CURRENT EVENTS - May 1 - May 7

  • Thread starter Thread starter nycfan
  • Start date Start date
  • Replies: 345
  • Views: 5K
  • Politics 
Status
Not open for further replies.
"Trump again teases a "very big announcement.""It's not about trade, it's about something else--but it's going to be a truly earth-shattering and positive development for this country and for the people of this country," he said in Oval. "And that'll take place sometime within the next few days."

His replacement for Obamacare will finally be ready. And his long-promised tax returns will be released (but only those from 1992-98)
Infrastructure week again?
 
"Trump again teases a "very big announcement.""It's not about trade, it's about something else--but it's going to be a truly earth-shattering and positive development for this country and for the people of this country," he said in Oval. "And that'll take place sometime within the next few days."

His replacement for Obamacare will finally be ready. And his long-promised tax returns will be released (but only those from 1992-98)
Scared Kermit The Frog GIF
 
Is this real life ?

and the beat goes on alienating one country after another...

 


I have been wondering lately what will happen as the large language models continue to train on an internet which is made up of an ever-increasing proportion of AI-generated content. Can only imagine the quality will get worse and further degrade the user experience of both the AI models and the internet as a whole- accelerating the enshittification of everything.
 
I have been wondering lately what will happen as the large language models continue to train on an internet which is made up of an ever-increasing proportion of AI-generated content. Can only imagine the quality will get worse and further degrade the user experience of both the AI models and the internet as a whole- accelerating the enshittification of everything.
There was a study I saw recently showing just that -- a degradation in performance for language models trained on the outputs of language models.

As the language models get bigger, though, I think this effect will fade out. My understanding (some of which is unreliable deduction or inference on my part) is that the problem comes from the necessity of approximation. The AI neural network is essentially trying to plot incomprehensibly many points on an unfathomably large manifold of semantic space, but I think the space is bigger than the points (assuming that the points have finite size to avoid a triviality). In other words, the representation of semantic space within the network will have holes, so to speak. If the AI trains primarily on AI-generated data, then it would have a tendency to replicate the holes in the training data and hence a degradation in performance.

But the plots will eventually catch up, I think. I thought I read OpenAI is now training a 1B parameter model -- or someone is, at least. Because the number of connections generally scales as a square, that would be a huge improvement in power. I think the architectures prevent it from being a full square, because they have many levels and levels do not connect except to the adjacent levels (among other reasons), so maybe it's more like 50% of the full square. That would still be an 8x increase in its information storage.

By the way, for readers not familiar with the terminology: a manifold is like a surface, except in many dimensions. In this case, millions of dimensions I would think, maybe a few hundred thousands. It's impossible to visualize -- you have to undo the metaphor, sort of. Also, I'm not using manifold in the technical sense of being locally Euclidean everywhere. I don't know a term that competes with manifold for a non-tech audience (and indeed manifold is pretty esoteric itself).
 
IMG_6874.jpeg

Translation:

The nomination of sketchy supplement promoter Nesheiwat for this post apparently died in the Senate, but she will be offered an unspecified role at HHS anyway. Meanwhile a medical influencer focused on metabolism and “optimal cellular biology” as the secret to health is nominated instead. She seems sort of on California granola eat clean guru side of things — hopefully genuinely interested in promoting healthy behavior (though her website looks like a bajillion other diet sites).

Dr. Means’ website: Casey Means MD

Nesheiwat is a certified quack, so hopefully this is an upgrade to harmless wellness guru.
 
IMG_6874.jpeg

Translation:

The nomination of sketchy supplement promoter Nesheiwat for this post apparently died in the Senate, but she will be offered an unspecified role at HHS anyway. Meanwhile a medical influencer focused on metabolism and “optimal cellular biology” as the secret to health is nominated instead. She seems sort of on California granola eat clean guru side of things — hopefully genuinely interested in promoting healthy behavior (though her website looks like a bajillion other diet sites).

Dr. Means’ website: Casey Means MD

Nesheiwat is a certified quack, so hopefully this is an upgrade to harmless wellness guru.
Interesting



Loomer’s beef is about support of the COVID vaccine and concern about vaccine hesitancy. But it also sounds like the former Fox personality was in hot water over misleading claims about her credentials. My guess is Loomer came hard after Nesheiwat after it became evident the nomination was in doubt, but if there finally was a scheduled hearing on the nomination in the Senate, then Loomer may well have tanked it since we know the Senate has been approving virtually anyone Trump nominates, and is unlikely to schedule a hearing if they lack the votes.
 
Interesting



Loomer’s beef is about support of the COVID vaccine and concern about vaccine hesitancy. But it also sounds like the former Fox personality was in hot water over misleading claims about her credentials. My guess is Loomer came hard after Nesheiwat after it became evident the nomination was in doubt, but if there finally was a scheduled hearing on the nomination in the Senate, then Loomer may well have tanked it since we know the Senate has been approving virtually anyone Trump nominates, and is unlikely to schedule a hearing if they lack the votes.

Because Loomer, probably the key “weakness” was the candidate’s melanin levels.
 
Is this real life ?

and the beat goes on alienating one country after another...

I don't think he is, but if Trump was secretly doing Putin's bidding and was deliberately trying to undermine our economy and international standing and political system he couldn't be doing a better job of it. In fact, if he was deliberately trying to wreck and undermine the country you'd have to say that he was doing a fabulous job. However, I don't think he is willingly trying to wreck the country, I think he's just a total idiot and moron who has no fucking clue what he's doing and whose greatest life success has been to convince half the country that he's some kind of patriotic savior and genius. Although he is no doubt a Putin stooge, albeit probably a clueless one, as well as a Saudi Arabian stooge, Orban stooge, and for pretty much every other global strongman.
 
Interesting



Loomer’s beef is about support of the COVID vaccine and concern about vaccine hesitancy. But it also sounds like the former Fox personality was in hot water over misleading claims about her credentials. My guess is Loomer came hard after Nesheiwat after it became evident the nomination was in doubt, but if there finally was a scheduled hearing on the nomination in the Senate, then Loomer may well have tanked it since we know the Senate has been approving virtually anyone Trump nominates, and is unlikely to schedule a hearing if they lack the votes.



 
Status
Not open for further replies.
Back
Top