jbsimpson

It's a fundamental limitations of LLMS - don't use them if you're worried about hallucinations.

Anyone using AI tools should assume that at least 20% of what it produces will be wrong: Missing details, inventing things, basic mistakes.

Use AI if being 80% right quickly is fine. Otherwise if you have to do the analysis anyway because accuracy is critical, there's little point to the AI - its too unreliable.

If we had the renewables to replace the coal politicians would love it to retire in a heartbeat. The reason it’s sticking it around longer is because politicians fear the backlash from blackouts and high prices more than the backlash from the bad PR of delaying closures of coal.

When Tassie dams get low, they import power from the mainland. It’s a common seasonal thing. Drought + Basslink outage had the gov buying and running diesel gen at huge cost to keep the grid running. The Basslink outage made a problem a crisis.

are you sure it wasn't prostitution but child abuse overseas that was illegal?

You also have to factor in how terrible post-doc pay and job security is, just to have a shot at getting on the tenure track. In the US it's half of that 80k number in bio research.

The effect is so significant that South Australia actually experiences intervals of negative operational demand.

> From the institution, not the authors.

Typically the authors pay the journal out of their own funding.

40% of Australia's population lives in just 2 cities.

Depending on the season Tasmania exports a lot of electricity to the mainland.

South Australia is a fascinating case study of this in action - average prices in the middle of the day are actually negative.

This has been clear for a while. The real challenge for grid operators at the moment is system security - we need sufficient storage to ensure that supply and demand are consistently balanced.