Interpreting language models can feel like stumbling through a dark forest - sometimes you just wish you had a flashlight! In our new post, we introduce HeadVis, our latest flashlight for studying attention heads.
New Anthropic research: Tracing the thoughts of a large language model.
We built a "microscope" to inspect what happens inside AI models and use it to understand Claude’s (often complex and surprising) internal mechanisms.
New Anthropic research paper: Scaling Monosemanticity.
The first ever detailed look inside a leading large language model.
Read the blog post here: https://t.co/6RYwxt6nWI
@CubeSmiling@RichardMCNgo @balfour_oliver Agreed. I think it’s different for CME which doesn’t really have competitors. Seems like most countries only have one dominant futures exchange
@CubeSmiling@RichardMCNgo @balfour_oliver One idea: Exchange could add 250ms delay to packets leaving their colo and also send their market data to other exchanges without delay. Then sell their real time market data and remove incentive for wireless towers (and still charge huge fees for the data!)
@RichardMCNgo Exchanges have strong network effects and end up being stagnant monopolies. Total hft profits are relatively small so the savings from this isn’t huge. For us stock market there are many exchanges but my understanding is regulation doesn’t let them try the solutions to this
@EpistemicHope@shakaz18 I tried exactly what you described with two different economics tutors from wyzant and was disappointed with the results. Seems like tutors there are optimized for passing various econ classes and not for a gears level understanding
The pace of the book completely changes when fossil fuels show up. It felt like I heard about every advance in farming, hydropower, and windmills then fossil fuels come along and suddenly things change so fast we can't keep up.