Also, given different AGI platforms had different graphics capabilities, I expect even the original Sierra AGI games had somewhat different graphics assets on different platforms – so whatever preprocessing of graphics they are doing to make them C64-friendly, Sierra quite possibly would have done something like that anyway had they made a C64 port back in the day.

Australia used to have a "White Australia Policy". It was introduced in 1901, but began to be progressively relaxed after WW2 before finally being completely abolished in 1973. The policy involved a general ban on non-European immigrants - and even at times Eastern and Southern Europeans too – although it was never exceptionless (e.g. Maori immigration was always allowed, to avoid upsetting New Zealand; originally only Ashkenazi Jews were allowed to immigrate, but in the 1950s they started allowing Middle Eastern Jews as well.)

So, in that respect, Australia does have certain parallels to US history on race; and the dismantling of the White Australia Policy roughly overlaps the US Civil Rights Era, and Australian developments were likely partially influenced by that concurrent struggle in the US.

However, the way post-1973 Australian officialdom responded to the abolition of the White Australian Policy was very different from subsequent developments in the US. The Australian official philosophy was to stop classifying people on the basis of race – the Australian census asks about ancestry, country of birth (for both self and parents), ethnic identity, indigenous status, language, nationality, religion–but never race; all these concepts partially overlap with notions of "race", but none of them can be straightforwardly identified with it. In fact, the Australian Standard Classification of Cultural and Ethnic Groups (ASCCEG), groups ethnicities together based on continent of origin, not "race" – so African-Americans and White Americans are grouped together as "peoples of the Americas", while Egyptians and Kenyans and Afrikaners are all "peoples of Africa".

Now, this decision didn't magically solve all issues with inequality and discrimination in Australia – but did it cause harm? My own honest impression is that, overall, its benefits have likely outweighed the downsides. And, due to the influence of US media, there has been a growing subset of Australians – primarily on the left – who have adopted more US-influenced ways of thinking about and speaking about race and ethnicity – and I'm not convinced that has been beneficial.

Of course, whatever Trump and the GOP are doing isn't necessarily the same thing as what Australian politicians and bureaucrats – on both sides of the political aisle – have done. There are significant differences in context and history. But they aren't completely different either. My own impression as an outside observer of the US, is a lot of narratives which Americans tell themselves (on both sides) make the most sense when you avoid looking at other countries as comparators.

Postscript: inevitably some will respond by bringing Indigneous matters into the discussion–but indigeneity and race are fundamentally different things. To quote Australian Indigenous activist Noel Pearson, "If we were in Lapland, we would be blond and blue-eyed and we’d still be Indigenous.” Consider someone like Dr Gordon Reid MP, the federal Member for Robertson – he publicly identifies as a "proud Wiradjuri man", but physically he appears indistinguishable from Anglo-Celtic Australians.

For a lot of use cases, you don't need a general purpose search engine – a search over a curated knowledge base works even better.

There are plenty of freely available data sets you can use, depending on the application; plus in many cases you will want to use internal-only knowledge bases containing non-public information (e.g. documentation for a corporation's internal systems and procedures). There are also many paid subscription domain-specific knowledge services available.

No, it literally isn't. The reverse cycle air con just blows hot air into the room–it is purely convective. The gas heater console provides radiant heating as well.

Also, the air feels different. Reverse cycle air con tends to deliver a higher volume of air at a lower temperature; gas heaters (especially if we are talking about console units as opposed to ducted systems) tend to deliver less air, but hotter air. It feels different on the skin.

Yes you can get bottled LPG gas, but it is a pain – you worry about running out, you have to call up and get them to deliver it. With piped gas you just set up a direct debit for the bill and never think about it ever again.

Also, some people prefer gas heaters because it is a different kind of heat. My father replaced his wood combustion heater with a gas one, even though he already had reverse cycle air-conditioning. Maybe just using the air conditioner would have been cheaper (he already has rooftop solar–I don't think he has a battery, but could easily put one in), but it wasn't what he wanted.

The main Australian electricity grid reaches from Cooktown in far northern QLD [0] through to the Lower Eyre peninsula in South Australia [1] and down to the Huon Valley in Southern Tasmania [2]. So if by "coast-to-coast" you mean "East Coast" to "South Coast", then literally we have it already.

Of course, I take it you actually mean including WA and NT into the main national grid. No, we don't have that, and probably won't any time soon.

But there is a project (Australia-Asia Power Link aka SunCable) [3] to run a HVDC cable from Central Australia to Singapore via Indonesia, to export solar power. Why not the same thing between Perth and Adelaide? It simply comes down to economics – Singapore and Indonesia have the energy demand to economically justify it, domestic Australian consumption doesn't.

However, there is a plan to extend the national grid to Mt Isa. [4] There's only around 600 km from Mt Isa to Tennant Creek (where the AAPL/SunCable project starts). It isn't impossible that they'll eventually link up. So we might actually see the northern NT connected to the national grid, eventually.

[0] https://www.aemo.com.au/-/media/files/electricity/nem/networ...

[1] https://www.aemo.com.au/-/media/files/electricity/nem/networ...

[2] https://www.aemo.com.au/-/media/files/electricity/nem/networ...

[3] https://en.wikipedia.org/wiki/Australia-Asia_Power_Link

[4] https://en.wikipedia.org/wiki/CopperString

More than 80% of coal mined in Australia is exported. So reduction of coal consumption internal to Australia has only a limited impact on Australia's coal industry.

Long-term, adoption of green power in emerging markets and developing countries is a huge threat to Australia's coal industry – but there is very little Australian industry lobbyists can do about that. And whether Australia goes down that route has only limited influence on whether the coal importing countries will.

There's a lot of people who would get gas if only it were available. My house (in NSW), the gas main ends two streets away, they won't connect me unless I paid them some insane amount of money ($100K or whatever) to extend it to us. My father (different part of NSW), his neighbour had gas, but the gas main stopped outside his neighbours house, and it would have cost him thousands to get it extended so he could connect; then his neighbour on the other side paid those thousands, and they extended the gas main right past his front door, and then he could get connected for $500. Anyway, I think the measured demand isn't always the same as the real demand.

Also, there's still a decent chance the Victorian government loses the state election this November–although with Ben Carroll replacing Jacinta Allan, it is less of a sure thing than it was–and the Liberals have already said they will wind back the new gas connection ban.

But if people are going to post it on HN, they should consider at least writing the README / main page of their website by hand

It is nice to hear the human authorial voice, at least sometimes

In the US, you normally don't recover costs even if you win; sometimes you can get an exception, especially if your opponent behaved in a particularly egregious manner.

In most of the rest of the English-speaking world, the normal rule is the loser pays a significant percentage of the winner's costs. I know here in Australia, the default is the loser pays (on average) 50–75% of the winner's costs ("standard costs")–the variation is because the judge sends the case to an auditor who reviews the winner's legal bill against the rule book, and so the exact percentage depends on factors like which tasks the winner's lawyer charged them for, their hourly rate, what the auditor viewed as reasonable given the complexity of the case, etc. Convince the judge the loser behaved egregiously, and then you get indemnity costs, which are closer to 85-100%.

This has pros and cons. For the "little guy", it makes lawsuits much more survivable if you are confident you are going to win; it makes you much more inclined to just fold if you think there is a significant risk you'll lose.

It all depends on what the use case is.

For example, consider a system which takes English questions from business users, translates them to SQL, then runs it (as the business user) against a reporting database

How does the attacker get access to it? And even if they somehow do, what harm can they actually do to it? The only tool it has is one to run SELECT against a DB, and it can only SELECT data the user already has access to, and we have timeouts/etc to protect against overly expensive queries

If your design involves lots of general purpose do-anything agents with very generic tools, the risk profile is very different from narrowly scoped purpose-specific AI services

Of course, there is a scaling problem here - doing it for N languages is N times the work of doing it for one, and you might not know some language well enough to do a good job of it

Two things you can do:

(1) Use an LLM for translate your regexes, classifier training/validation datasets, etc, from English to N target languages

(2) Use an LLM to translate the input to English, then run your validations against that English

Of course, these approaches aren’t foolproof - the translation might be flawed/etc - but a defence doesn’t have to be foolproof to add real value

Also, of course using an LLM to translate hostile input to English before checking it runs the risk the translation LLM itself might be jailbroken. But how about using a classical statistical machine translation model instead? They don’t tend to be vulnerable to these kinds of attacks, because they aren’t trained to follow instructions

I think this is binary categorical thinking. In the real world, safety systems (even in domains like aviation or nuclear power) are never foolproof-the point is you reduce the probability of failure to an acceptable level given the costs of doing so and the potential consequences of that failure

And there is the risk people say “there is no foolproof solution, so I’m not going to invest in probabilistic countermeasures” - which would sound like utter madness to a bank’s antifraud department, but for some reason a lot of people seem to think it isn’t when it comes to AI

Now, of course, there is a long tail of elaborate variations that those techniques won’t be able to stop. But have the published vulnerabilities come from that long tail or from not doing enough to address the simpler cases?

For several of them, it is possibly the only surviving publicly accessible code.

For example, the GUTS (Gothenburg University Timesharing System) implementation-I doubt you will find any other code written for GUTS anywhere online. Maybe some offline archive at some former user still has some code for it, maybe this is all that survives

My wife and son then told me I "broke" the front door. I was confused, I thought I had fixed it

Turns out, with the door coming off the hinges, it wasn't closing properly, which meant it was ajar most of the time, so our dog could push it open with her body.

Now I'd fixed it, it closed properly all the time, so our dog started whining for someone to open the front screen door for her

So I loosened the screws, and hence "fixed" it by breaking it again

The big reason for picking the 68K version is it was mainly written in C (initially in Pascal, but they later switched to C.) By contrast, the more famous 8080 versions were largely written in assembly, as was the 8086 version.

Well, CP/M-80 1.x was mostly written in PL/M, a dialect of PL/I developed by Gary Kildall, but CP/M-80 2.x+ was rewritten in assembly for speed; PL/M was still used for some of the utilities. Obviously, C compilers are much easier to come by nowadays than PL/M compilers are.

It is false that it doesn't have the concept of processes and threads, or the distinction between them

What I was saying "isn't really true" is the second claim, not the first claim

- Create a skill to perform some common task

- Watch the agent use it

- Wait until the agent makes a mistake

- Ask it to update the skill to warn it against making that mistake again

I think that, using that iterative process, I end up with something better than just asking it to perform the task plainly

One observation: ask an LLM to write a skill, it tends to make them overly verbose and prescriptive. Often, something briefer and human-written actually works better

Another: I generally let the LLM propose edits to the skill, but I review them carefully and often modify them, because I find it has a tendency to solve the current problem at the price of worsening the solution to a previous one