There’s usually no good reason to remove a panel for 60 years. A modern panel will still output over 60% of its original capacity. Oil wells which are not required a “cleanup deposit” generally run “economically dry” after 15 years. Cleanup liabilities are much more serious in O&G sector, and a big budget problem in Alberta.

While solar is highly recyclable, dumping them in a land fill is fine if they are not from First Solar.

While a critic might argue that’s just swapping one federal government for another, Lorusso argues that’s not the case in the US, where states have powers that Canadian provinces do not.

States don’t have the power to secede, so it’s a hotel california situation. Health care is not under control of any state. $100k in extra debt per person = $4000+/year in interest. $3000/year per capita military spending, about to increase to $5000/year. Higher interest rates and home building costs, including O&G drilling costs due to tariffs on Canada.

If negotiating secession with Canada, Crown land should stay with Canada or at least form a land bridge within Canada. Canadian policies would charge more for transporting Alberta exports, and reduce their energy use. Alberta secession economic optimism is based on going all in on dead ender energy without any real friends. Don’t expect keys to the store open arms invitation to being 51st state, either.

Training an LLM isn’t really even about compute speed, it’s about access to good training material.

Consumer can’t train from scratch but can fine tune/modify open weights model with their data. There is significant open source training material available, without needing/wanting Harry Potter knowledge.

I’m unsure the defense industry analogy is correct, although fewer hires does matter. Instead of hires, some people (often deeply unqualified) are now building software for spec to sell to industries that might be charmed/but unqualified to judge the software. Other’s making software for their own personal productivity, with self monetization ambitions, often copying what seems useful to others. It’s unclear whether there are fewer people working on software, or just bigger monolith companies shifting engineers to take on a more entrepreneurial role. That can mean trouble for the larger software companies.

You could (many have) make the same argument against higher level languages all these years. If no one actually knows assembler and metal networking (or GPU for that matter now) protocols, they can’t possibly make optimized/knowledgeable architecture decisions. The only people allowed to program should be compiler contributors.

If fwend, why pitbull shaped?

what does github copilot do? ELI20ish

This screws over the north east with less energy. Trump/GOP likes to cancel their projects then point that them for mismanagement. North east has to import energy, and boats from TX/LA count as imports. With Jones act shipping rates.

!sciencememes@lemmy.world is very high quality.

This is fine! I get paid to write code that passes tests. if Format F: and recreating test environment passes the most tests…

Huawei’s clusters have close to 4x the ram as NVIDIAs, and TFLOPs is most relevant to training. Huawei has better interconnect technology than NVIDIA, but incompatible with H200s, and so for China/friends use, it’s a much better package. Price/performance of 910 vs 5090 or 6000ada is much higher at single card level. The power cost/availability in China gives them much higher potential deployment rates. Chinese cloud rates tend to be lower than the same model on US clouds.

3.6 27b is probably most powerful/efficient (to size) model out there. Qwen has a history of leveraging deepseek power as well. (deepseek creating small models with Qwen as the base), and Alibaba is main hosting service for deepseek. Alibaba/Qwen in talks to invest in Deepseek, atm.

Huawei outperforms NVIDIA at the “cluster” level. Which are mostly turnkey systems for datacenter units. And promises truck container level cluster for next generation that is 30x the zetaflops as NVIDIA rubin cluster. China currently operates at 50% electric production capacity, and energy extremely abundant and low price, which make the per level card performance deficit irrelevant.

It is for inference, which is generally good at open backends. Most models done in pytorch, with backend library.

“other stuff” is the opposite of specific criticism. Extreme smears and lies exist in propaganda to diminish others, and just google all the accusations, is not a good way to be informed.

can’t address “other stuff”

It takes 34 states to declare constitutional convention. Once it starts, no rules. If one of the agenda items include redrawing state lines for more unity, can give more power to groups/states to join separate unions that advance their values.

At 32% approval, and blatant election rigging plans, a constitutional convention should occur instead of planning around impeachment for last 2 years of term.

Canadian or Provincial annexation is a terrible deal for 2 main reasons.

US debt, and power projection, is on a near term collapse trajectory. Insane healthcare, military spending levels for no social services is higher effective tax/oppression rate on people/small business. Changes in provincial (and/or federal) government structure would be extremely disruptive.

Territorial possession of the US is far more likely than statehood. Far less generous autonomy (Quebec especially) granted by Federal government, and autonomy eroded continuously like native treaties have been. Increased north/south trade seems nice, but “real states” will get federal manufacturing incentives, and possessions will be hollowed out to 3rd world level. Jones act regulations on sea shipping means high cost of living.

While tariffs decrease unity as they should, the whole “protect steel and auto workers for outrageously expensive cars in states that don’t have those industries” is reason for highest GDP states to secede from genuinely corrupt political process, that propagandizes swing states on national socialist domination of uneconomic operations delivering no results for them, and shit product for the oppressed states.

salinity changes, not temperature

temperature melt driven salinity changes.

how much heat AMOC carries from the equator to northern europe, and also try to see how much air it’s heating in the atmosphere.

the physics experiment should include heat in oceans trapped below surface. The different heat flow from 35C Florida oceans in October vs what used to be 28C even at slower flow rate. How a 2 delta melt rate one season could reduce the melt rate next season, how Greenland melt into 10C water instead of 2C water would increase the southward flow rate that increases/maintains the north flow. Salinity normalizes within 3 months. Early summer melt will be low at low winter freeze rates, by theory increasing Carribean north flow, late summer south flow will be warmer.

Observations since 2016 data paper have been the opposite warming of Europe in winter.

smoking withdrawal and not letting the expletives flow together with therapy mandated to deflect social responsibility for your problems increases the frequency of those flying off the handle. Vidya game complexity increases from pong could be a factor too. :P