Forgot your password?
typodupeerror

Comment FlashAttention (Score 2) 43

I did some math the other day on running local AI models and the net result is most homes can't afford to run the current median models.

They don't just need 80GB of VRAM, they need newer architectures - to be supported by CUDA, to be supported by pytorch, etc.

These problems may well be solvable with more clever use of hardware, MoE, acceptable quantization, etc., but today you're in for several grand and something north of 100W idle to use what is effectively a $20/mo plan.

A small enterprise can afford local, so that's good. We paid more than that for one SGI machine back in the day.

The point of the exercise was to plot the position on the curve. We're at something like 2006 YouTube where nobody could afford the drives or bandwidth that YouTube/Google was giving away for free (aka with VC money). Eventually hard drives got cheaper, people got gigabit at home, FlashServer was replaced with h.264/HTML5, phones could stabilize video locally, etc.

So it looks like these AI companies need to stay alive for about seven more years giving away product at a loss, or at least highly oversubscribed, to turn a profit. Hence the low token allowance, the banning of OpenClaw, etc.

On the other hand, I read the blog of a security researcher yesterday who found an exploit with (IIRC) Claude, tried to refine the PoC, but got dinged on "out of tokens" before he could finalize it. So he just deleted the work and moved on.

It sounds like they're trying to not lose money at such a velocity and are trying to find a sweet spot where people don't just declare it too underpowered to use.

A global energy depression may well take out the supermajority of the companies that believe they can burn investment money for seven more years. There is circular financing money, then there is real return on capital money. One is to fool the markets, the other is grounded in current physics.

Comment Re:training may be legal (Score 3, Insightful) 75

One might imagine that buying a million books would get a buyer a very good discount from the publisher.

They could have paid $5 or whatever for each book they trained on. $15M or so for 3 million books - they could totally afford that but "why pay when you can steal?"

Then at least they would be defending fair use rather than defending a 'theft' lawsuit.

If we had Constitutional Copyright they at least would have millions of 14-year-old books to train on. That would be quite sufficient to train and refine models.

Comment Continuous Sanbag Domes (Score 2) 35

The continuous sandbag dome systems are actually structurally and seismically stable. Good in desert climates, at least. They use local materials to coat the surfaces with stucco.

It would be very amendable to automation.

See here but many other videos on the Tubes too:

https://www.ameripacific.com/c...

Comment Re:Should be easy to find the users (Score 3, Interesting) 134

Yeah, this article is too cute by half.

Per reports SpaceX has been arming Ukraine with terminals for several years so Russia has put a lot of engineering into detecting, characterizing, and targeting the signals. They've provided this technology to Iran.

Trump recently bragged about CIA providing automatic weapons to the "protesters" ahead of the "protests" (over Bessent's currency war) which Iran shut down using the SL detectors.

Allegedly large shipments of terminals by Mossad were interdicted and those agents were hanged.

These spooks are willing to "fight to the last Iranian". Glorifying this is complicity in their entrapment.

There are much better ways to freedomtech than broadcasting a beacon unless a rapid color revolution is the goal.

Slashdot Top Deals

If a thing's worth having, it's worth cheating for. -- W.C. Fields

Working...