Gatsky

Do you think GPT-4 was trained and then immediately released to the public? Training finished Aug 2022. They spent the next 6 months improving it in other ways (eg human feedback). What he is saying is already evident therefore.

I guess the analogy would be if every year the same people keep finishing in the top quartile in your coin league, at the same time as failing to make the expected losses dictated by chance.

Agree, ~7% seems ok? The funny thing about the profession of active fund management is that there is a ‘scoreboard’. Almost no other profession can or would score itself like this. If they did, I doubt more than 7% would beat a suitable benchmark for hairdressers, plumbers, doctors etc.

This ‘score’ has other useful properties. A successful fund manager has to operate under difficult and changeable conditions for long periods of time. They can’t get there just by default or by being lucky. Knowing which people can do this very challenging task means the advice they give about investing, managing your own psychology and life in general is imbued with a legitimacy that is difficult to find elsewhere.

There is that story about KLF actually setting fire to a million pounds [1].

Why don’t the ultra rich give away more money? Why do they spend it on super yachts, houses with 50 bathrooms, and horridly expensive luxury trinkets? Even more to the point, why do they just hang on to vast sums of money they don’t need in bank accounts or other investments? My conclusion is that there isn’t a reason, in the same way that there wasn’t really a reason KLF burned a million quid. Counterfactualising the lives of the rich and powerful is a profound waste of time.

[1] https://en.m.wikipedia.org/wiki/K_Foundation_Burn_a_Million_...

I think schizophrenia could be prevented… most people manifest psychosis in teenage or young adult years. There is at least a good decade before it develops where something could be done, maybe. Just worth thinking about, at least.

The evidence isn't that clear as I read it [1]. But who cares really, playing music is great.

[1] https://www.psychologytoday.com/au/blog/the-athletes-way/201...

Looking at the paper here [1], it appears there was stability, then a rise up to 2011 followed by a global drop in 2012/2013, and then more stability (Fig 1). It is implausible that things could change so quickly for everyone. A spurious effect related to methodology is surely the most likely answer? I haven't read the text in detail. Fitting a linear relationship to the above as they do in the paper seems a bit crazy to me, but I'm not a psychologist.

[1] https://www.sciencedirect.com/science/article/pii/S016028962...

This is reversing the causality I think.

The word ‘chorea’ is derived from the greek word for ‘dance’ as it happens.

Seems so, although it says here [1] you can just email them for a link.

[1] https://singlelogin.me/blog/39

Thanks, yes this adds even more data.

As far as I know, there is no proven way to improve equality in a school system. Even in Norway, which has the highest GDP per capita in the world, the students from poor/less educated families still underperform [1]:

>Norway is a wealthy and egalitarian country with a homogeneous educational system, yet achievement gaps between students at the 90th and 10th percentiles of parental income and between students whose parents have at least a master and at most a high school degree are found to be large (0.55–0.93 and 0.70–0.99 SD), equivalent to about 2 to 2.5 years of schooling, and increasing by grade level. Achievement gaps by parental income, but not by parental education, increased over the time period, underscoring the different ways these two socioeconomic status components relate to achievement and the potential for policy to alter gaps.

Part of the answer could be that genetics play a larger role than anyone would like. Plomin for example claims that from twin studies a large part of educational attainment is heritable. I suspect the current school system will burn itself down before admitting that, however.

[1] https://journals.sagepub.com/doi/full/10.3102/0013189X221142...

I would have disagreed with you 5 years ago but actually I think you could be right. There are adjacent examples of transcendent human experiences being synthesized - eg people falling in love with that Replika chatbot.

I feel grateful that I heard the Mahler symphonies before the great flood of AI content, at least I will always have that. But perhaps AI could complete his 10th symphony…

The point is about which companies are worth investing in, not which companies benefit the world the most. Take biotech, where there is a lot of 'innovation' but you will generally lose your shirt investing in such companies.

The other point to make is that innovation tends to either be not obvious or too obvious. Innovation at Tesla is 'too obvious' and hence the company has very high valuations and a low margin of safety for an investor. Amazon on the other hand had a lot of innovation in their business model but this wasn't very obvious to an outsider.

Given the rapidly ageing Italian demographics, this is particularly crazy. Italy needs to make some changes or risk being a defunct nation state before the middle of the century.

A class action lawsuit with only 4 students? Ok.

Universities charge more and guarantee nothing in terms of job placement, but nobody sues them. Just goes to show how much people will happily pay for a status symbol.

Well that leaves aspartame and stevia as the cleanest sweeteners, at least using human data. Thaumatin is another one that should be benign (it’s a large protein, unlike the others), but largely untested.

The ‘last mile’ in healthcare isn’t really amenable to any technological cost reduction, especially for the most serious (and costly) illnesses. Any practising healthcare professional can tell you that. We are talking about actually seeing patients, trying to deal with their numerous problems across physical, mental and social spheres, supporting and advocating for them and their carers, tailoring treatment and investigation to their needs/preferences/beliefs, and participating in very difficult decisions about limiting treatment when that is appropriate. There is nothing to ‘optimize’ here. Good healthcare requires someone to take responsibility for the wellbeing of a patient.

I maintain that the best analogy for life sciences research is naval exploration. Some people tell tales of their great voyages but never set foot on a boat. Some sailed, yet found nothing, and still spin a story to avoid embarrassment. And some actually found America. Science is obsessed with these stories enshrined in journal articles, which are elevated to the status of artefacts of knowledge, complete with Excel spreadsheets in the supplementary info. Creating such artefacts determines everything in your scientific career. That is the actual problem in my opinion. Of course people will want to fake these powerful status symbols.

I do enjoy it, but you can see how it compares poorly to other alternatives for the very youthful. Particularly for women. But this is not the point - it isn't even a conscious decision for the most part to have kids late. Modern life is just set up that way. None of my peers were having their first kid at 20. None of my peers were even in a serious relationship at 20. Most weren't in any sort of relationship.

Predictably, you are being downvoted.

Fact is, some people can just work harder than others. This is obvious is it not? It is not because some people are lazy and others aren't. It is not a moral issue. Some people just have higher energy levels in my experience (I'm not one of them, as a type this at 8pm nearly comatose). They just have more capacity. I think it is genetic.

It’s mainly driven by gender equality I would have thought. The more educated and autonomous women are, the less they tend to have kids. The most fertile years are spent (quite rightly) enjoying life, getting an education and engaging in productive economic activity. It is the clash of culture with biological reality.

It’s in the bottom 15 problem index horizontal bar chart.

Well, there is this chap on twitter who makes amazing charts to do with equities and macroeconomics (eg https://twitter.com/TimmerFidelity/status/161865104862910873...). Turns out he just uses Excel. Maybe that's not the answer you are looking for.

Reading the paper [1], the genetically modified stem cells overexpress the ARSA enzyme. This seems to create a ‘bystander effect’ which increases the total body enzymatic activity above some critical threshold sufficient to arrest the disease process. Myeloid cells associated with neuronal structures are probably critical to this benefit. A standard bone marrow transplant doesn’t reach this threshold, and has demonstrated poor efficacy in early onset disease.

> The dataset was randomly divided into 70% training subset (1420 for prostate and 692 for pancreatic), 15% validation subset (304 and 148), and 15% test subset (304 and 148). The batch training task was performed based on conjugate gradient model for 100 epochs with a learning rate of 1e−5. The RNN deep learning model trained for discrimination of normal and prostate cancer urines showed the best validation performance at 93 epochs (Figs. S12a–c). As shown in Fig. S12b, using the total of 304 human urine test datasets, our RNN binary classification model achieved a 0.9997 AUC from the receiver operating characteristic (ROC) curve and a sensing performance with a sensitivity of 99.4%, a specificity of 100%, and an accuracy of 99.7% (confusion matrix). In the same manner, the trained deep learning for pancreatic cancer urine dataset (Figs. S13a–c) clearly showed superior classification performance with a sensitivity of 98.6% and a specificity of 100% (99.3% accuracy, 0.9892 AUC, 59 epochs). The false discovery rate (FDR) of RNN-assisted binary classification was 0%, 0.7%, and 1.3% for normal, prostate cancer, and pancreatic cancer, respectively. In CNN-assisted three-class classification, the dataset was randomly divided into 70% training subset (1,557) and 30% test subset (666). The batch training task was performed based on conjugate gradient model for 100 epochs with single conventional layer. To find more general diagnostic model for unknown urine sample analysis, multi-class classification method was applied on normal, prostate and pancreatic cancer urine based on CNN deep leaning model. The CNN deep learning model trained for normal and cancer type classification showed excellent performance with <1% loss and 100% accuracy (Fig. S14) after 33 epochs. Our CNN multi-class classification model achieved a sensing performance with a sensitivity of 96.8%, a specificity of 99.6% (prostate) and 97.7% (pancreatic), and an accuracy of 96.8% (confusion matrix) from test dataset. The FDR of CNN-assisted multi-class classification was 4.4%, 1.8%, and 3.6% for normal, prostate cancer, and pancreatic cancer, respectively. Therefore, our 3D-PCNUS platform supported by a fast deep learning bioclassification method successfully distinguishes cancer urine from normal urine with high clinical sensitivity and specificity, suggesting the use of the developed system for on-site cancer diagnostic technology. The developed system is suitable for patient screening with higher pathogenic probability.

> However, as I spell out in my book, the concept of AGI is inseparable from the kind of hierarchy of intelligence that has underpinned ideas of innate supremacy since the days of empire and colonialism.

I think this line discredits the entire article, which was already a few obvious points blown at gale force into the reader’s ear.

Good to hear.

IMHO the recent clinical trial results of psilocybin for depression were disappointing [1]. A single dose didn’t seem to have a lasting effect.

[1] https://www.nejm.org/doi/full/10.1056/NEJMoa2206443

This is indeed big news, and an important proof of principle. I suspect in the long term these cell therapies may have their most fruitful applications in non-malignant disease.

This is not identical to the SJC list.