Fact-Checking CSIRO’s Communications about the Trustworthy Systems Group
microkerneldude.org
microkerneldude.org
Management doesn't seem to appreciate what a good thing was going on here. Sooner or later, security research will pay off in a big way. It is a pity we can't measure catastrophic hacks avoided.
To be fair I'm not sure what options D61 management actually have - they could focus on supporting solid programmes like TS but it'll only last a year or two before the government yanks their funding, because if D61 has nothing to put in the background of a ministerial announcement then its of no value to the government at all.
Their observations suggest the latter - a lot of very smart people, but also some serious conflicts of interest when it comes to funding allocation.
Apparently this was always an issue, but accelerated in the mid-2010s after their funding was cut by the Abbott government (this is not intended as a partisan statement; for what it's worth the person who told me this was quite politically apathetic).
> Australia is a lucky country run mainly by second rate people who share its luck. It lives on other people's ideas, and, although its ordinary people are adaptable, most of its leaders (in all fields) so lack curiosity about the events that surround them that they are often taken by surprise
The 2nd item is "Voted against a binding code of conduct designed to ensure politicians act with integrity" - that isn't corruption. The list author just doesn't like their decision. Honestly, none of the top 6 dot points really look concerning and it is arguable if they represent corruption. Some are probably politically motivated ... but most things politicians do are politically motivated.
Whether they’re the right calls or not…let’s just say they didn’t hang around long enough to allow anyone to make a definitive judgement.
But this is a terrible blog piece. It's trying to parse English language statements into verifiable statements - when natural language is all about context.
The key thing with Data61's strategy is "Data61 will be all about AI".
This post tries to argue that sel.4 could support "cybersecurity for AI systems" but the points that are made could equally be applied to any cybersecurity research.
Trying to find some dishonesty in Data61's statements is pointless. Data61's policy is to be about AI and trusted systems research aren't in the AI field.
Argue the policy is bad! Or argue there should be a separate thing to support sel.4. Both of these are reasonable stances.
But as soon as Data61 said their future was in AI the writing was on the wall for sel.4.
The SEL decision was made on false premises and dishonestly represented to its staff, to the media and to parliament. I think that's pretty bad.
(Disclaimer: I worked in CSIRO in 1988)
But it wasn't at all?
Data61 said they were pivoting to AI. Trusted Systems research just isn't AI research. As the other reply here says, all the quotes in this post are micro-quotes taken out of context.
The context is "Data61 is AI". It's very possible to disagree with that policy decision, but saying it is lies is just wrong?
It is understood CSIRO has a strong bias now to self-funded work, which goes to commercialisation. its dogged it since the formation. The problem is that the IPR vesting and profit outcomes of work done can sometimes be misunderstood against wider gains. Standards work (for instance) often generate no profit. Does that mean CSIRO should stop doing them?
Data61 "was" formal methods. A lot of their work was in proofs. They applied the proofs to etherium, and they generated a huge amount of interest in formally verified blockchain contracts (which btw I think are bunkum) but the actual value was in the understanding of proof systems and verification. Australia is going into the Aerospace sector quite heavily. This is an area where verified systems are very important.
The "AI" component of proof is slim. Arguably there is an ML component, but having continued role in SEL and trusted platforms and applying proof systems and verification was a good thing in my opinion, and was not justified in being ditched. It looks to me like a poor long-term engagement strategy to walk away from a foundation.
It's fairly easy to take any piece of text, and present a compelling list of tiny sentence-by-sentence factual rebuttals that may or may not be correct, and then present them as a list of reasons why the text is wrong, and make it look really wrong because look at how many factual errors there are! The longer the source text, the better.
You can do this with almost any text, and while it is possible to respond to all these claims, it's incredibly time consuming, and by the time you have compiled a response, nobody will care and the damage is already done.