An Alzheimer's Bet Based on Biology

Alzheimer's research has been accused of being hijacked by a misleading hypothesis that only survives to enrich drug companies.  The amyloid hypothesis has been extensively critiqued, including by this essay on the blog Astral Codex Ten that looked at problems with the original paper that started (by some accounts) 30 years of misguided research.  

In response to a call for counter-arguments, David Schneider-Joseph, presented a compelling argument in favor of the amyloid hypothesis.  He even went so far as to propose a bet; that a drug targeting amyloid would achieve at least a 75% slowdown in Alzheimer's in the next 12 years.

I wouldn’t take David’s proposed bet, because I’m sure some company will find a way to gerrymander clinical endpoints to get to 75% “slowdown”.  Probably with a drug that has horrible side effects and costs north of $100,000/year.  

I would bet on the following:  the leading* therapy in 12 years will not be one whose sole intended mechanism involves amyloid production or clearance, e.g. monoclonal antibodies, small molecules, or gene editing that work directly on amyloid.  

    * “Leading” means some combination of efficacious and most widely used.

This is the important sense in which the amyloid hypothesis is wrong: it may be “correct” in some narrow biochemical sense, but it is a dead end.

In a way this is an easy bet on the status quo continuing indefinitely, because the current drugs used to treat Alzheimer's do not target amyloid.  However, I’m also bullish on emerging therapies such as creatine, lithium, calcium channels, and mitochondrial therapies including far infrared, ketones, and other ways to generally affect metabolism and autophagy/mitophagy.  Some of these might be shown to improve clearance of amyloid, but the amyloid hypothesis is not necessary or very helpful in the development of these “general" therapies.  

My prediction is based on a general model of how biology works.  Simple infectious and traumatic medical conditions seem most amenable to targeted therapies, while general conditions like mental illness, cancer, and Alzheimer’s are so complicated and depend on so many general health processes that targeted therapies tend to miss the mark.  

Ideally, we would pursue both lines of treatment, but our pharma system is biased toward patentable targeted treatments to the exclusion of all else.  The “Amyloid Mafia” is shorthand for this trend in Alzheimer’s research.  

However, even given the funding disparity between amyloid research and other treatments, non-amyloid treatments are arguably already more efficacious.  As a good Bayesian, I predict this will continue, and so this seems like a safe bet to me!

AI Developments... and Predictions

 I've been watching the AI developments... and the bloggers writing about the AI "foom".

• Scott Alexander on the feeling of being on the verge of AI Apocalypse.
• Scott Alexander on the new Janus-faced AI simulators, a weird new breed of alien intelligence we didn't expect.
• Erik Hoel on why we shouldn't be cavalier about the potential impact of super-intelligent AI.  (Item 4 of 9 on the page).

But in my opinion, the new AI chatbots may replace Google's current search, but they will stop far short of replacing anything but the most mundane human jobs. 

Its so hilarious that people are worried about artificial intelligence when we haven't even made a robot that can flip hamburgers yet!  Wake up and smell the grease:  the real world is complicated and messy.  We can't even replace the jobs that everyone, even the people currently working them, want replaced.  I don't think we're going to be replacing professionals anytime soon.

Just look at driverless cars.  Driving is a great example of something you'd think computers would be good at.... if you'd never driven a car in the real world!  Computers are great at driving cars in video games, but the real world is messy.  That's the Real World problem.

The second problem is even bigger: the messiness of the human social world!   Even if driverless cars were better than human drivers, that's not good enough.  We expect computers to be perfect.  If a company sells me a car and I wreck it, I'm responsible.  But if the company sells me a driverless car that crashes, the company is responsible.    

For every task that matters we want a human to be responsible.  Even if they solve the Real World problem with bigger and better AIs, the Responsibility problem will always prevent human replacement.  Imagine if my boss could replace me with a computer: he would then be 100% liable for any mistakes the computer made!  Is that the kind of liability any Pointy Haired Boss wants??  As long as he has real humans working for him, there's always someone to share the responsibility.  

Example: imagine something that's even simpler than driving a car: flying a passenger airplane!  I'm sure autopilot could fly an airplane as well as a human.  But would anyone want to fly on an airplane with no human pilot?  We'll always have the human pilots, even if the autopilot does more and more of the routine work.

Humans are social creatures.  The humans who think AI will rule the world must be living inside a computer world.  Its interesting that, as more and more of our world is computerized, we hear more and more from the people living inside the computer world.  But its not the real world!

Climate Prediction Skill

The US Climate Prediction Center issues forecasts beyond the normal National Weather Service's 10-14 day window.  They provide weekly and monthly forecasts out to 3 months.  Given the timeframe and the fact that their forecasts cover the entire contintental US, its not surprising that the forecasts are often wrong.  But how wrong?  And is their skill improving over time?

I analyzed their 3 month temperature and precipitation forecast skill using data provided on their "Gridded Seasonal Verifications" webpage.  

Note that skill is measured on a scale from -50 to 100, where -50 would be a forecast that was exactly wrong in every area, 0 would be a prediction that did no better than chance, and 100 is a prediction that was exactly right in every area.  

They provide data starting in 1995.  Since that time in the mid 1990's, linear trendlines show that their forecast skill has slightly improved for both Temperature and Precipitation.  Precipitation skill started out lower, but has almost doubled (from 10 to 20) while Temperature skill started higher but has not increased as much (from 22 to 28).

However, the last 10 years have not been as successful:

Since 2012, neither Precipitation nor Temperature skill have increased.  In fact, mean temperature forecast skill has decreased markedly since 2018.  Before that, Temperature skill had been doing quite well in the period 2014-2018.  It is not clear what changed in 2018.  A similar transition may be happening with Precipitation, where the period 2019-2022 saw consistently good predictions, but since the beginning of 2023 the forecast skill has fallen off a cliff.

With increased use of machine learning, it seems likely that long-term weather forecast skill should increase.  However, complex chaotic weather patterns are most impactful to climate predictions in the 1-3 month time frame, so this area of weather/climate prediction may continue to have lower than hoped for success.  

Covid and evidence for/against the Efficient Market Hypothesis

https://www.bloomberg.com/opinion/articles/2020-06-01/stocks-are-trying-to-forget-2020

Why is the stock market doing so well?

"If in 2019 you built a stock-price model that projected out 100 years of income, you will have to adjust one or two columns for 2020 and 2021, but after that everything can stay exactly the same."

The market movement this year seems to support this simplistic hypothesis and refute "short term-ism" critique of market.

https://www.lesswrong.com/posts/utySCY9nJt9xGYGGQ/the-emh-aten-t-dead

Argues that just because some people could make some money "beating" the market doesn't mean that they didn't get lucky and apply flawed post hoc reasoning.

https://www.lesswrong.com/posts/tonKatiDTzTP8LrEk/zoom-technologies-inc-vs-the-efficient-markets-hypothesis

Argues that EMH needs to be restated, as market predicts the future value of the market, which could be totally "irrational" compared to the real world.

Weather Prediction: Enough Said

February 2019

Moore's Law Redux: Connectivity is More Important that Processing Power

While Moore's law continues to slow, traffic on international internet links continues to grow faster than capacity. 

Moore's law, which predicted a doubling of the number of transistors per chip every 18 months, lasted from the 1960's through the early 2000s.  Over the last 10 years, newer chips have cost less and use less energy, but they have not been noticeably faster. Yet, even as Moore's law has slowed, the rise of the internet has made individual computer chips less important.

Chip density is still increasing, but progress has slowed as chip designers bump up against fundamental physical constraints, like the size of silicon atoms.  Also, as chips get smaller and denser the problem of dark silicon has increased.  To combat these limitations, manufacturers have added multiple cores and specialized architecture for graphics processing and other specialized operations. But multicore processors are hard for software to use efficiently. (Ars Technica article.)

So the current state-of-the-art 14nm Broadwell chips will remain best for the next 18+ months.  Intel promises to go to 10nm and then 7nm with new non-silicon technology.  Non-silicon technology may also finally allow faster chip speeds.  3D chip architecture is the next big advance that will improve power efficiency.  

In the meantime, manufacturers have focused on building specialized chips for mobile applications and the internet of things.  It seems the brave new frontiers are really in new mobile and cloud-based applications where energy efficiency matters more than processing power.  Every person in the world will have, on average, three internet-connected devices by 2019.

Global peak Internet traffic volumes rose 37 percent overall in 2015.  In years past, peak  internet traffic increased in excess of 41 percent each year, a rate which implies that traffic is more than doubling every two years. Over the next three years, overall growth is projected to slow to 23% a year (Telegeography 2015 Annual Report, Executive Summary.)

Even with the relative slowing of internet growth, global internet traffic in 2019 will be equivalent to 64 times the volume of the entire global internet in 2005. Globally, Internet traffic will reach 18 gigabytes (GB) per capita by 2019, up from 6 GB per capita in 2014.

IP traffic is growing fastest in the Middle East and Africa, followed by Asia Pacific. Traffic in the Middle East and Africa will grow at 44 percent annually between 2014 and 2019. By 2019, there will be more internet traffic in east Asia than in North America.
(Cisco 2015 analysis.) (Ars Technica has an older analysis)

On track for increased solar activity

Earth Observatory.

Doctoring is a Doomed Profession

Although medical doctors have quite a large amount of power in our economy, I believe that their bargaining power is being diminished by technological innovations and managerial restructuring. Robin Hanson's idea of negative health care would significantly raise the prestige of massage and personal training at the cost of doctors.

Jason I. Altman predicts: "By 2030 all surgical anesthesia will be administered and monitored by computers, with no need for professional medical supervision beyond the surgeon." Even sooner than that we are likely to see outsourcing of surgery, as robots controlled from the next room are replaced by robots controlled from India.

Its only a matter of time until large, centrally-controlled and ineffecient hospitals are replaced by "the Trader Joes of healthcare", a business model that goes straight from the drug companies to the client, and "cuts out the middleman" who in this case is none other than the good Doctor.

DNA screening is the wave of the future. Ever seen GATTACA? You can currently buy limited DNA tests that will soon be cheaper than a doctor's visit to tell you everything from whether your kids are likely to have Crutzfield-Jacobson disease to what your life expectancy is based on genetic markers for cardiac disease.