Tuesday, June 11, 2024

Situational Awareness

The AI world is buzzing with the predictions of Leopold Aschenbrenner. It has some overlap with the AI doomsayers who say AGI superintelligence will take over the world, except that he has inside knowledge from OpenAI, and he has specific arguments how it is all going to happen in the next five years.

Scott Aaronson raves about the analysis.

Interestingly, Aschenbrenner sees his fellow AI researchers as being among the first to be put out of work by the new superintelligence. There is a global arms race to create the best AI, and by 2030, the winner will be doing its own research to improve itself. No humans needed, except to keep mounting Nvidia computers into server farms.

Monday, June 10, 2024

Carroll tries to defend Many-Worlds again

Physicis Sean M. Carroll tries to answer this question:
Joel remok says if we can't access any of the many worlds, what 2:31:07 is the purpose of studying it, when there are still many discoveries to be investigated in the reality we exist in?
Good question. There is no point in talking about parallel worlds that we can never see or know anthing about.

The transcript of his answer:

2:31:13 look I've said before I don't care about the other worlds I care about the laws of physics the question is we have 2:31:20 problems with quantum mechanics as it is taught it is not a sensible rigorous coherent Theory. we make things up like 2:31:28 observations and wave function collapse that aren't rigorously defined. many worlds is a well-defined theory that 2:31:37 replaces the ill defined Copenhagen interpretation and lets us actually do science with it. the point of many worlds 2:31:44 is not the other worlds. it's that we've answered the question of what is a measurement, why do you get probability, 2:31:50 and things like that, and equally importantly there's the fact that we don't know the fundamental laws of 2:31:56 physics. we're not done with physics yet we're trying to build better laws of physics and I strongly think that taking 2:32:02 quantum mechanics seriously and thinking about what is the correct foundational version of quantum mechanics will be 2:32:09 useful to that program. so again it's not about the worlds. people who really 2:32:15 obsess about the other worlds are the ones who haven't really internalized many world worlds many worlds is just 2:32:22 it's always obeying the Sher equation that's the essence of the theory
No, this is nonsense on every level.

Many-worlds does not answer the question of what is a measurement. It says that a measurement splits the worlds, but there is no explanation of how the worlds split, how many worlds there are, how you knowt he worlds have split, or anything. It is no better than how bohr explained it.

Many-words says nothing about how we get probabilities. The many-worlds folks do not even believe in probabilities.

Saying that Copenhagen is ill-defined to do science is backwards. Copenhagen has had successes worth trillions of dollars. No one has ever done any science with many-worlds.

I don't know how Carroll says this junk with a straight face. It is like saying Astrology is the only true science.

Here is the distinction. Copenhagen says you can predict an experiment, with probabilities for different outcomes. Once you observet the outcome, you discard the other possibilites as events that did not happen.

Many-worlds is just like that, except that there are no probabilities for the outcomes, because they all happen, and no world is any more likely than any other. Once you see an outcome, the parallel worlds with the other outcomes go out of reach. A prediction might seem true or false, but that is only because you are trapped on one of the worlds. It is not possible to do any science, because no prediction can ever be falsified.

Copenhagen makes sense and is scientific. Many-worlds is an unscientific fantasy.

Saturday, June 8, 2024

AI Models will maintain String Research

Peter Woit reports that the top string theorists are retiring, and they are worried that the younger generation will not carry on as before.

Not to worry! One suggests that AI large language models like ChatGPT will come to the rescue, because they can

train an LLM with the very best papers written by the founding members, so that it can continue to set the trend of the community.
Okay, maybe this was a joke, but it is likely to happen anyway. We may have reached peak knowledge. Already journals are being tricked into publishing AI-generated articles. The LLMs of the future will be trained on the LLMs of the past.

There is no experimental data that has any bearing on string research. Ed Witten cannot live forever. Advanced AI LLMs may be able to carry on string papers indefinitely.

Wednesday, June 5, 2024

SciAm gets Trump Derangement Syndrome

Scientific American used to be the worlds greatest popular science magazine, but now it is filled with politics and pseudoscience.

Now it has this article from an Australian Politics professor:

Donald Trump is now leading in many polls and could retake the U.S. presidency, despite numerous scandals, indictments and erratic behavior. Explanations for his popularity focus on factors like white identity, right-wing authoritarianism, nationalism and populism.

U.S. politics has included people with these predilections for many decades, however, so the puzzle of Trump’s appeal remains. ...

We differentiated the most loyal Trump supporters from the rest of the American population, including those who merely voted for Trump and supported his policies, by measuring three aspects of Trump support: unquestioning credulity, exemplified by the belief that he was in fact the legitimate winner of the 2020 election; ...

Most strikingly, we find they are highly conscientious, a measure of self-reported characteristics including carefulness, dependability, orderliness and self-discipline. Conscientiousness is significantly associated with all the three cultlike aspects of Trump support.

Support for Pres. Trump is not so tricky to explain. Pres. Biden is one of the worst Presidents ever. He is responsible for foreign policy disasters in Afghanistan, Ukraine, and Gaza. His reckless spending has caused high inflation. He has allowed the USA to be invaded by millions of foreigners. He is too senile to do the job. His VP is even worse. He has pursued bogus prosecutions of his political enemics. He appoints incompetent leftist minorities. He supports the sexual mutilation of children.

Pres. Trump, on the other hand, presided over four years of peace and prosperity. He would have been re-elected, but for changed voting rules that allowed the collection of million of votes whose casting and counting could not be verified. Trump won a majority of the votes cast and counted on Election Day.

This professor even found that most Trump supporters had conscientiously examined the evidence, and found Trump to be the better man. But this article makes no mention of all the objective reasons for concluding that Trump was a better President than Biden. He blames authoritarianism, while Biden is much more authoritarian then Biden.

As an example, those who resisted covid vaccine mandates nearly all support Trump over Biden. This shows that Biden is the authoritarian, and his supporters like his authoritarianism. Also, only an auhoritarian would prosecute his political enemies.

The Scientific American of a few years ago would never have published such a trashy partisan and illogical article.