A minor quibble with Linux/Ubuntu is that when I update my packages, it seems like > 50% of the download volume and time are always related to libreoffice, which I have never used

I assume since libreoffice is used by some governments/schools, it gets many security updates

My favorite sign at Monday's #PauseAI protest at OpenAI, from Conrad Barski!

Thinking of writing an app that renders my computer unusable in the morning, until it has confirmed that I've reviewed my todo list and my inbox has zero messages in it

As we grow up, we develop models of the world

Your models & your friends' models are similar, molded by a shared world

When I meet tech folks in the Bay Area, our models are indeed very similar

But occasionally, models give a completely different prediction

It is startling

remember when I spent six months working on a novel WebAssembly Lisp machine in Zig with fully serializable state, first class recursive stepwise debugging, and delimited continuation control flow integrated with JavaScript async with a self-hosting DOM-based structure editor

OpenAI's safety experts keep JUMPING SHIP, citing extreme P(doom) and lack of confidence that humanity will survive AGI

CEOs like Sam Altman, saying risk is manageable, are hiding the truth.

The truth is that their own top safety experts have freaked out about P(doom) and quit!

Humanity’s astonishing ability to empathize with absolutely anything is our greatest strength and our greatest weakness.

At long last, we have built the Her from the movie 'Don't build the Her'

Admittedly I got this wrong

They didn't seem like they were aiming to make their products seem as useful as possible today

Not sure what to make of it

'the evidence preserved to us by the passage of time constitutes but a small fraction of that which must have once existed. Each important vessel must represent scores of others that have vanished'

- Art Historian Helen Gardner

it'll be great to see some of you at the protest tomorrow!

You should call your mom and make her aware of the safety benefits of open source AI models

I called mine yesterday, and she now promises to set up an artisanal fine tune of llama3 on a compute cluster

That should keep her safe from the corporate AI and rogue actor AI

What's particularly twisted about our protest on Monday is that I'll have to bring a blank sign & markers

So that I can make a new sign in real time

Protesting the gain-of-function hazards OpenAI will cheerfully announce at 10am

For almost any big project, my two most common unblocking questions are:

1. How would I do this, as intelligently as possible, with years of time and unlimited money?

2. How would I do this, as stupidly as possible, in very little time and with very little money?

ok I've seen all the new tweets...

... but what

... if there are even newer tweets

... that arrived

... while I was reading all the new tweets

???

One rather strange event of the past ten days: like ~600,000 Australians, much of my retirement savings is with a superannuation (think 401-k) firm called Unisuper

Unisuper uses Google Cloud to store their data. Well, ~10 days ago, Google Cloud *accidentally* *deleted* *their*

This will be the presentation where they show us all the cool things we could have if we just give the AI access to our bank/photos/emails/accounts/etc

Hopefully the worm didn't get the part of the brain that a person needs to pardon a turkey on thanksgiving