Reliability At Scale // Conundrum
This is phrased as if it is an objective theory which has found truth in the world. It is not. It is written in that style because that style lends itself to being a heavily abstracted theory. All mistakes are my own.
A framing device for the role of Informatics in our shared world. This is phrased as if it is an objective theory which has found truth in the world. It is not. It is written in that style because that style lends itself to being a heavily abstracted theory. All mistakes are my own. Theory Informatics is principally the study of manipulation of information. This manipulation is typically done along a given goal, which consists of the description of the input and the desired output. A so-called computation of this output can be called reliable if from the described input the output is produced
related reading
- Computers can be understood - Made of Bugsblog.nelhage.com
- How can LLM RL Work Despite Information-Theoretic Inefficiencyberen.io
- Computational Complexity Theory (Stanford Encyclopedia of Philosophy)plato.stanford.edu
- Scaling is subtler than it seemsberen.io
- aaronson-philosophy.pdfcs.princeton.edu
- Towards a scale-free theory of intelligent agencymindthefuture.info
- The necessity of machine learning theory in mitigating AI riskmishabelkin.substack.com
- Embedded Agency (full-text version) — LessWronglesswrong.com
- The Tragedies of Reality Are Coming for Youalexirpan.com
- Scaling in the service of reasoning & model-based ML | Yoshua Bengioyoshuabengio.org
- Out of the Tar Pitcurtclifton.net
- Taking LLMs Seriously (As Language Models) — LessWronglesswrong.com