Survey researchers: I'll discuss LLM "synthetic respondents" in a free webinar this week.
➡️ Synthetic Survey Data? It's Not Data
Thursday Feb 19, 12pm Eastern, hosted by Sawtooth
Free registration: sawtoothsoftware.com/resources/ev...
(P.S., I give no quarter on this topic 😆)
Not all examples are straightforwardly applicable to the social media context, but many are.
Compare this: en.wikipedia.org/wiki/New_Car...
To the AI equivalent: futureoflife.org/ai-safety-in...
Pick your favorite and we could discuss. Several reasonable, concrete comparisons to choose from: required crash test is red teaming, speed governors are kill switches, etc.
There are no shortage of concrete, realistic proposals drawn from industries that are already regulated: e.g., safety mechanisms (cars), marketing limitations (drugs), identity verification (finance), etc. Each proposal has benefits and drawbacks. We need to determine the acceptable balance.
Your comparison is strange, when the text generation machine can produce orders of magnitude more words in the same amount of time it takes a human to write 500 words on paper. Surely relative ease of production changes the calculus on the relative impact?
