There is a big distance between a set of basic methods (backprop, ConvNet, transformer, LLM…) and a *practical* system that can be deployed to assist soldiers. The former can be published and open sourced. The latter not so much.
ETHICS
-
Every Take Is Biased: Choosing Diverse Perspectives
By
–
Every take is biased.
The trick is to have a choice of takes and to not deliberately ignore what your own bias would consider bad takes. -
Google Exaggerates AI Code Generation Claims Says Leak
By
–
It seems as if Google is trying to fan the hype a little. Recently, the news circulated that Google would now have around 25% of its code written by AI. If the leak below is correct, Google would probably have exaggerated. It seems as if Google is trying to get out of the bad
-
Uncensored AI Models: Weighing Benefits Against Risks
By
–
I've often wondered whether we shouldn't have the uncensored models, so to speak. But I'm not sure. Of course, I understand the problems associated with uncensored models. People worry about creating diseases for which there is no cure. There is concern that crimes can either
-
Open Source AI Benefits Outweigh Geopolitical Risks
By
–
We are on the same side of almost every issue, Vinod.
I do think the benefits of open source/open weight AI platforms far outweigh the risks related to China and other geopolitical rivals having access to it as well.
AI foundation models are not some sort of doomsday device.
They -
Automation Will Eliminate Professional Jobs, Contrary to Expectations
By
–
Once again, we must prepare ourselves for the fact that many professions will disappear in the future. In a way, it is an irony of history that it was always believed that physical jobs would be the first to be eliminated by automation so that we could devote ourselves to the
-
Unable to distinguish between real and AI content anymore
By
–
Here we go again: i cannot figure out of real or AI. Crazy times ahead.. https://
x.com/i/status/18523
30998219829508/video/1
… -
China Develops Military AI Using Meta’s Llama Model
By
–
China develops AI for military use, violating Meta's policies „Chinese scientists announced the development of an AI model for military purposes, utilizing Meta's publicly available Llama model.“
-
CLEAR: First Benchmark for Machine Unlearning
By
–
> CLEAR: first benchmark to make models forget what we want them to forget! With privacy concerns rising, we sometimes need our models to "forget" specific information – like a person's data – while keeping everything else intact. Researchers just released CLEAR, the first