Things like racism or sexism, or over-generalising on a class of people is one of the more socially inappropriate things you can do. However, depending on how your logic system works, it’s not an entirely unreasonable method of thinking (the word “unreasonable” chosen purposefully) – and for any other subject, where the things being reasoned about are not humans, we wouldn’t particularly care. In fact, certain subjects like religion and spirituality are held to less strict standards of reasoning… there’s actually more defense in being racist/sexist then being a practitioner of certain religions. Perhaps this is why these occasionally go hand in hand[1].

So what do I actually mean by this? I’m going to use two methods of reasoning, deduction and induction, and then explain them in terms of uncertain truth. Nothing in this world is ultimately absolute[2] and so it behooves us to include probabilistic uncertainty in to any conclusion or relationship within our logic set.

Deduction can be summarised as inferring the specific from the general. e.g.

Note that obviously there are exceptions to the starting generalisation (1) of four-legged cats. Poor Aristotle could have been in a car accident and had a leg amputated making (2) a false conclusion. That’s where the probability and uncertain truth comes in…

Here we’ll use two parts to truth values (TVs), strength and confidence[3]. The first, strength, is how often the relationship holds true. E.g. with strength 0.9 for the we’d expect approximately 90% of cats we saw to have four legs. Confidence however, indicates how sure we are of the strength. The exact semantics of confidence are not important for the current discussion, but basically if the confidence is low then we’re not very sure about the number of legs cats have – perhaps because we haven’t seen many.

On the other hand induction is used to infer the general from the specific. e.g.

In other words, if, within all our experience, all cats we’ve seen have four legs it’s sensible to assume all cats do. With such a small sample size of only 3 cats, and without any other background knowledge (like animals of a particular species generally all have the same number of legs) we’d generally not afford much confidence in the result of this deduction[5].

So where does the sex/race/other-isms come in? Well if you haven’t worked it out yet, it’s when you use too small a sample set to do induction or you adopt generalisations from others without using your own evidence[4]. Or perhaps you do know lots of one sex that all fit your model, in that case the problem is using a single generalisation to deduct information about a new member of that sex that you know nothing about. Essentially sexism/racism is due to too large a bias towards sex/race determining an overall conclusion or understanding of the other.

I guess one thing should be clarified, the relationship between generalisation of sexism/racism. When does a generalisation turn into one of these categories. I’m not sure I really know, after all if you made generalisations about the basic anatomy of males/female or say something superficial like the colour of someone’s skin, then there wouldn’t be a whole lot of controversy there[6] but I guess it’s when you use that information to fully define everyone that falls into a category. The thing that I’m grappling with, is that, in the absence of any extra information on someone it’s sensible to use what you have to start trying to represent and understand someone new. In fact, if you didn’t, you’re demonstrating you don’t care enough about the other person to think about them at all!

And then we come to insurance brokers. They make huge generalizations. Law usually prevents them from using race as a factor, but for some reason they are still allowed to use your sex to determine the premium you’ll pay. They are also allowed to generalise on your age among many other things. In fact, pretty much the entire way the insurance industry works is through generalisations.

So what makes it okay for businesses to blatantly be sexist and ageist among many other categorical assumptions? I don’t know.

Ahem… got a little side-tracked there. Sorry…

So when is this generalization okay? If every interaction you have with people that fall into a category supports the conclusion then your confidence will increase – and you’ll have no examples of any alternative. Few people actually have such a life history to conclude this. I suspect many people who display racism or sexist tendencies probably get this strong bias from another human or information source. If it’s someone they highly respect, such as a parent, or someone who’s the social leader amongst a group, then they may adopt conclusions with high confidence without having the experience to conclude things themselves.

How can we protect against broad conclusions about people?

• By making it exceedingly difficult to increase our confidence of very broad generalizations. Realise that when making any large scale conclusion that confidence must be limited – especially when reasoning about dynamic entities like humans.
• Favour the use of more specific categories when reasoning. This can be hard, as when you don’t know someone and all you have is their apparent race and sex then any generalisations you’ve made will be your sole source of information. You can argue that this shouldn’t be the case, but any pre-consolidated knowledge you have will come to the fore until you learn more about the person to develop a more complete representation of them.

Why is it hard to change the behavior or views of people who are already displaying ism-ness?

• I guess that confidence in belief generally erodes slowly, unless presented with directly conflicting evidence. And sometimes conflicting evidence just makes people stop listening to you because it radically challenges their world view.
• Perhaps it’s physically more effort/energy to update generalisations on large groups? If you have to adjust your confidence of conclusion on a large group, then they’ll be many many connections from the neuron that potentially represents that group and the instances and associations within that group. Since the brain is massively connected, larger groups probably require more effort to shift. Think of all those tens of thousands of synapses having to reconfigure themselves or die.
• Generalisations may be self-reinforcing. While someone might not fit your generalisation, that may slightly alter your confidence in it, but not enough to completely remove it from your mind.

I hope I’ve raised some questions in the readers head. I don’t like sexism or racism, but at the same time I can (kind of) understand the logical reasoning that might lead to it and I haven’t completely resolved how it’d sit in a fully functional artificial intelligence. After all, I really don’t want OpenCog to turn into a racist bigot!

[1] I was going to link to an appropriate example here, but couldn’t easily think of one I could back up, feel free to suggest one.
[2] If people think otherwise, I suggest they study enough quantum physics or Gödel’s incompleteness theorem.
[3] This is based on the simple truth value type of OpenCog.
[4] Not that accepting generalizations from others is bad thing. Everyone does it because we cannot experience everything directly.
[5] Although if your sum experience of the universe you live in is merely these three cats then you’d be very confident and not only that but induct that everything has four legs (and is a cat).
[6] Yeah yeah, I know, some people are transgendered. I’m keeping this simple okay?

One evening, I had an interesting discussion in bed with my partner that was related to licensing such technology. OpenCog is open source, which is my personal preference for the development of AGI, even if the jury on whether it’s beneficial or needless reckless is still out. As open source software, it means that if we sold expert systems based on OpenCog to end-users, we’d have to also provide the source. Even though we can license it under different terms from SIAI, this isn’t entirely needed since my current viewpoint is that the real value will be in data within the system.

As an analogy, the biological design for the brain isn’t what makes us unique or what encompasses our knowledge and experience of the world. The pattern that’s formed during our childhood and education is what is really valuable, otherwise we’d make no distinction between twins and not particularly care if one twin passed away.

So, the digital mind’s pattern would be the important part that we’d license or sell. However dealing with a dynamic system makes that interesting, since the pattern that was sold/licensed would inevitably change. Learning software could well have the valuable part (the “identity” if you will) morph and change beyond the original deployment. In fact, the software could learn new things which makes the individual deployment “smarter” than the original or any other deployment.

In that case, who owns those improvements? Should we get
the rights, since it was our software that altered itself, or does it belong to the license-holder since the AI learnt the improvement in their environment?

I’m sure with sufficiently rigorous legal-work one could protect towards one view over another, but I’m more interested in what seems right.

“The Singularity Institutes “main purpose” is meant to be to investigate whether a recursively improving intelligence can maintain “friendliness” towards human kind. “

Okay, but my standpoint is: Why does the recursively improving intelligence need to be non-human? It seems counter-intuitive to me to devolve this power to something outside of ourselves – and also a bit like we’re just trying vainly to become a kind of God, creating another type of being.

I think the main reason there is a focus on AI rather than improvement of human intelligence is because it’s so damn hard to do experiments on people’s brains. It’s ethically difficult to justify various experiments, and it only gets harder as things become more regulated (and rightfully so for the most case). I think they’ll definitely be continuing research into this stuff though. For myself, occasionally taking Modafinil enhances my productivity significantly (so long as I maintain focus on what I’m meant to be doing, it’s easy to get enthralled with something that interests me, but isn’t related to my work).

But there’s no exclusion of human intelligence amplification from the singularity concept. If we create smarter humans, then this begets even smarter humans. Again we can’t really predict what those enhanced “humans” would do, because they are a significant step smarter than us.

Human intelligence amplification has a whole raft of other ethical issues associated with it though too. When it becomes more mainstream/available it’s going to be a major political and social issue. What happens when not everyone can afford (or wants) to enhance themselves? Will we develop two classes? One of naturals and another of post-humans? Will employers require certain professions to use performance enhancement (say for example, for brain surgeons performing long surgeries)? It’s also going to raise the question about the ownership of our bodies. There are laws against taking recreational drugs, but for some, LSD helps with certain types of thought and could be seen as a form of an intelligence manipulator (or an amplifier of certain facets of intelligence).

At the moment at least, governments and enforcement agencies seem completely uninterested in actively stopping this, due to the prevalence of various performance enhancement drugs throughout academia and other cognitively demanding professions. Obviously it’s not necessary, but for some the edge or boost it gives them is sufficient to outweigh the risks of off-prescription drug use.

That was my main beef with Kurzweil: He assumed that the intelligence beyond the horizon would be non-human. This, of course, begs the deeper philosophical question: what *is* “human?” But I’m sure you’ve mulled that one over plenty. Indeed, it would be interesting to hear what you have to say on the subject.

It simply strikes me that, the human ego being what it is, we would naturally be trying to improve our own intelligence and not worrying too much about creating AI as a standalone entity. Am I wrong? Is the human ego instead more interested in giving birth to a new species?

There is one other reason there is a focus on AI. Which is related to human ego. Once humans have an advantage over others, such as a significant step up in intelligence, then there’s a good
chance some will use that power over others in a negative way. It’s the old adage, “power corrupts, and absolute power corrupts absolutely” – and it’s my opinion, that this could lead to the same power/intelligence imbalance between an intelligence-amplified human and humans, as that between humans and dogs.

Or more succinctly, humans have plenty of evolutionary baggage that we
don’t necessarily want to amplify!

Yeah I’ve thought about that a little as well. What I wonder (and this really is just pure speculation) is whether hyper-intelligent humans would have more faculty for reason and logic; and, if so, if they would see the benefits in spreading the gift and/or using their gifts in more compassionate or benevolent ways…? Because, surely, that’s evolution as well.

Again, it would seem somehow counter-intuitive to improve intelligence without, for example, attempting to improve our capacity for processing emotions — our ‘emotional intelligence’, as it were.

I tend to think that many of the world’s problems today come down to a focus on what is scientifically possible rather than what is philosophically ‘good’ or ‘true’.

Basic agreement, improving intelligence is really just a catch all for all sort of cognition improvements. Depending on where we focussed neuron regeneration, we could expect to see improvements in different aspects of cognition.

Science and technology will inevitably advance, and while we can regulate technology to some extent, we can’t stop it. Telling people they can’t work on AI would merely push it underground, and one of the reason I work on OpenCog is because it’s an open-source framework. It can be inspected by other experts for flaws to ensure there isn’t some hidden time-bomb sitting in the code, and allows an international approach which might otherwise be over-regulated within a university-specific project (this ignores the actual issue of friendly vs. unfriendly AI, since the latter doesn’t have to be intentionally designed – I may write more about that one day, but SIAI and it’s fellows have already written plenty on that).

To bring it back to AI, what place is there for emotions in an artifically constructed intelligence? Is there a capacity for compassion or fear, if they serve a function for processing input and generating responses? I am guessing that cognitive scientists have thought about this one…

That’s a BIG topic, so all I’ll say is that my opinion is that emotions are just a particular mind state. Sure this mind state is influenced by hormones and neurotransmitters, but so is the rest of the brain’s functioning.

My opinion is that emotions are just extremely strong aspects of a human’s mental world. If we gave an AI an extremely strong desire/goal to make humans happy, and the AI design had some kind of reward based system (such that the AI was trying to maximise these) and achieving these rewards caused other effects in the AI’s mind, such a propensity to use positive phrases to describe the world, then is there any reason to believe that the AI isn’t happy itself?

Disclaimer: I’m no longer in the employ of the SIAI, so my thoughts are not endorsed by them at all.

This interesting fact and this analysis, thus this is the point.

Except that made for very short essays that were no where near the word limit.

Part of that was because I was used to maths, physics, and computer science problems. These fields don’t tend to push students to writing supposition about theory, at least not at the University I went to, and it’s usually pretty concrete for the level taught to undergraduates. Then midway through my study I switched tracks into Biology and flailed around for a while. They expected you to write more than a paragraph at a time, and form coherent arguments in exams without the benefit of cut-and-paste! The horror. After while I got over my initial shock, and actually got reasonably good at writing (my recent resurgence of posts on this blog is an attempt to practice those skills which are getting rusty from too much code and analysis of experimental results). Even though I managed to get pretty good, I still didn’t quite understand. I thought it was a waste of time when I could construct the argument for my 2000 word essay in a paragraph.

I got over that though. I saw that one has to also explain their reasoning, because any conclusions about the real world based on science usually contain assumptions. It’s important to not only be explicit about these assumptions, but to also explain why these are the assumptions chosen.

Even more recently, I’ve worked on economic attention allocation (ECAN) in OpenCog, which controls the flow of attention within an artificial mind. Why is this necessary? Because the mind’s resources are finite – based on the computers available memory and processing ability, and we obviously also have a finite capacity for focussed thought too (well, yes, it’s finite because we only have so many neurons, but I’m referring more to the limited number of concepts and relations we can consciously work with at any one time). Much of how the attention allocation works in OpenCog is based on importance diffusion between close concepts, but the concepts also need to be primed with importance in some way. If diffusion was the only process that moved attentional importance, you’d eventually end up with a homogeneous soup where everything was equally important!

The way this happens is somewhat complicated in OpenCog, and not immediately relevant to the point, but one thing to note is that external stimulus excites contextually relevant information. What that means is, if OpenCog sees that a cat walked past… then knowledge about cats (they are furry 4 legged felines) and perhaps locomotion/kinematics is stimulated (if the cat keeps walking in that direction it will fall off that balcony and fall due to gravity… okay, so maybe that’s more about prediction, but general knowledge about walking would be available in order to be used to make that prediction). By stimulating this knowledge, it’s made more important in the systems mind. Beyond the cat example[1] the same goes for reading someone’s ideas. If someone just tells you “the sky is green, purple and sparkly” and leaves it at that, you’d probably just think they were taking some particular effective hallucinogenics… especially if you were inside, at a conference about refrigerators (if it was a metereological conference, you might be willing to concede that they were right, because they are an expert in the field and have a reason for believing as such). But if the the refrigerator conference attendee first prefixed their statement with “Y’know, I was just holidaying in Alaska, and I saw the Northern lights, up there… “, then assuming you vaguely knew about auroras changing the colour of the sky, then you’d be able to believe the refrigeration expert. Or at least believe they weren’t in the habit of attending refrigeration events under the influence of mind-altering substances.

[1] It’s okay, the cat didn’t actually fall off the balcony.

I’ll be attending Burning man. I’m immensely looking forward to this as this is the first year in several that’s actually been feasible for me to get there from New Zealand. I’ll be with an Australian theme camp called Straya that a friend of mine put me in contact with, and who’ll also be there.

As well as Burning man, I plan to hang out in Washington D.C with Ben to talk about our work on OpenCog. Then I’ll stay in San Francisco for 5-6 weeks (end of Sep till start of Nov) to attend the Singularity Summit followed by the CogDev Workshop (an OpenCog coding jam, details to be finalised, but likely to be just after the Summit).

If you’ll be at any of these events and want to chat, drop me a line

Now that doesn’t mean a consciousness can’t be revived, the pattern is still there, and if the process can be restarted then I suspect the consciousness would continue as if nothing happen. One moment about to die, the next revived. This is essentially what proponents of cryogenics expect to occur.

Did I just contradict myself, by saying that consciousness can be revived from the pattern, even though I claimed the pattern wasn’t the mind? I don’t believe so. The pattern is the painting, the mind is painter. In humans, the painter is the physiological processes that generate the electrical signals shooting through our body and that update the neuronal structure in our brain.

Want to work on AI/language-processing over the Northern Hemisphere summer? Here are some of the ideas for projects proposed. Applications to Google open on the 24th of March.