Ulrich Junker@chrisoffner3d exactly. Similar to mathematics teachers who explain errors in reasoning to pupils, we will need #informationdetectives who will spot errors in the output of #transformernetworks. It will be a gigantic task to establish #soundness in the coming age of #disinformation. Alternatively, we could simply avoid using an immature and problematic technology in products. But many people don’t get the issue. May be these are those who didn’t get the teachers’ explanations?
The next big #legal dispute. If #TransformerNetworks reproduce data points used for their training, they may indeed violate #copyright. Will be interesting to see how judges will be defining the frontier between replication and origination.
Ownership of AI-Generated Code Hotly Disputed https://flip.it/tR91SU
