While no one in their right mind would reject reductionism (cf. Douglas Hofstadter), in this case the sum of many small research projects can never be as helpful as a single, collective research project. 4/4

If we want academic research to accurately assess AI security risks, we need to equip them with sufficient resources to simulate large-scale capability. This could be achieved by having research insitutions across nations work together and provide resources. 3/x

If researchers cannot simulate methods used by large corporations or governments like large-scale fine-tuning and large agent-community approaches, they will consistently underestimate the persuuasibility of AI models, communicating a false sense of security towards policymakers. 2/x

I do agree. However, for (1) I do think it is a relevant metric as the software itself influences which questions are being asked through affordances. And OpenAI chooses to display certain „example prompts“ and advertise ChatGPT in a specific way which does not rule out „wrong“ questions.

Yes. But. 1) a valid test could be constructed by sampling real user questions and thus answering whether user behaviour matches software capability and 2) even OpenAI doesn’t quite know what question types work, and none work consistently, so how could users be expected to know?

Depends on the party affiliation of their CEO I guess.

Maybe you should offer a seminar on writing term papers using AI. Or just on how to write term papers in general. Just a thought.

Those are the people able to pull the „I“ out of „Information Society“

Über Prag schreibt ja auch niemand

Es gibt, weiß nicht ob du’s weißt, die revolutionäre Option: nicht draufklicken

Man könnte meinen es ist Weltkultur!

BahnCard ab 16, aber unter 27. Bin da altersdiskriminierend.

Let’s have them grade bachelor thesis’s based on one look at the bibliography.

Eine Topfpflanze, die über 6 Jahre alt wird, wäre für mich ein so großer Erfolg, dass ich bereitwillig ein Flexticket für sie kaufen würde.

💯

Maybe it would spark a debate on whether blood transfers heighten PFAS levels. So a medicine student tells me.