Monday, July 22, 2024

Welcome to the Next War: the AI Triple Black Box and Accountability

Professor Ashley Deeks
Greeting from "The Double Black Box: National Security, Artificial Intelligence, and the Struggle for Democratic Accountability" by Professor Ashley Deeks, University of Virginia. This is a public part of the conference "Anticipating the Future of War: AI, Automated Systems, and Resort-to-Force Decision Making" hosted by the Coral Bell School of Asia Pacific Affairs at the Australian National University. Professor Deeks' thesis is that defence AI is a block box both because even the programmers don't know what it is doing and if they did it would be secret. I suggest the situation is not that bad: it is possible to build AI systems which can be asked why they made a decision. But as the recent Microsoft/Crowdstrike failure shows, even non-AI systems can do surprising things. There is also cause concern, as Professor Deeks pointed out, due to the scale of use.

At a practical level it is not that difficult to test if an AI weapon is at least as reliable as a human operator. This could improve procedures by making explicit the decision making processes. There will be pressure to use advanced automated systems, just as there are for current simple ones, such as mines. 

Professor Deeks is presenting a US-centric view of the issues. However, the US is not a leader in development of AI weapons. Any country with a university having a computing school has the capability to make advanced AI weapons. Recently I was assessing a university student project for a small autonomous vehicle. This was for civilian purposes, but one version was tracked, and just needed a weapon added to be a robot tank.

The problem, I suggest, could be far harder than Professor Deeks suggests. The magic sauce for an AI weapon is in the software. The physical weapon can be upgraded over the air to have new capabilities. Some of this has been seen with missiles, where air launched missiles have been adapted for surface launch & surface for air. An example is the US Navy's SM-6 ship missile adapted for air launch against surface, air and space targets. Deciding of something is an anti-satellite weapon or not is a matter of software. 

Professor Deeks mentioned her paper "The judicial demand for explainable artificial intelligence" (2019) which argued for lawyers to get AI savvy. Some are thinking tech, such as Herbert Smith Freehills.

Saturday, July 20, 2024

Commenting on the Great Computer Outage of 2024

My phone started ringing on Friday afternoon. First it was the SciMex Science Media Center in Adelaide: could I talk to the media on the computer outage? What computer outage? I had been driving interstate and hadn't noticed many shop, airline, and other systems were down. What I had noticed were that both ABC Radio Sydney, & ABC Radio National were off the air (more correctly  broadcasting  apology loops, indicating the transmitters were okay but programs were getting to them).

At this stage the indication was it wasn't a cyber attack, and was not at the network level (my mobile phone still worked). It was the operating system. So I made a general comment to go out from SciMex to the media. At this stage the ABC had a report suggesting it was from the Crowdstrike security software

My phone then starting ringing. Sky new wanted to interview me, but their Zoom and phone interview facilities were not working (due to the it outage?), and it was not feasible to get to the studio. I talked to ABC Radio Queensland, who said they had one microphone and a CD player working. A little known fact is that if all the fancy automation fails in an ABC studio, one microphone is connected to the transmitter for emergency broadcasts. At the end of the interview I asked them to play "A Walk in the Black Forest" (the only track Radio Goodies had), but the joke went flat. 

One 24 hour TV news network wanted me to come to their studio across town because they could not do a Zoom or phone interview (presumably because the equipment for that used Microsoft Windows). I was tempted to suggest they hold the phone up to the camera.

I made the right call to say it was not a cyber attack, & resisted the temptation to criticize Microsoft Windows. Something I found surprising was the range of devices apparently running Microsoft Windows. Why would you use it for an airline or supermarket machine, rather than an operating system designed for real time embedded applications (such as one of the Linux variants)? 


See also (updates):

Friday, July 19, 2024

Chatbots for More Rounded Employable Graduates?

Greetings from the weekly ASCILITE MLSIG webinar. One of the members had a positive report on using Cogniti (developed at University of Sydney), to build chatbots to help students. With this, the software simulates a patient in conversation with the student acting as a therapist. The chat-bot then switches to tutor more and provides feedback and advice to the student. It occurred to me the same would be useful for students "soft" skills.

Many STEM students have difficulty with the part of the job where they have to talk to people, especially non-technical clients. This also creates problems when talking to potential employers. It may seem odd to suggest the students talk to a machine to imp[rove personal communication skills. However, this way students can get a lot of practice with an infinitely patient tutor. Also client and work communication is increasingly using digital technology. In a way reality s becoming more like the simulation: you apply for a job not by writing a letter but via a web form, do online tests & get interviewed via Zoom. The graduate will likely communicate with their client, and perhaps colleagues, mostly online. So talking to a chat-bot online will be a more realistic simulation of the workplace, than talking face to face in a classroom.

Wednesday, July 17, 2024

Thinking assessment in the large as an answer to AI

Greetings from the CRADLE Seminar "Assessment beyond the individual unit/module and AI". Associate Professor Jason Lodge is talking about how to see how students progress over time, rather than assessing in small packets of instruction. While he did not explicitly say it, I assume he envisions this makes it harder for the student to cheat, using AI, or otherwise, as they would not be showing consistent progress. 

Mentioned by someone was "Assessment reform for the age of artificial intelligence" (TEQSA, 2023). 

Margaret Bearman took us through the logic of current unit based assessment and asks about "big picture" outcomes. My reaction was "Inst that what capstones are for?". You have the student do a big project at the end of their study, where they have to demonstrate the skills needed. 

Surprisingly, there was little mention of AI, which is refreshing. The approach is to get the assessment right and cheating will be harder, however it is done.

I asked the panel:

'Will technology help? Could we give the AI each student's CV and have it suggest what degree requirements they have already met? I help out with applications for course credit and there is a lot of stuff students have done they really don't have to do again. More than once I have thought we should have the student teach the course. ;-)


The next seminar is: "Second Handbook of Academic Integrity (2024) launch".

Wednesday, July 10, 2024

CIT Timber High-rise Campus Going Up


Greetings from the Woden Town Centre in Canberra, where the timber frame of the new Canberra Institute of Technology is going up.

Digital Technologies education in Australian schools

The Australian Computer society has released "Tech skills for the next generation: Digital Technologies education in Australian schools". The report makes 11 recommendations, in 4 categories. Initial Teacher Education is the area where I suggest there is the most scope. Teachers could be trained using digital technology to effectively use it in their teaching and administration. Rather than this being seen as adding to the burden of teaching and administrative responsibilities teachers already have, digital technology could be a way to make their jobs easier.

Ensuring there are accessible ready-to-use teaching resources

1. Expand support for, and increase visibility of, the online Digital Technologies Hub to ensure teachers have access to best practice exemplar teaching modules for the DTC.

2. Improve schools’ internal information management processes regarding digital teaching resources to ensure they reach teachers who need them in the classroom.

3. Support cross-fertilisation amongst professional associations and communities of practice for the DTC.

Embedding digital-readiness training in Initial Teacher Education (ITE)

4. The Australian Institute for Teaching and School Leadership (AITSL) should incorporate into ITE accreditation a requirement that ITE programs demonstrate their capacity to prepare our future teachers to: 
• teach with digital technologies (as expected by AITSL standards)
• use digital technologies within all learning areas (including Digital Literacy development)
• teach the F–10 Digital Technologies subject and/or senior secondary computer education courses.

This could be supported through the Australian Technologies Teacher Educators Network (ATTEN) to provide end-user input from Digital Technologies teachers based in each state and territory.

Supporting ongoing professional development and training for teachers

5. Ensure that training courses suitable for teachers are available and accessible across all essential areas of digital technologies knowledge and skills.

6. Identify and promote existing recommended courses that provide training in software tools and core principles of digital technologies for teachers of all year levels.

7. Invest in initiatives that support teachers to attend suitable training for digital technologies skills and in turn this will increase the number of skilled teachers at each school.

Elevating awareness of the Digital Technologies Curriculum in the community

8. Empower parents with the tools and capabilities to understand and communicate at home the value of digital technologies, including the types of technology careers that can be pursued and how the skills can be applied to solve problems in a range of industries.

9. Ensure that tools and capabilities that empower parents are inclusive and increase visibility of underrepresented groups in STEM fields, such as women and girls and Aboriginal and Torres Strait Islander Peoples.

10. Establish a national coordinated data collection of DTC learning outcomes, and communicate these outcomes to the community to build better understanding and awareness of learning and career outcomes.

11. Recognise and reward excellence in digital technologies education to increase visibility to parents and the education community and promote best practice-teaching in Australian schools.

Thursday, July 4, 2024

The Cafe and the Education Revolution

Greetings from the just opened TASA Cafe in the sport building at the Australian National University in Canberra. TASA adds Philippine cuisine to the selections on campus. I am having the pulled pork with coleslaw on a bun. More importantly, it is just across the road from my office in the school of computing. The campus cafes have an important role for informal discussions.

Already I have scheduled a meeting on exemptions and credit for recognition prior learning and experience, in the cafe. Granting students credit for what they did somewhere else is something academics are reluctant to do. This is partly out of a concern for standards, but also because it is not something part of academic training. Some of this is relatively simple: a course in discrete mathematics is much the same in Sydney or Shenzhen. However, soft skills are another matter: a course where students work in a team is not the same as one where they just read books about working in a team. Is work experience at a computer company in another country equivalent to Australia?

Wednesday, July 3, 2024

Innovation in Canberra



Greetings from the foyer of ACT Government HQ where they are hosting the 109th First Wednesday of Canberra Innovation Network. The ACT Chief Minister is doing the honours before the pitches.

ps: Interesting building.



High ceilings give low exam results

I suggest Bower, Broadbent, Coussens and Enticott receive an Ig Nobel Prize for their paper "Elevated ceiling heights reduce the cognitive performance of higher-education students during exams". The researchers found the higher the ceiling, the worse students do in exams. This is based on analysis of 15,400  exam results at three Australian campuses over eight years. At first I thought this an April fools day joke, but it seems to be real serious research. The work is deserving of an Ig Noble, recognizing research which has humor, but provokes thought. The key point of the research for me is not that high ceilings disadvantage some students, but that exams do. This can be corrected, I suggest, not by lowering ceilings, but by replacing exams with better assessment techniques.

The research comes at the time when exams should be on their way out. This is  like inventing a more efficient steam engine in the 21st century: an obsolete dangerous technology which no amount of technical improvement can save. Exams cause students stress (I have spent decades avoiding any course which had an exam). That a high ceiling might increase stress is interesting, but I doubt it could be lowered enough to make me comfortable with an exam. I stopped setting exams around the time I stopped giving lectures (2018).

The obvious factor which would cause the effect is the size of the room (bigger rooms having higher ceilings). However, the authors appear to have considered room size as a factor, and controlled for many other possible causes. One not mentioned might be that large rooms may have students from different classes taking tests at the same time, which could make students feel less comfortable.

If this is a real effect it could be easily corrected for by lowering the perceived height of the exam room ceiling. This could be done with lighting.

Reference

Bower, I. S., Broadbent, J., Coussens, S., & Enticott, P. G. (2024). Elevated ceiling heights reduce the cognitive performance of higher-education students during exams. Journal of Environmental Psychology, 102367. https://doi.org/10.1016/j.jenvp.2024.102367