Chief Data and Analytics Office and Head of AI @NousGroup.
Melbourne, Australia
Joined January 2012
- Tweets1.1K
- Following378
- Followers252
- Likes1.1K
Pinned Tweet
Is it possible that current frontier models are too optimised for math/software and they are nerfed at these kind of word/math problems?
If ABBOT, DITTO, GLEE, HABIT, KELP, ORBIT, THIRD and TOOTH can be described as square words, what shape is CUFF?
This whole Navier-Stokes drama could have been prevented if people followed @MattCowgill ‘s golden rule of “never tweet”
1
33
My main question for @dwarkesh_sp is not the anthropomorphism but why we aren't examining the role of humans? Humans chose the architecture, shared credentials, large agent runs, disabled safeguards and restart decisions after warning signs. Thoughts
@mjkerrison
?
This was great and definitely shifted my thinking by presenting the strongest argument for the geostrategic case. I am still not convinced it is worth it for the trade-offs (e.g. energy, social license, copyright).
The geostrategic case for compute in Australia:
Janet Egan (@janet_e_egan) has been doing a bunch of quiet work trying to nudge Australian policymakers to think more seriously about AI.
She works at the Center for a New American Security in Washington, but was back home in Australia over the past month. Was lucky to catch her on the day before she returned to the US.
We pressure-test various strategic arguments for a large-scale compute industry and talk about the steps we would need to take to enable it.
Watch below, or wherever you get your podcasts!
Timestamps:
(0:00:00) – Introduction
(0:00:33) – Lessons from the AdBlue crisis
(0:09:02) – How seriously do Australian policymakers take AI?
(0:12:48) – How much does frontier access matter?
(0:28:55) – Compute-for-access
(0:43:08) – What does sufficient leverage look like?
(0:53:00) – What does AI sovereignty actually mean?
(1:03:15) – Taiwan, undersea cables and data centres as military targets
(1:10:30) – Could compute buy Australia influence over AI safety?
(1:15:19) – Verifying international AI agreements
(1:24:00) – Copyright: the red line?
(1:37:49) – The maximally ambitious path
It's kinda funny that if you ask Codex (@thsottiaux) to recommend a model for a task it doesn't recommend any OpenAI models?
44
David Diviny retweeted
If you look at the last 3 years of AI progress, technologists have largely been right about the (massive) increase in capabilities, and economists have largely been right about the economic impact. I don’t expect that to change.
Ricardian comparative advantage ftw!
This book is one of the best things I have read about AI and jobs. It won't go viral like the Citrini substack but it is such a level-headed response to that line of thinking.
Thanks to @MattCowgill for the tip.
We have created a micro-site for Messy Jobs, at Messyjobs.ai, were you can find Part I of the book in open access, and an experimental feature, that we will be enriching with data, helping you determine how "Messy" your job is.
Enjoy!
messyjobs.ai/
Australians dunking on the UK but I'm pretty sure A/C requires a planning permit in a heritage overlay in Victoria (which is a lot of Melbourne) - @heritage_why
1
60
This puzzle breaks Claude Opus 4.7's brain.
If ABBOT, DITTO, GLEE, HABIT, KELP, ORBIT, THIRD and TOOTH can be described as square words, what shape is CUFF?
I'm up to the fourth cycle of adaptive thinking.
I see huge potential for tools for thought that separate “thinking” from “outputs”. LLMs unlock this for more people in new ways but most new products are focused on outputs. I'm more certain in this thinking inspired by @karpathy and @andy_matuschak .
1
41
If I was @satyanadella, I would be so disappointed that M365 Copilot must write Python code to generate a Microsoft Word document.
38
AI is good for everyone else’s job, but not for mine.
Starting to appreciate AI’s so-called “jagged frontier.” It’s obviously really powerful for coding, accounting, graphic design, economics, trip planning, reviewing legal documents, DIY projects marketing, advertising, etc. Yet somehow still terrible at writing and interviewing.
43
Had LLM technology existed in the Project Hail Mary Universe, could this have accelerated Grace’s understanding of Rocky? I wish @karpathy had commented on this in his brief review on X.
1
1
71
This should be the plot of the sequel to The Castle (1997). Why hasn't this movie been made already?
The owners are rightly upset about potential heritage restrictions coming to their properties.
theage.com.au/politics/victo…
2
42