Bubbles and Technical Debt

It's reasonable old news now, but the Auditor General is investigating the investment of nearly $1 billion in public money. Thank goodness, the whole thing has always stunk. This is a good time to remind everyone that, which an incredible feat of engineering, Quantum Computing is a bubble and still can't even do basic things.

In AI shit-posting news, AI is worse than humans in every way (and the government agrees). Molly White make the convincing case that, while AI isn't useless, it comes at a huge cost and it just isn't anywhere near useful enough to justify that cost.

Ezra Kline did a series of podcasts on AI that I enjoyed because of the nuance and candidacy of them (which is all I ask of these discussions). Two that stood out was the interview with Holly Herndon about her use of AI in composing music which is quite amazing. It's a hyper localised small scale project which is where AI projects tend to be most useful and have the highest technical barrier to entry (a point I believe Herndon would agree with). The other one was the interview with Dario Amodei, the CEO of Anthorpic because it confirms for me that these LLMs aren't really gaining any efficiencies. There is a linear relationship between the quality of the LLM and the amount of computing power (and size of the data set) you throw at it.

While we're on it, here is yet another comprehensive explanation of why AI is really quite limited.

The Federal Government released it's Privacy Act updates and, in no shock to anyone, is a disappointment and fails to bring Australia anywhere near up to date with a modern internet environment. This stuff is basic. In related news Meta is scraping all non-EU resident's data to train AI models, because the EU has a modern Privacy Act.

It's been doing the rounds, but this episode of Risky Business is an interview with ASIO Director Mike Burgess is worth a listen in terms of the way he views privacy and surveillance.

I've been thinking a lot about technical debt lately and was reminded of this very old piece by Joel Spolsky:

There’s a subtle reason that programmers always want to throw away the code and start over. The reason is that they think the old code is a mess. And here is the interesting observation: they are probably wrong. The reason that they think the old code is a mess is because of a cardinal, fundamental law of programming:

It’s harder to read code than to write it.

The other tech debt article I read was this piece which breaks down how to tackle tech debt and finding the productivity boosts as well as delivering value which I think is well worth the extended read.