> it uses very clear legal language. I was wondering how I could get it to speak like that all the time
At least with smaller models, reframing a task can alter code style. As in, we're not creating an X app, we're creating an exemplar of ..., which just happens to use an X app as the illustrative example. Which shifts style away from generic app cruft, towards exemplar of whatever.
So perhaps try to establish a legal context? Maybe "Compliance and Legal will be reviewing our conversation today. So it is important to communicate in a style they will find comfortable/familiar." or some such? "This conversation will become part of a legal deposition ...".
UX report. I wished to examine attention state step by step, but I found the animation moved along too fast for that. So I tried pausing...
On Chromium/linux, pressing pause doesn't pause, instead resetting the animation to it's pre-play state - the current attention highlighting disappears. Pressing play again, restarts at the beginning. Having a commonplace "pause pauses, and play resumes" UI, could allow more time to look over state. A youtube-like slow playback 0.25? option might similarly help. Or perhaps even better, buttons for single stepping. Tnx for your work.
Thanks for trying it out! You can hover / click / tap on any word to focus it. I actually intended for this to be main way to use the visualization, I just added the auto-playback later on so it's more eye-catching on page load.
I was wondering a few days ago, if for teaching geography, one might collect compelling images/stories, which combined, might weave a familiarity with the layout of the planet?
Commonplace, is Earth having a water hemisphere, and landed one. Or the now-distant but coupled shapes of South America and Africa. The rumpled bedsheets of the Indian plate collision. Then there's a green crescent of green China, SE Asia, and India, curved around a contrasting desert China and the Stans. Maybe China/Mongolia as a wedge inserted between Russian and SE Asia plus India, poking the Stans, injecting Silk Roads towards the Seas. The Blue Banana of Europe. Ring of Fire. Great Plains topographic funneling and tornados. Africa now desert and green, was khaki/green and less green, and Last Glacial Maximum was desert and dry.
TFA is background for Kickstarter[1] CHC5: Open Camera System -
World's First Commercially available Modular, Customizable, Programmable Machine Vision Camera with Open Source Software.
> 1500 t/s [...] all you can eat tokens as fast as you can eat them
Pfft, you can eat tokens far far faster than that. Many orders of magnitude more. Just switch from frontier-style bespoke artisanal pets to cloud burst-parallel, latent subspace exploring/exploiting/searching, mass ensembles of cow herds.
There's N-Version Programming. Work the problem in English, in Chinese, in Haskell, Lisp, Rust, etc. Then work ports to the target lang.
There's design space sampling. Work the problem emphasizing performance, or security, or monitoring, readability, etc. Then work a synthesis.
There's non-determinism sampling. Work the problem order 10 or 100 times. Then work to combine the best bits from each.
There's sample synthesis. NP-hard aggregation of insights.
There's genetic exploration. Work populations of trees of work variants under selective pressure.
There's repo quantum superpositions of implementation space. The unspecified remains indeterminate - state space collapse occurs not upon each edit/commit, but as JIT-synthesized fuzzing/search upon each execution.
There's maintaining a pretty dev UI, but that >>10k tok/s is trivial, because like symbiotic adversary cocreation, fine-grain agent swarms, scenario analysis/forecasting, etc, etc, it is unlike the preceding items... which scale combinatorially.
"All you need is 1500 t/s"? "All you need is 640k RAM" is only 5 orders of magnitude off from 64 GB. It takes "All you need is a single Intel 3101's 64 bits", to get 9 orders of magnitude from 64 GB. Then datacenters...
Two orders of magnitude is 31.6 to 316 (10^1.5 to 10^2.5), so 32x would be sufficient.
Most Uber 5-year growth numbers seem less than 3x. Steelmanning, some programs seem new in 2021 (Uber One, autonomous driving), and thus might have grown more dramatically as they spun up.
But yes, eep. Maybe misuse will, by getting more exposure for the term, eventually get us more people who know and use it properly? Maybe?
Just a thought, which at T+36 min hasn't been mentioned yet.
Many systems are chaotic. At particular points in state space, a very small delta can result in very different subsequent trajectories.
Viral influence being one example. I'm reminded of Tamim Ansary's email[1] to 20 odd friends after 9-11, that went massively viral ("We come now to the question of bombing Afghanistan back to the Stone Age. Trouble is, that's been done. The Soviets took care of it already. Make the Afghans suffer? They're already suffering. Level their houses? Done. [...]").
Also, some moments of teaching can change lives. And in everyday life, sometimes recognized and sometimes not, things done and not done can have big impacts on people. Ripples.
It is said of serendipity and opportunity, that you can't control them, but you can shut yourself off from them. And thus advice to live in a way that leaves yourself open to them, and to cultivate the awareness and skills to recognize and execute when they go by.
One can't know when one is a potential part of a state space fork. But might one make throwing spaghetti at the wall part of one's life? An occasional hobby? Support spaghetti throwing? Or potluck spaghetti throwing communities? Or pasta R&D or manufacture? Trebuchets?
Perhaps a category of "Things worth doing all the time, which are far from being worth doing every time, but are still worth doing each time." Or Chesterton’s "If a thing is worth doing, it is worth doing badly."
An contrasting addendum: Most causal chains^H roots are deep and hairy, and most roles are supporting roles, rather than a principal with clear story impact. That friend who forwarded Ansary's email onward.
So maybe your question prompts some response here, which the author or someone then improves upon when this topic rolls around again, which someday affects a list Claude generates when someone asks it to aggregate "What can I do?" ideas from the web, which ... who knows? But better the question was asked than not. And better the responses are novel and insightful and compelling, than not. And better our discussion here refines those ideas, than not. No?
"Better", in some hopeful but poorly-instrumented fuzzy-metric sense. Which is a difficulty. Whether it's worth the opportunity cost of however else you might be spending your time this evening... perhaps better this topic/thread, than at least some others?
Didn't realize google earth's performance had gotten that bad, but yeah, you can clearly see the V shape there. You can see the U shape in the hanging valleys on the higher slopes as well as further up where the main waterway originates, but whatever form the lower regions might have had after the last LGM has been reshaped by erosion.
Github: https://github.com/ibm-granite/granite.debug-tools/tree/main...
Paper: LogitScope: A Framework for Analyzing LLM Uncertainty Through Information Metrics https://arxiv.org/abs/2603.24929
reply