The Third Wave (covid)

 So in my previous post on the subject, I noted that the small third waves in the various states were not large or coordinated enough to make another big bump in the graph. They have now begun to do so, but there are a couple of factors in play that make it seem perhaps more significant than it really is. 

The first one is fairly straightforward: the expected number of deaths goes up this time of year every year. To see the second one, let's look a bit more closely at the CDC mortality statistics:

(This is from the same CDC page I've been using all along.) So far I have been looking only at all-cause mortality, compared to average, since that finesses a lot of issues about whether covid was a major factor in any given death. 

But a closer look shows an interesting phenomenon I hadn't noticed before. Since April the non-covid deaths ran well above normal, contributing to both peaks, if you accept the CDC's classification. But since the September minimum (and using the same classifications), non-covid mortality dropped back into the normal range. 

As a result, while covid deaths are increasing (and that's all you will see, out of context, in the alarmist press), total all-cause mortality is still basically tracking average, seasonally increasing, rates.

And it still looks like that if nothing fundamental changes, your chance of dying next year is 1.1% instead of the normal 1%. Of course, if all goes well, you will be able to drop that back down to 1% by taking a vaccine. 

Ammonia, the fuel of the future: a refrain

 A couple of years back I did a post here about a post I had done at Foresight about a decade or so back.

Ammonia (NH3)

The basic idea is that ammonia is a close to ideal fuel for a near-nanotech chemical technology, in that it carries hydrogen which can be oxidized in a fuel cell (producing nothing but water), and the rest is merely nitrogen, which is already the majority component of air.

One of the key reasons for revisiting something like this is to check up on your prowess as a technological forecaster. My recent post here was to point out some recent advance that seemed to bring us closer to the ability to use ammonia as envisioned. So when I saw another reference in the technical literature, I thought to look around and see how far the efforts had got. 

Imagine my surprise when I discovered a substantial literature, conferences, and companies at the point of viable commercialization. Rather than talk about that myself, I will refer you to the Ammonia Energy Association's review of ammonia for fuel cells. Rummage around the site for much more information.

A little brief background: the hydrogen from the ammonia molecule is what fuels the cell and generates electricity. In the most common standard type of fuel cell, the PEM (proton exchange membrane), ammonia itself poisons the chemistry and has to be separated into nitrogen and hydrogen ahead of time. There has been substantial progress in separation devices. 

But even more important, at a best guess, is the fact that there are other kinds of fuel cell than the (acid-chemistry) PEM, but which are not poisoned by ammonia and can be fed it directly, without a separate separation phase. These include alkaline-chemistry fuel cells, and solid oxide fuel cells (SOFCs). These have come a long way in the past decade. The state of the art is that SOFCs are efficient enough for small stationary power units (1-10 kW range) but still too heavy for transportation, although approaching something that might be usable on a ship or locomotive fairly soon.

Your flying car, not so soon. But we're probably only talking another decade.

For a futurist, this is close enough to qualify as a win. In my previous post, I pointed out that "Ammonia, the Fuel of the Future" scanned with a classic American tune, "Columbia, the Gem of the Ocean." To celebrate, I sat down and penned out a couple of full verses. You may now sing them with your friends in full glee:

Ammonia, the fuel of the future!
your vector of useful energy
for your house, for your car, your computer,
for airplanes; even for the ships at sea.
It's a liquid you pour into a fuel tank,
Yet ammonia's entirely carbon free;
From the good old days of Bosch and Haber,
We understand the right chemistry.

Understand the right chemistry!
Understand the right chemistry!
From the good old days of Bosch and Haber,
We understand the right chemistry.

Ammonia, the fuel of the future!
With power for you and for me;
'Tis silent and couldn't be smoother;
when piped into an SOFC.
Do not thermalize your potential;
We know how to handle NH3 --
No need for Carnot and his limits
to generate electricity!

Generating electricity!
Generating electricity!
Ammonia, the fuel of the future,
generating electricity!

... and there you have it!

One more look at COVID numbers

 Given that my fit of a lognormal to excess deaths in my previous post worked so well, 

I thought I would revisit the stats and see how things were going two months later and whether I might need to add a third wave to the fit.

The actual national numbers don't appear to support that, so I thought perhaps I would take the state-by-state data (from the CDC) and see if there were separate patterns that made up the whole curve. I still couldn't see much in the way of a third wave, but this illuminated the first and second waves quite a bit:

Each trace is a single state (plus DC and PR with NYC separated out) of the absolute number by which all-cause mortality exceeded the high end of the expected range. Note that you would not even see the normal winter flu mortality peak in such a graph: the "expected" number goes up along with the actual deaths.

It's the orange line on this graph. It cycles between 50 to 60 thousand deaths per week over the course of a normal year. Note in particular that it's going up now, as people are not outside as much and not getting as much Vitamin D. Also note that the total number of people dying from any cause whatsoever is flattening and they will soon cross.

The major thing to note in the first graph is that the states making up the bulk of the first wave are different than the states making up the second wave. To a rough approximation, each state only gets one wave. States that didn't participate in the April or August waves are now getting their own waves, incoherently, so there's no distinct third wave but a statistical mush.

Here's the same graph again, but instead of absolute numbers we have the percentage over expected deaths for the particular state:

Now the vertical scale is percent. New York got up to 6 times the expected death rate in April, but nowhere got much more than 50% in August, and the average in the statistical mush is getting back down to zero, with a spread of 10-20%.

Note that that's not quite normal: normal is that the TOP of the spread is zero, as on the left-hand side of the graph a month before the pandemic hit. But remember the average American has a roughly 1% chance of dying in a given year; now he has a 1.1% chance of dying next year. Unless the vaccines work out, of course.

Eleven days in May, revisited

 Well, it's 9/11/2020, so a young man's thoughts flit lightly towards major disasters. A few months (!) back I wrote a post attempting to put the mortality of COVID in perspective, and here we sit; perhaps a longer perspective is now possible.

In the meantime the CDC has released data indicating that the vast majority of Covid deaths had major co-morbidity factors, which bolsters the interpretation that it best be looked at as essentially an acceleration of your remaining days in this vale of tears.

Rather than try to guess how many people actually died of, rather than with, COVID, I simply took the number of people who died of any cause, and compared that to the average non-Covid numbers. You can these stats at the CDC website here.

It now looks like this: 

The initial spike got a longer tail, and there was a second wave, and of course there will be a tail from the second wave you can't see yet due to reporting lag. 

Back in May I wrote that the overall effect of the pandemic had been to reduce everyone's life expectancy by 11 days. What does it look like now?

The low curves are the number of days we were losing per week, the upper ones are cumulative, and the two curves represent the differences from average and statistically significant figures. The "real" number should be between them.

So we lost eleven days in May (actually mostly April), then it slacked off some, and then picked up a bit for the second wave, and is going down again. As of right now we are losing about a day of life expectancy a week, but that is declining. The total amount of life expectancy we've lost so far is closing in on a month.

By the way, remember that average American life expectancy has been improving throughout history. How far would you have to go back to have a life expectancy one month shorter? About to 2010, it turns out. And this decade has been one of abnormally slow life expectancy growth; over the 20th century we typically gained 2 or 3 months a year.

Well, that's one way to look at it, but what's gone is gone. How much more life expectancy can you expect to lose?

To predict the remainder of the loss I fit a curve to the excess deaths. A lognormal curve fits the spikes quite well, as you can see:

The heavy blue curve is actual excess deaths, the orange is a lognormal fit to the first spike, and we add a second lognormal to it for the green, total, curve.  Note that these are in thousands of excess deaths in the US on a weekly basis. You need to compare them to an average 55,000 deaths per week or nearly 3 million a year. 

The total area under the green curve from current date on is just 14,000 deaths, which works out to be 43 hours of life expectancy loss.

 You can all relax now.

AGI: What, how, and when?

 Recently, Robin Hanson posted the following predictions:

 AGI isn’t coming in the next thirty years. Neither are Moon or Mars colonies, or starships. Or immortality. Or nano-assemblers or ems. ... So if you are looking for science-fiction-level excitement re dramatic changes over this period, due to a big change we can foresee today, you’ll be disappointed.

I agree about starships and immortality, and have different probabilities for some of the other things. But I replied (on his Facebook page) as follows:

We will almost certainly have AGI in 30 years, and probably in 10. A teleworker directing a machine looks, from the outside, a lot like an AI running it. The big tech issue of the next decade or two will be working out how the two can be seamlessly integrated as the balance shifts from telehumans to pure machine. This applies as much to Rosey the Robot as to self-driving trucks.

To which he rejoined

Care to bet on that "probably in 10" claim?

and me:

Give odds and a definition.

Robin:

AGI could take away most jobs from humans, right? So what odds do you give to >50% unemployment rate in over year 15 from now?

me: 

Wrong. To displace humans, AI would have to be cheaper, and that was not part of the ""probably in 10" claim". Furthermore, it's not necessary that there is a fixed amount of work to be done; instead of doing the same with less humans, we could simply make and accomplish more things. Like colonies on the moon, etc. But again that's a prediction not of what will be technically feasible, but of whether we collectively want to do it. And again, not part of my claim.

More directly to the point, I do not accept " >50% unemployment rate " as a definition of "AGI exists."

Robin:

... but do you really think AGI could exist yet stay expensive for a long time? I'm much less interested in when it "exists" than when it is useful.

Me:

You might want to read Matt Ridley's "How Innovation Works." AGI is going to come out of the confluence of a bunch of precursor technologies, and it will evolve, not be invented in a flash in some ivory tower. Telework is very likely one of the precursors; it will force people to learn how to break intelligent action into components, and then begin to try to automate some of the components. As this process goes forward, there will be niches where comparative advantage puts AI and others where it puts people. So AI will evolve toward being useful for people. But at the same time, people will evolve (culturally, not genetically, given timeframe) toward being more compatible with AI. Imagine you have a bunch of people working for you. At first, they are unskilled laborers. They save you time by mowing your lawn. But over time, they get smarter. They begin to be able to help with your research and writing. Now they are like graduate students. Pretty soon they are as smart as you are but they still don't replace you; they still relieve you of work but that simply allows you more scope, done at a higher level of abstraction. Somewhere in there your job changes from doing X to managing robots that do X. But managers make more than workers; the robots have become more valuable but so have you.

Tim Tyler puts in:

We were talking about 10 years from now. We will likely have some pretty sophisticated machine intelligence by then - in agreement with JoSH, but not Hanson - but machines won't be able to do all jobs more efficiently and more cheaply at that stage - and there likely won't be lots of unemployed humans knocking around either. Ricardo [a reference to the economic principle of comparative advantage] will not save the human race from obsolescence - but in 10 years, humans will still be needed.

Robin replies

I agree AI will improve, but at the same rate we've seen for 50 years. The question is what YOU can mean by "AGI in ten years" if it isn't cashed out in what they actually do in society.

Tim Tyler:

JoSH said: "To displace humans, AI would have to be cheaper". He's talking about the existence of the machines, not their cost-competitiveness. There are other issues that may hinder adoption of intelligent machines: regulations, preferences for humans in some jobs, sensor/actuator progress, training time/cost, etc.

and Adam Ford asks:

J Storrs Hall it seems your picture of AGI doesn't require AI understanding. Is that correct?

Not sure if this question can be answered directly atm - if not treat it as an intuition pump -  how far can generality be taken without there being understanding in the agent?

Since generality is a spectrum, and not an ideal state, I'm not confident that if we did achieve generality in AI, it would quickly force humans out of most jobs.

So I thought I should compose a blog post putting together my thoughts on the subject of AI, AGI, how, what, and particularly when. Come to look at the blog and lo and behold, I had already done a good part of that, four years ago: The Coming AI Phase Change 

... which did a fairly good job predicting the kind of progress we've seen in AI in the interim. 

Robin's notion of  "AI will improve, but at the same rate we've seen for 50 years" might well be summed up by the change in the state of the art between the famous Dartmouth AI workshop in 1956 and the 50th anniversary conference, also at Dartmouth, in 2006. I was there (for the second one), and so was Geoff Hinton, who gave a very nice talk on the history and future of "neural networks" and what has come to be called "deep learning" in the meantime. I can assure you that the rank and file of classical symbolic AI researchers very much thought of Hinton as a second-class citizen. But the actual achievements of AI since, ranging from Alpha-Go to GPT-3, show us that Hinton had something they had missed. 

I gave a paper at that conference too, which I mention because it goes to the answer of Adam's question above. AI had always meant "building a machine that could think like a human," but over the 50 years in academia it got subjected to so much grade inflation that it had come to mean essentially "the clever programming that captures some specific skill" such as playing chess or driving a car. By the 21st century various people, notably Ben Goertzel, had come to use "AGI" to try and recapture the original meaning. An AGI would be simply a human-level AI, one that could be expected to be able to, or be able to learn to, change a diaper, plan an invasion, butcher a hog, conn a ship, design a building, write a sonnet, balance accounts, build a wall, set a bone, comfort the dying, take orders, give orders, cooperate, act alone, solve equations, analyze a new problem, pitch manure, program a computer, cook a tasty meal, fight efficiently, and maybe even die gallantly.

More directly to the point, an AI would be able to have a deep brandy-and-cigars discussion with you, examine its own motives, know the limits of its own knowledge, and so forth. It would not necessarily actually get this right: as Robin propounds, humans don't get this right all the time, but we went for millennia with a geocentric cosmology as well. 

So how would you go about building an A(G)I? I have plenty of ideas on this too, but unfortunately the margin is too small to contain them.

 

Eleven Days in May

Amidst all the hype surrounding the coronavirus, I thought it would be useful to find some perspective on how bad it really is. One of the big problems with the reporting is that everyone seems to have a different way to count who actually died of it, or who got it and died for some some other reason.
One good way to finesse the issue is simply to look at the numbers of people who died from any reason at all, and see if that is significantly different from before the virus got loose.
There is a very informative page at the CDC website which lists exactly that.
Here is the graph it shows for the past few years:

Note several things it shows you: on an average, about 56,000 Americans die every week. There is an annual cycle that varies this by about 10%. The statistically significant border, (orange line) is about 2000 higher than the average; what that means is that variations lower than the line are to be expected and shouldn't be interpreted as proving anything.
Back in the winter of '18 we had a bad flu season and deaths broke through the line for about a month. We can compare that to the current peak and say several things. First, the coronavirus is definitely worse than even a bad flu season. But it's a judgement call whether it's enough worse to be considered a qualitatively different thing.
Second, the effect of the virus is very different from place to place. Here is New York City (you can call up all these from the page):



Yikes. The virus was an unmitigated disaster there. But look at, say, North Carolina:







There is no discernible impact from the virus at all. (Note that the falloff in the last few weeks, in all the graphs, is not real and an artifact of reporting lag.)

So what does this all mean? Raw numbers of deaths need a bit of perspective themselves. Here's a graph of how long you can expect to live as an American:


It works like this: God picks a point for you on the left axis, and you extend a line out horizontally until you hit the curve, at which point you die.
It turns out that given the numbers above, we can calculate where the curve would be given the coronavirus. In fact, I have done so, and the results are shown in the red line on the above graph. Can't see it? Neither can I; it's only 11 days shorter than the original black one. You have to zoom in to a scale of a couple of years even to see the difference:



The best way to think about the virus for the US population as a whole is that it has cost each of us 11 days of life. For most of us, that's in the noise. If you are in New York, or way out to the right side nearing the curve, it looms larger.

The spherical 5-bar linkage

I've been working on one, and amused to find that the physical construction and control software took about the same amount of time. Here's a movie, under control of the mouse in the other hand:

The idea of a spherical linkage is that all the motion takes place on the surface of an imaginary sphere, but in reality the links can be at different radii as long as their motion is well-described by rigid shapes moving on the sphere.  That makes it possible to construct in the real world.
There's some fairly cute math involved in converting a position you want the end axis to be, to angles to set the driving servos.  Quaternions are involved.