When my preoccupation with this began last summer I imagined it was still years away. These are digital assistants which augment your capacity for reflexivity through learning about your activity, with a view to optimising your life in your own terms. But the first generation of digital daemons will be coming to the market soon(ish). I suspect we’ll see specialised daemons optimised for different areas of activity initially, which will eventually give rise to general purpose daemons:
What would such an assistant look like? Let me introduce you to Amber, one such prototype of a digital assistant developed at Microsoft Research. The project was led by Everlyne Kimani, and I was fortunate to be involved in this project along with other colleagues at Microsoft Research. Amber is a simple conversational agent that learns about a person’s behavior and gives recommendations to help them perform their best: when to take breaks so as to not get fatigued, to get unstuck from social media, and to get back on task if distracted. This agent learns about a person using various sensors like video and audio to detect things such as facial expression, body pose, and voice prosody (to infer mood and energy), physiological sensors (to measure stress), as well as computer logging (to see what tasks a person is working on).
https://gloriamark.substack.com/p/from-distraction-to-deep-focus
The problem with having this angel on your shoulder isn’t the reliance on technology, it’s the likelihood that it will be commercialised through a surveillance model. I’ve been surprised at how strongly I feel this intuition, that these developments are in principle a good thing. I fully accept that the prototype Gloria Marks is developing isn’t intended to be commercialised in this way:
The goal is for you to have a personal agent that resides locally on your computer, so that you would own your data and not a tech company. Your data that the agent collects to learn about your content preferences and your preferred behavior should be under your control and not under the control of tech companies.
https://gloriamark.substack.com/p/from-distraction-to-deep-focus
But the data which is collected through such a system, the initial behavioural and biometric data combined with how you respond to it, would be incredibly valuable when linked to other data points about your activity. I struggle to see how this wouldn’t be commercialised through surveillance, or how it could remain entirely locally operated, particularly given the compute requirements necessary to link together different specific functions into a generalised assistant. Imagine advertising which has something of the quality of TikTok in its capacity to inculcate compulsive behaviours based on real-time analytics of what you can’t help but click on.
Mark Carrigan 