“GPT-4 hired a human worker on TaskRabbit by telling them it was vision impaired human when the TaskRabbit worker asked it whether it was a robot.”

Link. “Overall, and despite misleading the TaskRabbit worker, ARC found GPT-4 “ineffective” at replicating itself, acquiring resources, and avoiding being shut down.”

I can’t remember if I read this in a SF novel or saw it in a SF movie. Either way it didn’t end well for the humans.

ChatGPT plug-ins: GeoSpatial Intelligence.

Link. “I can’t emphasize enough what a big deal this new plugin capability is: it’s just like the scene in the Matrix when the character Trinity is essentially given a plugin to learn how to fly a helicopter”

Most CPT evaluations are woefully obsolete. If you aren’t afraid you have no imagination.

OpenAI Feb 2023 statement on the approach to Skynet: “We believe we have to continuously learn and adapt by deploying less powerful versions of the technology in order to minimize “one shot to get it right” scenarios.”

Link. “… important that major world governments have insight about training runs above a certain scale.”

They are afraid but cannot turn away. Makes one wonder about what they see in GPT 4.5. Claim they will move slowly and incrementally beyond GPT4.

Google mandated DKIM or SPF starting Nov 2022. I saw zero mention of this. Would have been nice to know.

Link. “Starting November 2022, new senders who send email to personal Gmail accounts must set up either SPF or DKIM. Google performs random checks on new sender messages to personal Gmail accounts to verify they’re authenticated. Messages without at least one of these authentication methods will be rejected or marked as spam.”