go follow my other account @my_old_acount]

Just made a handy tool to use #GPT4 on macOS!

Select text, right click (or keyboard shortcut), & a pop-up window appears for quick commands.🪄✨

Summarize, translate, reformat, debug -- without switching apps.

Meet Lookie⬇️

[lowki jammin @0xlowki]

@DramaAlert when capitalism victimizes you with free money

[lowki jammin @0xlowki]

@lsukernik need to define fully onboarded, i often feel like a lot of "crypto native" devs still have a good amt to go

[lowki jammin @0xlowki]

@SubsetTopology we have no idea what signatures are real and which aren't lmao, it's a meaningless list alone

[lowki jammin @0xlowki]

with LLMs, sales teams everywhere have finally cracked engineer-to-customer translation

[lowki jammin @0xlowki]

I expect to see the US military increasingly leveraged as dollar reserve status is questioned more and more