The result should be treated with a bit of caution, as a content analysis of forums and online articles, but an important finding nonetheless. It’s striking how this development features almost nowhere in debates about LLM use by students within higher education!
It suggests to me that LLMs are rapidly diffusing as a reflexive technology and that we urgently need to understand how to distinguish between harmful and helpful uses, in order to steer users in one direction rather than the other.
In general there’s an urgent need at this stage to crack open the black box of practice in order to identify relational modes of engagement, rather than treating singular practices (in effect thematic generalisations) as the organising principle for understanding real world LLM use.
See also Sam Altman saying that young people don’t make a decision without asking ChatGPT what to do, claiming the memory feature enables a context to emerge across their life and all the people in it:
He goes on to say (17 mins) that the single sign in which OpenAI are launching will be a basis to enable this personalisation to follow a user around the internet. The context accumulated through ChatGPT will in effect follow you around the internet.
Mark Carrigan 