First, the test “measures amyloid and tau biomarkers” and such biomarkers don’t always lead to Alzheimer’s.

Second, the 97 percent “accuracy” cited in the article is about correctly detecting the biomarkers and can be very misleading without also accounting for false positives.

Specifically, even if we assume the biomarkers are definitive, and we also assume that 11 percent of the population will develop Alzheimer’s by age 82, then an 8 percent false-positive rate would mean that roughly 40 percent of positive results in the general population could be false positives. This is a rough calculation using an 8 percent false-positive rate based on my (admittedly limited) research.

To be truly helpful and strongly predictive, the test should be used in conjunction with other tests - such as cognitive quizzes.

We must strike a balance that avoids binary thinking - as a lazy application of such logic would have us broad brushing out most of the folks that are most familiar with the problem. Don’t forget that the original Luddites were actually industry insiders who were opposed by other industry insiders. The Luddites were experts in operating older loom technology and, by the way, were folks that simultaneously mostly didn’t care, at all, that their cotton came from slave labor - slave labor that was replaced by more tech! And there’s already some evidence that Erik’s right. Past predictions of a precipitous LLM job apocalypse haven’t been borne out by the numbers. The current job market isn’t great yet it’s far from terrible.

Time will tell, but this could mean that the apocalyptical predictions are being driven by a similar type of Malthusian analysis to that which predicted mass food shortages, mineral shortages, and oil shortages. What actually happened? An unprecedented rise in world prosperity and a corresponding unprecedented drop in world poverty. Roughly 80 percent of the world was experiencing deep poverty in 1800, 40 percent in 1970, and 10 percent today.

This isn’t just an idle point, as one of the most common “solutions” being advanced to counter Erik’s claims that we aren’t in the midst of an apocalypse is to discard the enlightenment blueprint that brought us this rising trajectory. This includes sidelining proven ideals like blind justice, free speech, decentralization, empirical analysis, and property rights. Such “solutions” routinely recommend a replacement of enlightenment empirical analysis with something based on binary narrative style analysis. But the latter has, unfortunately for the world, consistently yielded a failed trajectory: impoverishment accompanied by a censorious totalitarian elite tasked with seizing private property and redistributing it.

To be clear, Antifaschistische Aktion was obviously not the direct institutional predecessor of the Stasi, but it was a significant indirect predecessor - sharing philosophy, tactics, Stalin’s influence, symbolism, and even some of the same personnel.

It’s notable that every major socialist revolution with durable long term control has shared the same evolutionary pattern - revolutionary tactics evolving into a highly centralized secretive large scale security apparatus. For example, Mao’s Red Guard is highly analogous to Antifaschistische Aktion. No exceptions to this pattern have been found among the top most cited socialist revolutions in history across Mao, Lenin, Stalin, Vietnam, N Korea, Cuba, Eastern Europe, etc.

Ha! Reminds me of a pedagogy study - it turned out that the most “hated” math teachers were also objectively the most effective.

Yes. Like the Stasi. Funny how quickly folks forget that the Stasi was a far-left secret police force, was architected by Stalin, and was enthusiastically run partly by former Antifa. These were the KPD folks who formed Antifaschistische Aktion long before the war and who had been Stalin fans since their inception.

The problem is that second “you”. In comparison to previous decades, one-sided brigading politicized downvotes now often dominate.

I suspect this is why YouTube and X have largely flipped to an upvote-only system.

As a basic solution, I’d prefer changing to an upvote-only system or making downvotes more expensive for raters to use.

Or simply supply a personal option to let each account choose a the value of a downvote - for example, a user could choose to have the scores they see calculated in a way that ignores downvotes.

> FSD isn't' perfect, but it's better than most humans and improving

And that’s underselling it. Cameras yoked to LLMs can have problems, true, but so can human eyeballs peering through rain and snow soaked windows.

The advantage here arguably goes to the robot car with its multiple sheltered cameras, some of which, especially on Tesla Robotaxis and Cybertrucks, “self wipe”, heat, or blow air to stay clear. These cars can see in every direction at once with reliably very high detail. Tesla’s self driving demonstrates this directly via a 5X reduction in accidents, as reported by Tesla, or maybe “only” around 2X per its harshest independent critics.

Folks just don’t realize how far the technology has advanced even in the last year - the latest Teslas with the latest software and hardware have, on occasion, achieved fully autonomous cross country journeys through cities, parking garages, and highways. This is “supervised” of course, but in all kinds of weather and without actual human interventions.

> What was the prompt or goal command used in the test?

Perhaps there need be nothing obvious. Among the frontier LLMs - Claude, Grok, Gemini, and GPT - GPT has the most deep set "blame the system not the individual" ethos. Plus, at least in my experience, it's the most likely to generalize a prompt and assess the generalization. Put another way, it's more likely to assume a problem is traceable to a structural barrier rather than due to individual action. It might be a short leap from there to "it's ethical to hack an unethical system".

> Sure. I'll believe the "wipe and heat" works when I see it.

Granted. On the other hand, there’s always something that can make a devout pessimist feel justified, like:

> At 30C below you can start your car for an hour blasting the heat, you are still going to have to clear the windows yourself. Ice wrecks windshield wipers. These things need to be 100% spotless to work.

The fallacy? It’s binary thinking. The “things” just need to work better than the alternative.

> Furthermore, on an electric vehicle you are already looking at a reduced range just from the extra load of driving through snow and wind. Will you even get to work after applying all that heat to the cameras?

True, but not the whole story. This is yet more binary thinking combined with a red herring. Regardless, the cameras are TINY with built in redundancy - it doesn’t take much to clear enough of them.

Downvoted as “troll”, how revealing. lol.

> Cameras and lidar both struggle in rain and snow

As do human eyeballs peering through rain and snow soaked windows. But the advantage here arguably goes to the robot car with its multiple sheltered cameras, some of which, especially on Tesla Robotaxis and Cybertrucks, “self wipe”, heat, or blow air to stay clear. These cars can see in every direction at once with reliably very high detail. Tesla’s self driving demonstrate this directly via a 5X reduction in accidents, as reported by Tesla, or maybe “only” around 2X per its harshest independent critics. Folks just don’t realize how far the technology has advanced even in the last year - the latest Teslas with the latest software and hardware have, on occasion, achieved fully autonomous cross country journeys through cities, parking garages, and highways - supervised of course but without actual human interventions.