Profile: ec55841e...
For those people who consider #AI a major helper when #programming, I don't know ...
Today's experience, caused by me being too lazy to look up the APIs myself:
Me: How do I do X?
Chatgpt: yadda yadda max_redirect(0).
Me: It doesn't have a max_redirect().
Chatgpt: Sorry, yes, it doesn't have it, use requests_redirectable() instead.
Me: It doesn't have a requests_redirectable() either.
Chatgpt: Sorry, you are right, it doesn't have that either.
Should I go on?
Sorry for the link to the hell site, but …
I have never seen a video about a serious presidential candidate that was even remotely like this.
I wonder whether the cyber warriors in armed forced around the world sit on a bunch of Crowdstrike-like “opportunities” and in case of war, waltz into Crowdstrike-like companies and order them to deploy the bad file. For them, that’s sort of the ideal scenario, no?
In all the hoopla about #AI you’d sort of think people would be debating strong vs weak AI vs the limits of “thinking” by any kind of machine as it has been debated for centuries. Like the Chinese Room thought experiment. https://en.wikipedia.org/wiki/Chinese_room
But I see nothing of that. Why is that? Do we simply now assume these issues have been resolved in favor of unlimited machine intelligence and all the doubters have been convinced?
Getting a laptop with 32G RAM was a really good idea. No memory pressure ever, in spite of running VMs, sometimes more than one.
Been a long time since my first computer that had 64kB, and where I recall I had to implement a custom swap-the-middle-portion-of-the-program-from-disk for something I wrote. In BASIC.
Have there been any protocol-level #attacks against the #Fediverse so far? Malformed ActivityPub messages, subtly crafted HTTP signatures, that kind of thing? (If not, there sure will be!!)
And more interestingly: if you were on team red, what would you do? 🙂 (On second thought, maybe we should have that conversation outside of the public eye, so DM.)
Is there something we ought to do with feditest.org to test implementations for resistance against attacks? #testing
Somebody recently used the term "permissionless" to describe the innovation that is possible in the open social network called the #Fediverse. (I forgot who, sadly.)
That term might be borrowed from the crypto crowd, but it appears applicable.
Any thoughts on adopting this term for the Fediverse, too? I'd like to, I think it's apt.
Pondering the today's "AI" and this famous saying:
"If after ten minutes at the poker table you do not know who the patsy is—you are the patsy."
Given the amounts of money involved, #AI looks like the biggest poker game of all times.
nostr:npub1melv683fw6n2mvhl5h6dhqd8mqfv3wmxnz4qph83ua4dk4006ezsrt5c24 So why aren't more people talking about it? Baffles me ...
I participated in a geek meetup earlier this week where people jointly try to learn about a new technical subject every couple of weeks.
This week: ActivityPub.
It's always fascinating what kind of information about this topic people came across, what kinds of questions that produces, and misconceptions. I think I was able to address some of those.
And every time the refrain: this could be big! Why aren't more people talking about this??
I don't actually have an answer for that one.
There is bullshit.
Then there is really bad bullshit.
And then there is #AI bullshit.
Apropos nothing in particular.
#AI as a topic is a marketer’s dream. Great promise! Great peril! Ban it! No, invest more in it! Everybody will be fired! No, we will all be gods! No, the new gods will exterminate us! The Americans / Chinese / … / but not us will use this to take over the world!
Everybody can have an opinion. Headline topic in every governmental meeting and board room and the pub at the corner. Even the pope gets quoted. With almost nobody having any idea what they are actually talking about.
If I type "Tell me the history of the Roman empire" into an #AI chat bot like #ChatGPT and hit return, it HTTP POSTs that to one of their web servers, and then ... what?
How many computers are involved in producing my response? What are their responsibilities? How is the response assembled from those parts?
It's a bit like the old interview question: if I type "cnn.com into a browser, what exactly happens?" I can answer that one, but I'm unclear on the AI chat version. Anybody know?
Enshittified Instagram is now forcing you to spend at least one ad worth of time in shit. No stepping around it allowed!
https://www.theverge.com/2024/6/3/24170682/instagram-unskippable-ads-test
nostr:npub1melv683fw6n2mvhl5h6dhqd8mqfv3wmxnz4qph83ua4dk4006ezsrt5c24
I understand the architectural differences :-)
I also believe that using the same account with multiple servers, and changing identifiers and "home base" for when interacting with the fediverse are very solvable problems. With the caveat that it hasn't really been broadly implemented.
nostr:npub1melv683fw6n2mvhl5h6dhqd8mqfv3wmxnz4qph83ua4dk4006ezsrt5c24 The video is a bit aspirational in that respect, but people are working on it. No reason it can't be made to work.
Nice video explanation of the fediverse.
Nothing is as practical as a good theory.
I don't know who said it first, but they were right.
Over on #bluesky, somebody pointed me to this very fascinating thread on this topic: https://bsky.app/profile/rahaeli.bsky.social/post/3kp7lfkssxl2q
Imagine two identical #3social #networks with mainstream users. Same number of users, same connectedness of the social graph, same demographics etc.
Social network A only has an algorithmic feed. Social network B only has a chronological people-I-follow feed.
What would the usage numbers look like in comparison? Numbers of posts, numbers of likes/replies/boosts, growth trajectory, etc?
I would expect somebody has done some research on this. Anybody know of any public results?