I recently subjected it to my favorite LLM test: can it write political and corporate feedback e-mails for me? I plan to put a full writeup on my blog later, but here's the gist:
Some of the e-mails were actually usable. I.e. I think I could send them to the real live target without feeling ashamed. However, not *all* of them were usable. Which means I'd still have to proofread everything to make sure GPT is not inserting falsehoods or going off on a weird tangent. It doesn't know anything about me personally, or about recent events that aren't in its training data (obviously). Trying to help it out by including that kind of information in the prompt had mixed results. Sometimes it liked to just recycle my prompt text, which makes me feel like I'm still doing a lot of the work of writing the e-mail. If I tried to generate multiple e-mails on the same topic (which would be the goal - I like to bother my representatives on the regular), they started looking very similar. Telling GPT to "rewrite that in different words" just produced another copy of the same output. ------------------------------------------ Artificial General Intelligence List: AGI Permalink: https://agi.topicbox.com/groups/agi/T357de2f46d742838-M953f49a470b19ab984675ae9 Delivery options: https://agi.topicbox.com/groups/agi/subscription
