Search This Blog

Tuesday 18 June 2024

Celebrating cute ChatGPT calendar creation capability causes calm contentment

A cute thing that ChatGPT can do: if you have a PDF (or similar) of a schedule, it can turn it into a .ics file that you can import into your calendar.

When you buy a ticket with NSW Trainlink you just get a PDF -- I'm surprised that it isn't still hand-written paper tickets to be honest -- so this just solved one of the petty annoyances in my life.

Saturday 8 June 2024

Generative AI and cybersecurity: governance, risk, compliance and management

This is a quick summary of a much longer substack: and a ~40 minute video:

The world of AI is rapidly evolving. AI researchers are divided on whether superhuman intelligence can be achieved by scaling up current technology or if fundamental breakthroughs are still needed. 

Key AI Concepts for Cybersecurity Professionals:

  1. AI Alignment: Controlling AI systems that may become smarter than humans.
  2. Explainable AI: Understanding why AI programs make certain decisions.
  3. Agentic AI: AI systems that can perform actions autonomously.
  4. Prompt Injection: A major security concern in AI systems.

Impact on Management:

  1. Distinguishing between genuine understanding and AI-generated artifacts.
  2. Using AI for performance improvement plans and automation.
  3. Dealing with potential fraud and impersonation using AI.
  4. Monitoring AI-driven automation by employees.
  5. Adapting to increased use of speech recognition and document management.

Corporate Governance and Responsible AI Use:

Pre-2022, AI governance focused on controlling the training process. Post-2022, governance must shift to maximizing benefits while managing risks. Key challenges include prompt injection, hallucination, and lack of moral sense in AI systems. Blocking AI usage may lead to data leaks to less reputable companies. Organizations must embrace AI while implementing proper guardrails.

Required Capabilities for AI Governance:

  1. Observatory: Monitoring AI usage and its impacts.
  2. Reward Giving: Incentivizing staff to automate tasks responsibly.
  3. Expansion: Propagating new ideas and methods for AI use.
  4. Financial: Balancing AI costs with staffing savings.
  5. Cybersecurity: Defending against new attack vectors.
  6. HR Responsiveness: Managing job role changes due to AI automation.

Key Processes:

  1. Audit/Discovery/Inventory Management: Identifying new AI activities.
  2. Incident Response: Handling prompt injection attacks and rogue employee actions.
  3. Rapid Iteration on Education and Training: Keeping staff up-to-date on AI capabilities.


Few regulations currently exist for generative AI usage. China requires AI systems to act in the benefit of social harmony. Japan allows training models on copyrighted works, potentially leading to faster AI adoption. India requires registration for AI model training, hindering the development of language models. The USA has proposed limitations on large-scale AI training.

Monday 27 May 2024

Reflections on student research projects

I was supervising 7 Masters of AI / Masters of Data Science / Masters of Cybersecurity students this semester.  Reach out to me if you are looking for people with these skills and I'll introduce you; also, if you have small industry projects that need doing, there are cohorts from Macquarie and ANU next semester.

On Friday we had the final presentations from my students and the rest of the cohort. Observations:

  • Very few students trained their own computer vision models. Viet (one of my supervision students) was one of the few that did, and was quite successful. Instead, research in computer vision is now often asking "how can I prompt Claude/ChatGPT/Gemini able to get the right answer?" As Marco found, this is often not a trivial exercise, is vastly more inefficient than a trained model, and not necessarily all that accurate. Yet.
  • Explainability is a big deal. Both Viet and Himanshu looked at explaining the results of their models. Shout out to Himanshu for delivering his presentation in the dark (I don't know why the lights weren't working...) after racing to get there after his car broke down.
  • We don't know whether blockchains can coexist with quantum computers. The problem is obvious: RSA or ECC public keys are too easy for quantum computers to break. Lots of solutions have been proposed: Suyash, Proshanta and Pradyumn all found problems in the different solutions they looked at.
  • In research, ChatGPT wasn't all that popular: about a third of the cohort that used an LLM in their research used ChatGPT. Open source models like llama and mixtral were far more popular in academia than in business (that I have seen).
  • About 90% of projects didn't use LLMs at all in their research. They might have used it to clean up their writing or other "mundane" tasks, but it didn't play a part in the research process itself. I am going to try to track that over the next few years as I expect it will drop.
Reflecting on the reflections:
  • For $20/month, you can have access to all the necessary tools to do world-leading computer vision tasks. It has been a long time since the forefront of technology has been that accessible and that cheap anywhere in anything. Truly we live in wondrous times.
  • As we move into a (possibly) AGI+quantum world, being confident that you can trust complex systems will become much more important than building the thing itself.

Monday 20 May 2024

10 years of this blog

I started this blog 10 years ago.

The most popular post was at the start of the pandemic, where I wrote a detailed post How I teach Remotely I had been doing that for many years before the pandemic; I could have written it before any lockdowns started.

The next most popular post was Something funny is about to happen to prices which talked about some of the weird things that I predicted would happen as we started getting negative electricity prices. I assumed that Australia would develop industries that made use of nature's subsidy; but alas, Australian industry is too uncompetitive even if electricity is free-er than free.

After that, there were some posts about how to fix things in HP DataProtector. I think I was expecting the blog to be mostly about data protection: the first post was about HP's confusing line-up of overlapping backup products, and identifying when it was cheaper to go with one choice over the other. HP no longer have overlapping backup products -- they sold them all to Hewlett-Packard Enterprise Entco Microfocus OpenText who now has a confusing line-up of overlapping backup products (different ones). Those a

Anyway, it looks like what I should be writing about is extrapolations into the future of what I'm seeing happening now. That's good. I have started writing a book on what we should expect in the next 5 years in AI, so I'll post excerpts as I write it.

Saturday 27 April 2024

Do Australian banks care about AI?

Both the 2022 and 2023 Expert Surveys on Progress in AI have identified that acting as a telephone banking operator (including card services) should be fully achievable by AI by 2026. So a forward looking bank would be reporting their AI expenditure as a way of signalling to the market that they aren't going to be left behind. Will any of them do that this year?

If you want to bet one way or the other on this question, here's a prediction market on it:

Sunday 21 April 2024

Mysterious latitude and longitude coordinates

There's the coincidence of the Pyramid of Giza's latitude being a very similar-looking number to the speed of light. It's nonsense of course, because when the Pyramid of Giza was being built (roughly 2600BCE):

  • No-one knew that the earth was round (so the concept of latitude didn't make any sense). It was more than 2000 years later Eratosthenes calculated the circumference of the earth (around 200BCE). He also made a map that connected the Atlantic Ocean with the Gulf of Arabia, and had the Equator substantially south of all of Africa.
  • No-one described angles in degrees, only ratios of lengths or ratios of quadrature (so describing a latitude in degrees wouldn't have been possible). Angles were probably Babylonian (2000 years later); Aristarchus of Samos (2300 years later) is the first writer we know who used them.
  • They didn't have decimal numbers; they didn't even have fractions with numerators. It's only in the Middle Kingdom period of Egypt (2000BCE, about 600 years later) that we see fractions of any kind.  Even then they couldn't represent 2/3 other than saying (1/3 + 1/3, or 1/3+1/5+1/12+1/20). It was the height of mathematics six centuries later than the Pyramid of Giza to be able to say that 1/5 + 1/12 + 1/20 = 1/3. So they wouldn't have been able to represent 29.979.
  • The length of the metre was defined as being the distance travelled by light in 1/299792458 seconds. That is, we chose that the speed of light would be that in 1983. If we had chosen 1/299792459 instead, presumably the great pyramid would have shifted north a bit. (I suppose some physicist might have been playing a prank on the world by making the speed of light match the latitude of the Great Pyramid of Giza. But that's not a mystery then, and it doesn't involve ancient aliens. It just involves a physicist with a sense of humour.)
    • Actually, the metre was defined before that. One of the complaints of the French revolution peasants was the inconsistent measures imposed by their pre-revolution feudal lords, who would change the size of measures when they were buying vs selling. So they defined the metre by saying "a metre is one ten-millionth of the distance from the equator to the north pole via Paris." They knew that the Earth wasn't completely round, and it solved the problem that they had, but they still got it a bit wrong anyway.
    • Before that, a metre was the length of a seconds pendulum -- a pendulum that swings every two seconds is roughly a metre long, as long is it only swings a little bit and the flexibility of the pendulum arm doesn't count.  Anyway, yes, the ancient Egyptians had pendulums, but they didn't have the concept of seconds. They probably had some way of dividing up time into units smaller than an hour using water clocks, but nothing as small as minutes. They didn't know that pendulums swung at a constant rate, that wasn't until later, so even if they had had the concept, they couldn't measure it.
The bigger problem is that the exact coordinates of the Great Pyramid of Giza are 29.979167N, 31.134167E and the speed of light in a vacuum is 299792458m/s. So it's close, but it's out by about 8 metres. (And why didn't they put it about 28km further east, at 31.415926E ? That might have been weirdly convincing of the existence of time travel.)

Which brings me to the point of this post. Here are some important things in the world that have either their latitude or longitude (measured in degrees) be a number that's close to a multiple of a 10th power of the speed of light. All of these are closer to this ideal than the Great Pyramid of Giza.

  • Port Neches Elementary School in Texas, Latitude=29.97925, Longitude=-93.95905, according to Geonames. Presumably this is the address of the entrance, if you just move 47cm south (18.5 inches), you'll be spot on. Fun fact for Port Neches Elementary School students: if you travel east or west (and don't deviate even a little bit), you'll eventually hit the Great Pyramid of Giza. Google maps puts Port Neches Elementary School well away from this, and says that the Subway opposite the Neches Federal Credit Union / Magnolia Church is the speed-of-light latitude.
  • Crowne Plaza Astor New Orleans Latitude=29.97922, Longitude=-90.1137. If you are the events manager for the Crowne Plaza Astor, I hope you have locked up the lucrative market for Physics Conferences at important latitudes. Just 2.87 metres north and you're on the light speed location. (Google Maps is again out-of-sync with geonames, and suggests the pedestrian crossing on North Tonti Street where it meets Aubry Street; it also suggests the Jewish Cemetry near the Hurrican Katrina Memorial... and the Mortuary Haunted House.) I don't know whether to believe Google maps or Geonames. I was on the Google Maps SRE team briefly in 2006--2007, but I spent the majority of my time on the maps team nowhere near any other maps team members, so I don't have an insider opinion on Google Maps accuracy.
  • The Omnipotent Missionary Baptist Church. This seemed to have disappeared from Google Maps, if it was ever there, which I guess is a thing you can do if you are an Omnipotent Missionary. (But what do Omnipotent Missionaries evangelise about? Themselves?) It was at Latitude=29.9792, Longitude=-90.03.
  • I tried to find a few places in China, but between inaccurate maps and other problems it's hard to be sure. The Wuhan Institute for Virology is just a little bit too far north to be on the light-speed latitude.
  • According to Google Maps, the Benedictine Monastery of Toumliline is very close. But Wikipedia puts it several degrees of latitude in a different direction. For Moroccoa, Geonames suggests the town of of Aguerd Issegane, which seems like cheating because it's a town... but it's surprisingly small.
  • Geonames also suggests New Orleans' East Bank Sewage Treatment Plant.
But why latitude? Why not longitude? (Well, we couldn't calculate longitude accurately until well after Galileo, because we had no accurate time keeping mechanism that could survive a journey. For a long time, the only option was to observe when eclipses of Jupiter's moons happened; compare that to an almanac of when they were supposed to happen, then look at the angle of the sun or a fixed star and work out what longitude you were at. Way beyond what any ancient Egyptian could have done.)

So for longitude 29.97925E:
  • It's quite close to the Chernobyl nuclear power plant (51°23′21″N 30°05′58″E) -- but the powerplant is about 8km too far east. 
  • Suri Suri Dam in Zimbabwe is very close to the light-speed longitude (-18.08597,29.97944)
  • Playing around with Google Maps, I found that the southern most shed of the University of Alexandria's Faculty of Agriculture Poultry Farm was is on the speed-of-light longitude.

But why latitude 29.97925? Surely 2.997925 would be better? And here we hit the jackpot:
  • Taman LEP 7 Stop (which I think is a bus stop in Kuala Lumpur). Latitude=2.99793, and longitude=101.64858. The KL bus transit authority has encoded the speed of light in the bus stop!

Here's how I did it. Download the Geonames database, unzip it, create a postgresql database:



 createdb geonames

Then in the postgresql session:

CREATE TABLE geoname (

        geonameid int,

        name varchar(200),

        asciiname varchar(200),

        alternatenames varchar,

        latitude float,

        longitude float,

        fclass char(1),

        fcode varchar(10),

        country varchar(2),

        cc2 varchar(60),

        admin1 varchar(20),

        admin2 varchar(80),

        admin3 varchar(20),

        admin4 varchar(20),

        population bigint,

        elevation int,

        gtopo30 int,

        timezone varchar(40),

        moddate date


\copy  geoname (geonameid,name,asciiname,alternatenames,latitude,longitude,fclass,fcode,country,cc2, admin1,admin2,admin3,admin4,population,elevation,gtopo30,timezone,moddate) from 'allCountries.txt'  null as '';

create index on geoname(latitude);

create index on geoname(longitude);

select * from geoname where latitude < 29.980 and latitude > 29.979 order by abs(latitude - 29.9792458);

And then likewise for different latitudes and longitudes.


  • Gungele is at Pi latitude. Latitude=3.14159, Longitude=28.14137. (So close to being 10 times e!)
  • VilledonnĂ© is at e longitude (to 6 decimal places) Latitude=47.27323, Longitude=2.71828
  • A beach in Tiomann Island (Pasir Gerengganin Malaysia is at e latitude. Latitude=2.718,  longitude=104.1724 

Monday 15 April 2024

Interview with the Australian Writers' Centre

I was interviewed about ChatGPT, Anthropic and other GenAI tools and their impact on jobs for writers. We ended up talking about speech recognition, why SEO might be becoming irrelevant and the computer gaming industry.