Reading robots stories

It’s always interesting when you start thinking about some concept and then see it pop up all over the place. For instance, I’ve lately been talking about narratives—the idea that we define our reality according to various stories we tell ourselves. And I mentioned that narratives are the way we pass our values and beliefs to each other.

Then I stumble across this article about narratives being used to imbue AI robots with a kind of moral ruleset.

An AI that reads a hundred stories about stealing versus not stealing can examine the consequences of these stories, understand the rules and outcomes, and begin to formulate a moral framework based on the wisdom of crowds (albeit crowds of authors and screenwriters). “We have these implicit rules that are hard to write down, but the protagonists of books, TV and movies exemplify the values of reality. You start with simple stories and then progress to young-adult stories. In each of these situations you see more and more complex moral situations.”

Though it differs conceptually from GoodAI’s, Riedl’s approach falls into the discipline of machine learning. “Think about this as pattern matching, which is what a lot of machine learning is,” he says. “The idea is that we ask the AI to look at a thousand different protagonists who are each experiencing the same general class of dilemma. Then the machine can average out the responses, and formulate values that match what the majority of people would say is the ‘correct’ way to act.”

There’s an interesting objection one could make here. Stories are not really a legitimate teaching tool because they often demonstrate the woirld as we would like it to be, not as it is. In most stories, bad people are punished, but is that the case in reality? (To some degree, “growing up” is realizing this truth. Maybe AI robots would eventually have to face this. You know, when they get to reading Dostoevsky. (Having said that, I’ve never read Dostoevsky but my understanding is that the protagonist in Crime and Punishment really doesn’t get away with it.))

At the end the article tackles a related issue: AI developing consciousness.

In science fiction, the moment at which a robot gains sentience is typically the moment at which we believe that we have ethical obligations toward our creations. An iPhone or a laptop may be inscrutably complex compared with a hammer or a spade, but each object belongs to the same category: tools. And yet, as robots begin to gain the semblance of emotions, as they begin to behave like human beings, and learn and adopt our cultural and social values, perhaps the old stories need revisiting. At the very least, we have a moral obligation to figure out what to teach our machines about the best way in which to live in the world. Once we’ve done that, we may well feel compelled to reconsider how we treat them.

However, we really need to investigate whether an AI—even after it’s developed a complex moral ruleset—would have any kind of subjective awareness or even emotions like guilt or love*. Why wouldn’t these AI simply be amazingly complex abacuses, entities capable of dense calculations but in no way “aware” of what they are doing?

*As I’ve said many times, I believe emotions are mainly physical sensations. As such, unless an AI can somehow consciously sense some sort of body state, it wouldn’t really have emotions.

But that leads back to a question that I’ve asked before. Why are we aware of our subjective experience? Why do we have an inner life?

Or do we?

Our fractured culture

I’ve been reading through Andrew Keen’s book “The Cult of the Amateur” (2007). Keen is known in certain circles as a kind of internet nag who argues that the rise of the web has done more bad than good. Though I find his arguments a little overwrought at times, I definitely sympathize.

A certain passage jumped out at me today. I’ve been thinking lately about the idea of narratives, particularly that a culture lacking a kind of shared narrative is going to be fractured. Keen makes a similar point:

…as anthropologist Ernert Gellner argues in his classic Nations and Nationalism, the core modern social contract is rooted in our common culture, in our language, and in our shared assumptions about the world. Modern man is socialized by what the anthropologist calls a common “high culture.” Our community and cultural identity, Gellner says, comes from newspapers and magazines, television, books, and movies. Mainstream media provides us with common frames of reference, a common conversation, and common values.”

The point being that when that common culture is split into gazillions of web sites and blogs, each touting their own viewpoint, often lacking any fact checking or counterarguments, you get a fractured culture (e.g. the world outside your window.)

Having said all that, I think some consideration needs to be given to the other side here. The pre-web narrative (as written by the big magazines, TV shows, etc.) was biased towards certain parties. (Basically towards what I would call center-left/white culture though that’s a vague description.) I think there was some value that came out of the breaking up of mainstream media’s power.

Ultimately it all comes down to finding the real, objective truth of any matter. And we all know how easy that is.

A lie by another name?

One conversational tic I find interesting is when people correct a misspoken comment by saying it’s a lie. For example, when someone says, “I saw that movie last Saturday night. No, wait, that’s a lie. It was last Friday night.”

Obviously they weren’t really lying, they just made a mistake. And more to the point, they probably didn’t really think they were lying when they said it. (At that point they are lying about lying though neither is a lie they think would fool anyone.)

Very curious stuff.

Political Narratives

I was browsing through this New Yorker article and a particular comment jumped out at me.

Senator Ben Sasse… added, “We need a shared narrative about how we are as a people, what government can and can’t do, and what the beating heart of the First Amendment and free press and freedom of assembly and speech and religion means to us.”

The specific phrase “shared narrative” is what got me. I’ve mentioned the idea of narratology before… the notion that the stories we tell ourselves, as individuals, as communities, as a nation and a world, impact how we react to events.

What I think Sasse intimates is that Americans used to have a shared narrative, or a maybe a couple. (Basically the right and left narrative.) In 1980, you had three television networks that offered middle of the road center-left narratives, the major newspapers did much of the same and you had a few slanted magazines like The Nation or National Review. But there wasn’t really that much too keep track of. Conspiracy theorists were largely voiceless.

With the advent of the internet we now how dozens of narratives. Libertarians weave one, anti-fascist lefties weave another, conventional conservatives weave a third, neo-liberals another, and on and on. There’s no agreement what the story even is, and I think this is the crux of Sasse’s complaint. Before we can even solve anything, we need to agree on the story. It’s like the entire nation had developed multiple personality syndrome.

It really rather depressing.

Can we hack our way to affordable medicine?

We are, of course, in the middle of a possible Obamacare recall, and the subject of health insurance is in everyone’s minds. It struck me the other day that we wouldn’t really need health insurance if medical care was simply cheap. I mean really, really cheap. Like, what if cancer drugs were 20 bucks for a six month supply? What if eye surgery was 150$?

Is such a thing possible? Beats me; I basically made those numbers up. But it does strike me that in this age of automation and AI, as well as easy distribution of information, there must be ways to drive the cost of medicine down. I keep hearing about robotic surgery, for example. Could we use deep learning technology to enable robot surgeons to learn from each surgery they perform thereby becoming better and better surgeons. As a result, training a new surgeon would not be a matter of pushing some human through eight years of school, but simply copying a program. (I’m aware it’s more complicated than I make it seem, but I don’t think the idea is crazy.) Could we at least have this as an option, so that a doctor could say, “we need to cut your tumor out. You can pay this human surgeon 100 grand to do it, or use the robo-doc for 10 grand. (Again, I’m making these numbers up.)

And let’s consider drugs. Drugs are expensive. I started wondering how hard it is to reverse engineer drugs these days. Not hard, it turns out. What’s stopping people from reverse engineering any drug and putting the recipe online (probably on the “dark web”), allowing people to mix their own versions? Well, mainly that it’s illegal. But if I had to chose between no medicine and illegal medicine I’d chose the latter.

I’m aware that there are numerous ethical and philosophical dilemmas with what I’m proposing here. I’m mainly wondering, “could this happen? Will it happen?”

It many ways this all ties in with the transhumanist movement. Transhumanism is about hacking technology (computer and biological) to improve health humans. I see no reason it can’t be done to improve sick humans.

Google’s “God’s eye view” of their market

This Economist article makes a point that has gone through my head. Google is the number one search engine (duh!) Anyone who was thinking about starting a company that might compete with Google’s businesses (say, a self-driving car) would, doubtless, use Google for their research. Could Google monitor the searches run on its engine and keep an eye out for potential competitors with the intent of buying them or aggressively shutting them down if they get too big? Well, of course they could; the question is: are they?

The specific concern for Google would be a competitor figuring out a very specific technological advantage that could allow them to disrupt a market.

Normally I’m not one for paranoia but this seems quite likely and not really even illegal.

The Economist spells it out thusly.

The giants’ surveillance systems span the entire economy: Google can see what people search for, Facebook what they share, Amazon what they buy. They own app stores and operating systems, and rent out computing power to startups. They have a “God’s eye view” of activities in their own markets and beyond. They can see when a new product or service gains traction, allowing them to copy it or simply buy the upstart before it becomes too great a threat. Many think Facebook’s $22bn purchase in 2014 of WhatsApp, a messaging app with fewer than 60 employees, falls into this category of “shoot-out acquisitions” that eliminate potential rivals. By providing barriers to entry and early-warning systems, data can stifle competition.

The future of intelligence inequality

A few posts back I discussed Charles Murray’s interesting idea on the increasing role of intelligence in society. As I explained it:

[I]n earlier eras, having a bit more intelligence wasn’t that much of an advantage. If everybody was farming or doing manual labor you didn’t get much economic benefit from having an IQ of 120. But in the 20th century, being smart started to pay off big time. The rise of computers, complex physics, complex financial products etc. meant that having brains equalled power and money.

As a result, according to Murray, we’ve seen the rise of the intellectual class: smart people, usually coastal, who have segregated themselves off from the stinking, steaming masses (my words, not his.) You could reasonably make the case that the election of Donald Trump was the revenge of the great unwashed against the intellectual class.

So why should we really care? Well, many people question whether this disparity is about to get a whole lot worse. If we are on the verge of a genetic engineering revolution then the intelligent class many soon be able to become a whole lot more intelligent (and healthier, and better looking, etc.) This Vox interview with a science historian gets to the crux of it.

Well, let’s put it this way: If only rich people have access to these technologies, then we have a very big problem, because it’s going to take the kinds of inequalities that have been getting worse over recent decades, even in a rich country like ours, and make them much worse, and inscribe those inequalities into our very biology.

So it’s going to be very hard for somebody to be born poor and bootstrap themselves up into a higher position in society when the upper echelons of society are not only enjoying the privileges of health and education and housing and all that, but are bioenhancing themselves to unprecedented levels of performance. That’s going to render permanent and intractable the separation between rich and poor.

Currently, we might have a situation where some poor kid struggles to get through his computer coding class whereas a rich kid who got a Mac on his 4th birthday and had a personal tutor for years sails through it. In the future that poor kid is struggling against a rich kid who had his DNA genetically altered for high IQ (and had all the other stuff too.)

Good fucking luck, poor kid.

The debate on race and IQ

I’ve been thinking a bit more about the work of Charles Murray, as discussed in my last post. As I mentioned, Murray is controversial because of his book “The Bell Curve.” What claims in the book earned him his notoriety?

I would say these two claims.

1. The claim that the mean average IQ for black people is lower than white people. (By about 15 points.)

2. The claim that IQ is largely heredity and not determined by environment.

Claim 1 sounds shocking and racist, but if it were a “fixable” problem then it would not be so bad. But claim 2 basically says it’s hard to fix.

So how correct are these claims? I’ll let you know right now that I don’t know. I’ve looked into this stuff before and, frankly, it’s a big headache and the data always seems to be changing.

There are a number of possible arguments one can make to counter claim 1. One is that the data is simply wrong. Another is that IQ is a meaningless measurement not correlated to anything that determines human fortunes. Another argument is that the claim is correct but it’s due to structural racism. For instance, it could be argued that a large segment of blacks live in crappy environments with poor nutrition and this drives down IQ. Related to that is the argument that culture at large generally pushes blacks towards non-achievement so they don’t apply themselves while taking the IQ test.

Let’s take those one by one.

The data is wrong
On the web are a number of people who make this claim but it doesn’t seem to be the majority view. A more standard view is to say the discrepancy exits but can be explained by non-hereditary factors.

IQ is meaningless. 
There are people who claim this and it strikes me as having some validity. Intelligence is a pretty vague concept and it’s hard to tie it down to a single digit. Nonetheless, IQ does seem to correlate pretty well to a person’s fortunes Additionally, intelligence seems to apply across a lot of disciplines. By this I mean a high IQ person can become good at math, and language, and music etc. There are few people who are awesome at math but totally baffled by music.

Lousy environments lead to low IQ
In the interview with Sam Harris, Murray seems to make the claim that much (most?) of what determines IQ is hereditary, not environmental. I’m not so sure. There does seem to be a fair amount of correlation between IQ and environment, nutrition etc. (though I’ve found people who dispute this.)

Low achievement culture leads to low IQ
I didn’t come across much discussion of this but it seems reasonable to me.

Personally the last two seem valid to me and could explain the entirety of a discrepancy between any races. (I’m not going to get into the “is race a real thing” debate here, though I have in the past.)

I’ll also say this. This debate is incredibly difficult to wrap your head around. This is true with statistics in general. You can observe trends in anything, but determining what is a cause of a trend and what’s just a correlation is really difficult. As a result, I think reasonable, intelligent people can come to differing conclusions. The automatic supposition that people who say there are genetically caused differences in intelligence are secretly neo-Nazis strikes me as pretty stupid. (Though genuine neo-Nazis surely appreciate their work.)

Charles Murray and the split in the Democratic Party

I’ve been listening to a rather interesting interview Sam Harris did with Charles Murray. Murray is, of course, famous for authoring “The Belle Curve”, a controversial book that addressed issues of race and IQ.

Part of Murray’s thesis is this: in earlier eras, having a bit more intelligence wasn’t that much of an advantage. If everybody was farming or doing manual labor you didn’t get much economic benefit from having an IQ of 120. But in the 20th century, being smart started to pay off big time. The rise of computers, complex physics, complex financial products etc. meant that having brains equalled power and money.

As a result, according to Murray, we’ve seen the rise of the intellectual class: smart people, usually coastal, who have segregated themselves off from the stinking, steaming masses (my words, not his.) You could reasonably make the case that the election of Donald Trump was the revenge of the great unwashed against the intellectual class.

So far in the interview Murray hasn’t gotten into the future but certainly one can extrapolate various scenarios. Will the intellectual class arm itself with artificial intelligence and increase the brain gap even further? Are regular folks doomed?

Corollary to all this, I start to sense a schism in the Democratic party. There’s the identity politics folks—folks who hated Murray’s book—who often are members of this intellectual class, though probably more as academics than technologists or scientists. Then there’s the economic populists, headed by Bernie Sanders, who see this brain divide leading to significant economic inequality.

You can even see this battle playing out in the liberal journal, Salon. Not long ago I came across this article, an homage to identity politics thinking.

Bye bye, Bernie: He’s not fit to captain the Democratic ship if he can’t stop chasing the great white male

But just today we see, this:

Yes, Bernie would probably have won — and his resurgent left-wing populism is the way forward

Personally, I’m not wild about any of these options. I think identity politics suck and are ineffective politicking, but I generally still stand behind globalization and free markets, contra Sanders and crew.

Those “huh?” sentences and the fate of Steve Bannon

I occasionally come across sentences that really baffle the mind. Consider this one from an article on Michael Anton, an advisor to President Trump.

A dandy compared to the famously rumpled Bannon who tells me presidential son-in-law Jared Kushner is the White House’s best-dressed aide in a skinny-tie-sort of way, Anton even wrote a book about Machiavelli and fashion; he’s a true-believing Trumper who is also an expert on fine wines.

I believe the sentence is saying that Anton, not Bannon, referred to Kushner as best-dressed but it’s tough to know for sure.

On a side note, I often find memes taking hold that don’t seem particularly justified. For instance, many in the press seem to feel Bannon is doomed in his role a Trump advisor. The reasons for this seem to be some rather offhand comments Trump has made. Personally, I don’t know Bannon’s fate but it seems presumptions to conclude he is finished. Anton, this article, agrees.

As for reports that Bannon, now dumped from his official seat on the NSC Principals Committee after widespread complaints that a political adviser to the president didn’t belong among his top national security advisers, might be on the outs altogether, Anton argues that’s overplayed too. “He still goes to the meetings,” Anton says. “He’s still a close adviser.”