Faux & Chafkin on FTX a grift from the get-go; Levine is on the point of abandoning belief in “fundamental values’; we are not quite completely ignorant about why MAMLMs do what they do; Joe...
I guess ACA would have been better without Lieberman and with a public option, but my main regret is that it didn't do more to erode the employer "provision," at least for the lowest-wage workers.
Let's be clear. A few examples of this may be evident, but that is certainly not true of the majority of Tory MPs or their enablers. If Rishi Sunak loses his seat, that would be fine, but be careful what you wish for. The next contender may be even more batshit than Liz Truss who is trying to run for the position again. [She blames everyone else but herself for her period in office being shorter than the shelf life of a lettuce.]
"if it walks like a Nazi and talks like a Nazi then it might as well be treated like a Nazi"
Absolutely. It has been shocking how the media have recoiled from doing this. Is this a variant of Godwin's Law that the media wishes to avoid? We have knowledge of the flavors of Fascism and their Venn diagrams. It isn't difficult to apply the label. Conversely, the right wing[nuts] have no problem calling out anything to their left as "socialisms" with there being any real overlap with actual socialism. It has just replaced the "pinko"/"commie" epithets of a former period.
I sure hope all those independents who think immigration is important are fired up by Biden's (admittedly belated) attempt to improve processing of asylum seekers only to be thwarted by Republicans on Trump's orders.
The prison sentence for SBF would be salve for the idea that billionaires don't go to jail, except that he wasn't a billionaire anymore. Nor was Elizabeth Holmes or Madoff. What are we to learn from this? Fraud gets at worst a large fine. Stealing from wealthy people gets jail time.
People have been recommending that the Middle Kingdom move away from fixed investment towards consumption for such a long time. I can remember that recommendation all the way back to when Stephen Roach was Morgan Stanley's Chief Economist, if not earlier. That change still hasn't happened.
Idk if Heaven is naive, or the interviewd researchers are naive, or just uneducated in linguitics and language acquisition. Humans have struggled with trying to understand how we acquire our native language. {c.f. https://en.wikipedia.org/wiki/Language_acquisition). The interviewed researchers seem to be fumbling about trying to understand how the computer does it, but using the tools of math. Just because we have greater access to the LLM model that in the human cortex - Broca's and Wernicke's areas) doesn't mean that we can tease out how the computer's algorithm and data manage to simulate the same feat (if it even does).
As for LLMs learning simple addition, even we don't do anything so stupid. We have rules for basic addition and substraction, and these can be applied to multiplication and division. Is the LLM really doing more than storing lots of near examples and applying them, rather than learning how to do math functions?
As for reasoning, we have evidence that LLMs can only do the simplest "reasoning", teh mental equivalent of breaking down a complex problem into baby steps and being led to the correct answer by prompts. LLMs are still "brains in vats" with no ability to truly understand concepts. Maybe that changes with LLMs embedded in robots able to interact with the real world (or simulations of it) but that is speculation.
There is a case I believe, for plugging an newborn LLM into a humanoid robot and treating it like a baby that needs to learn the world and human interaction over time, just as we do. OK, that is too slow for the researchers (and their investors) wanting to make a $ ROI, but we might possibly end up with a better result, and even learn something useful about cognition in te hprocess.
to analyze Chinese growth rates. The key is do Chinese institutions permit a transition from catch-up growth, investment in already known technologies to application of new technologies.
If China succeeds in increasing its manufacture exports to developed countries, it will only be because a) allowed their saving investment aggregates to allow larger trade deficits or b) imports of Chinese manufactured goods displace imports of other goods and services or c) developed countries increase their exports. The first, a), would be bad, c) could be good and b) ???
But increased supply does not necessarily amount to increased amount supplied.
Saule Amarova, Jonas Algers, & Isabel Estevez: If this were text, I would be able to form and opinion about how close this industrial policy comes to mimicking taxation of net CO2 emissions.
Three legs of China's economy are broken: real estate, infrastructure, and consumption. Now they really need exports, so charm replaces wolf warrior relations. In a better future, China's manufacturing builds renewable energy alternatives for the world. But FDI in China has moved to other nations, and both China's neighbors and the West remember Xi's recent belligerence.
I assume importers raise tariffs on China, but is there a better approach that doesn't have the big dead weight loss, inflation, and re-labeling?
Will Dogulas: This was the source of Markov's work in the first place, and described in detail by Shannon in 1948, who commented that one would lead a lot more computer power to do it properly. (Probably more than we have if one stuck strictly to the original method, which would have worked directly with an insanely large database and calculated all transition probabilities accurately.)
I guess ACA would have been better without Lieberman and with a public option, but my main regret is that it didn't do more to erode the employer "provision," at least for the lowest-wage workers.
"Even Tories seem ashamed of their record"
Let's be clear. A few examples of this may be evident, but that is certainly not true of the majority of Tory MPs or their enablers. If Rishi Sunak loses his seat, that would be fine, but be careful what you wish for. The next contender may be even more batshit than Liz Truss who is trying to run for the position again. [She blames everyone else but herself for her period in office being shorter than the shelf life of a lettuce.]
"if it walks like a Nazi and talks like a Nazi then it might as well be treated like a Nazi"
Absolutely. It has been shocking how the media have recoiled from doing this. Is this a variant of Godwin's Law that the media wishes to avoid? We have knowledge of the flavors of Fascism and their Venn diagrams. It isn't difficult to apply the label. Conversely, the right wing[nuts] have no problem calling out anything to their left as "socialisms" with there being any real overlap with actual socialism. It has just replaced the "pinko"/"commie" epithets of a former period.
"pleasant fantasy that enables the self-identified Left to reject any notion of a crisis that requires solidarity with mainstream liberalism;"
Or if solidarity is too much to ask, just to stop drawing huge bull's eyes around Liberals.
Republicans Are Easily-Grifted Morons Department:
I sure hope all those independents who think immigration is important are fired up by Biden's (admittedly belated) attempt to improve processing of asylum seekers only to be thwarted by Republicans on Trump's orders.
The prison sentence for SBF would be salve for the idea that billionaires don't go to jail, except that he wasn't a billionaire anymore. Nor was Elizabeth Holmes or Madoff. What are we to learn from this? Fraud gets at worst a large fine. Stealing from wealthy people gets jail time.
Barry Eichengreen, Middle Income Trap:
People have been recommending that the Middle Kingdom move away from fixed investment towards consumption for such a long time. I can remember that recommendation all the way back to when Stephen Roach was Morgan Stanley's Chief Economist, if not earlier. That change still hasn't happened.
The unreasonable effectiveness of MAMLM
Idk if Heaven is naive, or the interviewd researchers are naive, or just uneducated in linguitics and language acquisition. Humans have struggled with trying to understand how we acquire our native language. {c.f. https://en.wikipedia.org/wiki/Language_acquisition). The interviewed researchers seem to be fumbling about trying to understand how the computer does it, but using the tools of math. Just because we have greater access to the LLM model that in the human cortex - Broca's and Wernicke's areas) doesn't mean that we can tease out how the computer's algorithm and data manage to simulate the same feat (if it even does).
As for LLMs learning simple addition, even we don't do anything so stupid. We have rules for basic addition and substraction, and these can be applied to multiplication and division. Is the LLM really doing more than storing lots of near examples and applying them, rather than learning how to do math functions?
As for reasoning, we have evidence that LLMs can only do the simplest "reasoning", teh mental equivalent of breaking down a complex problem into baby steps and being led to the correct answer by prompts. LLMs are still "brains in vats" with no ability to truly understand concepts. Maybe that changes with LLMs embedded in robots able to interact with the real world (or simulations of it) but that is speculation.
There is a case I believe, for plugging an newborn LLM into a humanoid robot and treating it like a baby that needs to learn the world and human interaction over time, just as we do. OK, that is too slow for the researchers (and their investors) wanting to make a $ ROI, but we might possibly end up with a better result, and even learn something useful about cognition in te hprocess.
Perlstein:
But can we blame reactionaries for not knowing what Socialism is when Sanders and AOC don't know either? :)
Conservatives is how eager they are to blame everyone else for their conservatism…. George Osborne… [on] the “lie” of Brexit….
Did I miss the Wednesday when a Tory PM announced negotiation of UK entry to the EU?
Eichengreen:
I like Tabarrok's exposition of the "simplified" Solow model https://marginalrevolution.com/marginalrevolution/2024/03/teaching-the-super-simple-solow-model.html
to analyze Chinese growth rates. The key is do Chinese institutions permit a transition from catch-up growth, investment in already known technologies to application of new technologies.
Edward White & Cheng Leng:
If China succeeds in increasing its manufacture exports to developed countries, it will only be because a) allowed their saving investment aggregates to allow larger trade deficits or b) imports of Chinese manufactured goods displace imports of other goods and services or c) developed countries increase their exports. The first, a), would be bad, c) could be good and b) ???
But increased supply does not necessarily amount to increased amount supplied.
Saule Amarova, Jonas Algers, & Isabel Estevez: If this were text, I would be able to form and opinion about how close this industrial policy comes to mimicking taxation of net CO2 emissions.
Three legs of China's economy are broken: real estate, infrastructure, and consumption. Now they really need exports, so charm replaces wolf warrior relations. In a better future, China's manufacturing builds renewable energy alternatives for the world. But FDI in China has moved to other nations, and both China's neighbors and the West remember Xi's recent belligerence.
I assume importers raise tariffs on China, but is there a better approach that doesn't have the big dead weight loss, inflation, and re-labeling?
I found the Odd Lots interview of David Autor quite insightful and hopeful on AI. I'm interested in your thoughts on Autor's AI opinion.
Will Dogulas: This was the source of Markov's work in the first place, and described in detail by Shannon in 1948, who commented that one would lead a lot more computer power to do it properly. (Probably more than we have if one stuck strictly to the original method, which would have worked directly with an insanely large database and calculated all transition probabilities accurately.)