Paper: New benchmark testing if models like GPT3 are truthful (= avoid generating false answers).
We find that models fail and they imitate human misconceptions. Larger models (with more params) do worse!
PDF: https://t.co/3zo3PNKrR5
with S.Lin (Oxford) + J.Hilton (OpenAI)
@OpenMotics keeps growing and is looking for backend developers in Ghent (Belgium), to work on #homeautomation and #IoT. #Python is used on the devices and in the cloud. https://t.co/FNEmZ3atqe
what if instead of getting the pfizer or moderna vaccine, you got a VaccineProvider in your arm instead? now you can swap out whichever vaccine you like
I call it... Dependency Injection
New fact: #CloudRun instances can now process up to 1,000 concurrent requests (default is 80). In comparison, FaaS (Cloud Functions) have a fixed concurrency of 1, making CR is the ideal option for fast scale up + service to service auth/communication👏
https://t.co/2AvZ14iIAn
A testing facility in Holland had their frontend connect directly to Firestore. Not only could you scan the entire DB that way, you could register a (fake) negative test using that connection ... 😳😬🙄
(see screenshots/video on thread)
Had some people ask about my recommendations for improving their code review skills after my tweet this morning.
So this is my attempt at writing them up 🤓
https://t.co/p6frj70h0Q
Where are you stashing all your app artifacts?
Today, @googlecloud extended our Artifact Registry (which already supported Docker images) to support Java, Node, and Python packages.
One managed service for containers and packages. Unique, and useful.
https://t.co/p7NXpYmeTo
🎟 I just bought my @DjangoConEurope ticket and I hope to get to listen in to half of each day for the next three days.
⏲ There is still time to pick up a ticket and tune in, but it starts in {hours} from now.