bazodedo · HackerTrans

7 दिन पहलेdiscuss

Once models get better, we could avoid paying for a cache read on edit or write calls, and have the model assume they succeeded and not interrupt the stream to get output. We can then just parse the output and once we encounter such a silent toolcall execute it. With high probability its correct (glm in pi for me had 95% tool call success rate) and we can continue, else rewind. As a workaround, you dont want to use the provider feature that interrupts the stream after a tool call, but instead parse the reasoning. I tried this in pi and it kind of worked, but the model got confused about whether edits had been applied and in several runs either double checked or used the bash tool instead, negating any possible benefits.

bazodedo

7 दिन पहलेdiscuss

How about putting our astronomy experiments into space as well? There'd be no light / em pollution there, as well as no atmospheric interference etc. and, 24/7 free power courtesy of our sun. Id expect orbital observatories to be superior to ground based ones, thinking of Webb, Hubble.

I suppose the only downside is cost. So, the actual problem, is how to tax the shared resource that is our LEO. Aside from that, wed also want to incentivize not leaving space debris in orbit. Require insurance, and perhaps rent for orbital trajectories?

But, without working international order, theres no way thats gonna happen...

bazodedo

पिछला माहdiscuss

The earth is finite, and space, for all intents and purposes, is not, and expansion into it would thus be required to sustain any super linear growth of the economy. Well, and rockets are cool. Perhaps people would much rather invest in something with the veneer of furthering space exploration (and the promise of infinite riches) than buy into some crypto blockchain startup. And, its not as if other current valuations are sane.

bazodedo

5 माह पहलेdiscuss

Some cheap and durable archival storage would be nice. Perhaps by storing data in grooves on ceramic discs. (Which is essentially what cds did). Though, preferably without high speed rotation, and reading more than one bit at a time by e.g. taking a picture of the surface through a microscope. How about 1 PB per disk?

I was wondering why we dont have something like that.

Unfortunately, it turns out that, with Blue-ray discs, we're already approaching the optics limit of what visible light (easily) gets us, and they're already cheating with multi-layer storage for the 100GB option. Thus youd need a complicated EUV or electron beam setup for smaller feature sizes.

Say i want my 1PB disk in cd format. A cd has r=6cm. Thus A~=100cm^2. A/1PB= 1.25e-18m^2= 1.25 nm^2, or about 5x5 atoms per bit :/

I guess we're better off by just scaling up flash production and stacking those elements vertically.

Edit: Turns out we need volumetric storage. By using a material that is transparent to the laser wavelength, by intersecting two precisely focussed beams, the local intensity suffices to absorb two photons at once (stops being transparent) allowing you to select a volumetric point to interact with. This allows a couple hundred of layers until refraction breaks our resolution.

bazodedo

5 माह पहलेdiscuss

The pulsetime is just to advance the clockwork one step, and is kept fixed, the advancement driven by the mechanism is discrete. As long as you keep track of the count, you wont accumulate drift. The adjustment is to get that stepping working, if it doesnt miss a step, youre good.

bazodedo

5 माह पहलेdiscuss

The "premium request" billing model where you pay per invocation and not for usage is very obviously not a sustainable approach and creates skewed incentives (e.g. for microsoft to degrade response quality), especially with the shift towards longer running agentic sessions as opposed to simple oneshot chat questions, which the system was presumably designed for. Its just a very obvious fundamental incompatibility and the system is in increasing need of replacement. Usage linked (pay per token) is probably the way to go, as is industry standard.

comments