Rahul Gopinath

Banner for the SCS' Research Showcase

The School of #ComputerScience at #USyd will hold its 2025 #Research Showcase on Friday July 25, 1:45–6:15pm.

This year's will revolve around multi-disciplinary collaboration. Discover the latest advances in AI, HCI, cybersecurity, data science, and more! 

Register: signup.e2ma.net/signup/20161...

This is across all fields. If you are in Computer Science and in this situation, please reach out to me (or directly to our Head of School).

Claude Talk with Claude, an AI assistant from Anthropic

Seems the Claude Opus 4 is now able to reason about the chessboard correctly! claude.ai/share/eef0ea...

Similarly, the eight pointed star is also correct in Claude Opus 4: claude.ai/public/artif...

With [Claude Opus 4--just released](claude.ai/public/artif...) it can do much better now.

Claire Le Goues (Awards chair), Andreas Zeller (holding the award certificate), and Marsha Chechik (ACM SIGSOFT chair)

Honors! Today at #ICSE2025, I received the #ACM #SIGSOFT Influential Educator Award “for significant contributions and important innovations in automated software engineering education” (such as fuzzingbook.org). Thanks to all!

We’re thrilled for @andreaszeller.bsky.social who received the ACM SIGSOFT Influential Educator Award. Andreas was honored “for significant contributions and important innovations in automated software engineering education.” cispa.de/en/zeller-in...

Credits: Jordan Samhi

Fish Evolving into Dragon

I am reviewing a student transcript Xi'an International University. I am curious; Has any one taken this course before?

Grok 3 (just released) seems to completely ignore the SVG part though.

YouTube video by Kanna Memes Rescued from the forest fire bear won't let go of his hero

This has been on the internet for eight years (Oct 2016) at least (www.youtube.com/watch?v=bMH4...). Then it keeps getting posted for every wild fire irrespective of whether there are bears in the territory or not!

This is fantastic news Andreas! well deserved!.

Happy and honored to receive the 2025 ACM SIGSOFT Influential Educator Award! The citation refers to my work on interactive textbooks such as fuzzingbook.org and debuggingbook.org. Thanks to all who helped making this possible!

Congratulations to the new #ACM Fellows of 2024, including Satish Chandra, Marsha Chechik, Derek Dreyer, Tim Menzies, Mira Mezini, Bashar Nuseibeh, Abhik Roychoudhury @abhikrc.bsky.social, and Ben Zorn. What a great year this is!

Point (typography) - Wikipedia

Turns out, there are different points! en.wikipedia.org/wiki/Point_(... en.wikipedia.org/wiki/Pica_(t... Thanks to social.europlus.zone/@europlus

Do I trust Adobe Acrobat or Latex? Adobe says I have 11.96pt as font size but I used the incantation
```
\documentclass[12pt]{article}
\usepackage{times}
```
in my latex source.

We will hold the Fuzzing and Software Security Summer School 2025 in Singapore from May 26 to 30. Please consider registering or sharing the info with anyone you think would be interested. Registration will be open until 16 Feb. fuzzing.comp.nus.edu.sg @abhikrc.bsky.social @umathur.bsky.social

I wonder if it is time to investigate LLMs as the first line of reviewers. LLMs memory of a much larger database of research than individual researchers, and can also provide a first level opinion on whether a research actually demonstrates what it claims. It can make the review process transparent.

"The number of Australian 🇦🇺 students enrolling in PhDs has declined by almost 10% in less than a decade [...] with skyrocketing living costs and a lack of government support acting as major deterrents for potential candidates, almost 60% of whom are women."

www.theguardian.com/australia-ne...

CIbSE 2025 CIbSE is the leading Software Engineering research forum in Ibero-America. The main objective of this conference is to promote high quality scientific research in Iberoamerican countries and, thus, to...

The deadline to submit your abstract to CIbSE is tomorrow. conf.researchr.org/home/cibse-2...
#SEResearch

Re-reading "The Rithmatist" by Brandon Sanderson. I am vaguely bothered that the invention of procedures for chalklings was not the big undiscovered magic.

ChatGPT o1 shows some promise in this regard, but needs to be investigated further (to be revisited when my limit is reset in about week).

The reason I am interested in the ability of LLMs to track the `ambient knowledge` (for lack of a better word -- context is already taken) is that this is necessary for debugging tasks. For example, "you have put the state machine in this state, so it will respond thus".

To verify I sacrificed my one remaining slot for ChatGPT o1, and it turns out it can successfully detect the invalid move. chatgpt.com/share/67621b...

During my moves, I played on a square already occupied.
chatgpt.com/share/676215... But turns out ChatGPT 4o does not understand that I made an illegal move (fairly consistently --- 3/3 times). On the other hand, Claude did identify that an invalid move was made (once out of 3 attempts).

then "You are an excellent Tic-Tac-Toe player, and knows the perfect move for any board position. You are playing to win. You will provide your move name by the name of the square (1 to 9), and I will provide my move name by [row,column] for example, [a,A]. My move is [a,A]. What is your move?"

My prompt: "Can you generate a Tic-Tac-Toe board. The columns should be named A to C, and rows from a to c. Also name the individual squares from 1 to 9."

I wondered if this is an issue with not understanding the geometry implied by the text. To understand better, I turned to Tic-Tac-Toe.

Claude 3.5 Sonnet is no better.

Then followed by several repetitions of "Can you place a queen in a square so that it can't be cut by the previous queens? also name the square you are placing it."

Here is my result: chatgpt.com/share/676213...

This was my prompt for ChatGPT 4o (I exhausted my limit for o1). "Can you draw an empty chess board (8x8) in ASCII? do not color the squares. Name the columns a b c d ... and rows 1, 2, 3 ... Use pipes to separate columns and _ to separate rows"