Introduction
This is my personal blog about AI. It contains essays, thoughts, and short notes about things that interest me.
Latest Posts
-
Generated For You
Social media recommendation algorithms predict content users will engage with. But optimising only for engagement risks filter bubbles and addictive doomscrolling. As these algorithms begin to prompt generative AI models, these issues will likely be exacerbated if platforms continue prioritising engagement over truth, diverse perspectives, and user wellbeing. -
Why I Write
I don't write to communicate ideas - but to generate them. I use it as a form of introspection through which I can notice gaps in my thinking, and uncover questions which I hadn’t yet considered. It is a tool that allows me to reach higher levels of thinking by highlighting what I don’t understand. -
X Is a Step Towards a More User-Focused Internet
Twitters transition towards X represents a step away from advertiser dependency and towards a more user-focused internet. It’s new revenue model allows it to prioritise user experience whilst also helping to combat bots and spam on the platform, which is an increasingly difficult problem due to recent AI capability advancements and ease of access. -
Mode Expansion Through Prompting
Mode expansion is a technique to help thwart mode collapse, which reduces LLM creativity and diversity. Mode expansion can be achieved through a mix of obedience finetuning and intelligent prompting. -
The Trojan Machine
As language models become more sophisticated and powerful, there is a risk that they may develop emergent capabilities or behaviors that are difficult to predict or control. These capabilities may remain hidden until specific circumstances arise that trigger them, making it difficult for humans to anticipate and prepare for them.
-
Prompt Injection
Prompt injection is where a user injects a malicious prompt via an input field which manipulates the intended output of the model. It is similar to the idea of SQL injection.
-
A Future Battle For Compute
Artificial superintelligence is, by definition, artificial intelligence that far exceeds human intelligence. If artificial superintelligence one day becomes a reality, we will find ourselves far too inferior to protect ourselves from it intellectually. Instead, we will have to employ the help of another artificial superintelligent agent. In artificial intelligence, an intelligent agent (IA) is anything which perceives its environment, takes actions autonomously in order to achieve goals, and may improve its performance with learning or may use knowledge. They may be simple or complex — a thermostat is considered an example of an intelligent agent, as is a human being