Content moderation and #RecommenderSystems are the most widely deployed #MachineLearning based action taking systems. By action taking, I mean an algorithm that suggests a particular decision that impacts human actions. Using this view of recommender systems, we can consider how to increase the #diversity of online content by re-ranking recommended items to encourage creation of diverse content in the long term.
The basic rationale is to use random split-half data to identify what's "true" versus sampling error. Scores are based on similarities between eigenvectors or cluster centres, rather than, e.g., the shape of the eigenvalue plot.
@aral I want to believe "AI" is probably just some Fred Wiggleton in his basement somewhere in Ohio, with a can of beer (or two) beside his keyboard, having the best time of his life.
Wait, if this scales up would this let us run these things on, say, our phones? If it replaces the transformer, could it replace other uses of the transformer, like Whisper? If it can do that and deliver equivalent or better results for those kinds of speed-ups, ML applications on edge computing will explode, again. And given the accessibility benefits of some existing tech, imagine being able to run that on something other than the cloud!
This new technology could blow away GPT-4 and everything like it https://www.zdnet.com/article/this-new-technology-could-blow-away-gpt-4-and-everything-like-it/#AI#MachineLearning
"Tech companies have grown secretive about what they feed the AI. So The Washington Post set out to analyze one of these data sets to fully reveal the types of proprietary, personal, and often offensive websites that go into an AI’s training data."
Such systems carry additional burdens that are foreign to more consumer/business-level #MachineLearning systems - in particular, the need to exhaustively quantify "the unseen" through objective analysis.
It is something that, most notably, #Tesla fails to recognize with respect to their #FSDBeta program, likely by design.
The term "safety" is tossed around quite a bit by Musk, by Tesla and by these untrained human drivers.
But a complete assessment of safety is not only what is seen (the #FSDBeta-active vehicle did not appear to collide with anything), but of the potential, future unhandled failure modes that are unseen.
That is the only pathway of progress, the only pathway towards a continuously safe system.
Anything else, like what is happening with the FSD Beta program, is just goofing around.
If you were going to hand someone a single resource to get a basic, workable knowledge of #MachineLearning, neural networks, deep learning, #AI, etc. what would it be? Not looking to prime someone for a doctoral class but enough for the person to be able to talk about the concepts and play with some pre-built libraries and tools. Could be a book, a class, a video, anything.
I recently moved from mastodon.social to hachyderm.io. I'm a Computer Scientist who moves between software engineering and research. I may post about code, papers, and conferences.
The attached image helps my colleagues differentiate my cats. 🙂 🐈⬛ 🐈