tangeli

Our mission is to ensure that artificial general intelligence benefits all of humanity.

They have a different understanding of ‘benefits’ than I do.

It should be made clear in the law that when LLM output is published a person or other legal entity is responsible for that publication with a duty of care to ensure the output is not harmful. At least until LLMs are recognized as legal entity in their own right. And that publishing the output of an LLM without knowing what was published and deliberately deciding to proceed with that knowledge is, at least, failing a fundamental duty of care as publisher.

The referenced article is not new. It was published in 2022.

I’m sceptical that it is the first-ever recording of a dying human brain. I recall significant interest in and controversy regarding changes in EEG recordings around the time of death back in the 1980s and EEG was in use and the issue was controversial for decades before that.

Time to loss of brain function and activity during circulatory arrest reports findings of a literature review published in 2016 and many of the reviewed publications were published much earlier than that.

Guideline 6: Minimum Technical Standards for EEG Recording in Suspected Cerebral Death records some details of early standards, practices and literature going back at least to the 1960s. It is focused on use of EEG to confirm death as opposed to EEG recording ‘of a dying human brain’, but it is inconceivable, given the intense interest at the time, that there were no EEG recordings around the time of death prior to the recording reported in 2022.

This is not to suggest that there was nothing novel in the reported findings, but the ‘first-ever recording’ in the headline is, I think, sensationalist misrepresentation.

There are guides available. Search for ‘Linux kernel debugging’ or ‘Linux module debugging’, depending on which you are more interested in. And, of course, learn about the relevant parts of the kernel.

You might have a look at Debugging kernel and modules via gdb¶. The kernel.org site has a wealth of information.

On Debian 12 and 13 with xfce, I am using ibus and Intelligent Pinyin (ibus-libpinyin) for Chinese and English. In the past I have used fcitx5 and various other IMEs. Once they’re configured there isn’t much difference between ibus and fcitx5, for my simple use. My Chinese is rudimentary but my Chinese wife is happy with the configuration. I switch input methods with a configurable keyboard shortcut (Ctrl-Space is my preference) or menu on the ibus item in the Status Tray Plugin of the xfce panel. Changing is easy.

I have task-chinese-s-desktop and task-chinese-t-desktop installed. These bring in fcitx5 and various fonts, which suggests that whoever created these tasks think fcitx5 is better than ibus. And I installed ibus-libpinyin, which brings in ibus. I don’t recall why now - it was long ago. So I have ibus and fcitx5 installed but have been using only ibus for the past few years. It works well enough that I haven’t revisited it. If I were installing again now, I might choose fcitx5 instead of ibus.

I see there are also task-japanese-desktop and task-korean-desktop, which you might find helpful.

I haven’t used 13/trixie but with 12/bookworm I use jmtpfs but it’s also available for trixie.