Hacker News Reader

You need to learn without someone telling you what the answer is. Most of the time for biological intelligence there isn't an oracle describing the truth at every moment of life to judge actions/decisions against.

There is: Death, without reproducing. If an organism dies quickly but left children, they were biologically successful. If you didn't, then your evolutionary branch goes away.

I'm not sure this can be simulated, but it might be worth trying. Biological brains are the product of a billion years of the above strategy, so hopefully it's feasible to do the same thing.

Don't people already do that? Well, no. The strategy implies some sort of physics simulation with general rules, not a specific ruleset tuned to your domain.

>> Most of the time for biological intelligence there isn't an oracle describing the truth at every moment of life to judge actions/decisions against. [emphasis mine]

> There is: Death, without reproducing.

The problem is that it's not describing the truth at every moment of life, just the total approximate fitness of that lifeform by the end of its lifetime. It would be time consuming and cost-inefficient to simulate many possible configurations of actors until the end of their lifetimes.

Yet it's the only strategy that we're absolutely certain produces intelligent life.

Well, that it did produce intelligence once; not that it necessarily always does.

Although I do tend to think that given enough time (some millions of years), other intelligence will evolve on earth, from parrots or elephants or dolphins or octipuses or something. Though there's no evidence of a brain growth spurt that led to us, in other species; and it's not settled why we had it.

Are you absolutely certain we qualify as intelligent life?

edit: This comment isn't a joke, I mean this seriously. Modeling a system after ourselves may not be the best idea if we're not the best possible system.

It's a good question, and I think the answer is - up to a point.

We're biased towards individualism and the illusion of personal intelligence making clever decisions using its own resources, but in fact our real intelligence is collective.

Individually, most humans are a little bit smarter than other animals, but not much.

Humans are also excellent mimics. We learn by copying, and a lot of what we consider intelligence is copied behaviour and copied beliefs - often applied unconsciously.

So we've created systems - spoken language, writing, science, technology - that can externalise, collect, and preserve the best of our learning, so the species as a whole, including average humans, can get the benefit of it.

This works fine as far as it does, but our political and economic systems are still founded on animal logic (i.e. personal advantage no matter the cumulative long-term cost) instead of conscious collective intelligence amplification and species optimisation.

Based on this, AI won't get far unless someone invents an analogous way to externalise and abstract learning, so different AI projects can bootstrap each other, and there are common systems for abstraction and generalisation.

Unsupervised learning isn't going to be all that interesting unless it can do that. It'll be a fun toy for certain classes of problems, but it won't be anywhere close to an AGI.

An interesting question, but if we make a program that makes wry jokes about what it means to be alive or argues persuasively about its own freedoms, the world will notice.

You can define 'intelligent life' however you want. No matter how high you set the bar though, an algorithm that reproduces the intelligence of the average human and can be run in parallel at speeds unhinged by biological constraints would provide immense power, wouldn't it?

I think I'm more scared of creating an intelligent life form that isn't modeled after us. Maybe an irrational fear, but one I have regardless.

Of course you need some objective function and feedback about how well you're doing wrt that. In RL this is true too. The feedback needed for backpropagation might be, for each action you take in life, what your chance of death is right after that action

That's a different problem. Some things are hard wired by biology, but individual organisms learn during their lives in spite of vague feedback. That's what we're trying to replicate.

Having just raised a human from newborn to walking, talking kid, I have to say that her life so far has been chock full of oracles guiding her learning.

Long before language, babies spend hours every day just interacting with the physical environment, which is "true": the toy is here, not there. The body weighs a certain amount. If you move the head this way, you'll roll over. If you move it that way, no rolling. Etc.

And parents provide a different kind of guidance. We react differently to different sounds, actions, movements of child. Even if these are not "truth" philosophically, they represent for the child emotional truth.

I'm just a guy who reads things about AI on the Internet, but I have to say it sometimes seems to me like AI folks have funny ideas of what "intelligence" is, and how it develops. The description of "unsupervised" learning above doesn't comport with any human experience I recognize. IME, humans receive constant immediate feedback on everything we do in life.

What you've described for unsupervised learning is really closer to reinforcement learning, where you get rewards without any information. True unsupervised learning is more like "figure out the clusters of these n-dimensional vectors".

Well your teachers, parents, society tell you what is right and wrong so we really aren't learning unsupervised.

Also things like pain, hunger and pleasure also are for the most part generated outside of the "learning" parts of the brain, so similar to a pre-defined utility function/oracle.

Yes a 1yo child has a good way of getting feedback from the environment when learning to walk. They fall and it hurts !

But who teaches the previous generations and how? Chicken or egg first?

I think the origin is communication and memory. When a being (X) signals to another being (Y) the origins of its pain (signal of increased death probability) and X and Y can remember this cause, they can begin to teach.

Very little of what human's learn is done "unsupervised".

This is codifying "instinct" which presumably has already been done at least once (via DNA in some species).

Pain and pleasure are the oracles. Maybe not at every moment, but within a given lag.

Well, the algorithm is death and reproduction, isn't it?

If the output of your networks performs well enough when it confronts the real world, you get to reproduce before you die, and some of the essence of your network gets to live for another generation.

If not, you die, and your network doesn't get to contribute to the next generation.

>> Most of the time for biological intelligence there isn't an oracle describing the truth at every moment of life to judge actions/decisions against.

Is it possible that your conscious self is the oracle that provides feedback to subconscious networks?

No, why would "your conscious self" know any of the answers? The problems biological organisms face aren't even well defined enough to have "answers" generally.

People learn many things by being told a set of rules and/or objectives. Over time they can self-evaluate their performance against those rules and objective.

I suppose an animal may have a need - it feel hungry. If it can satisfy that need then it should reinforce whatever it did to get not-hungry. I suppose that doesn't require conscious awareness. But at some level of complexity we have to be able to plan more complex actions (hunting, tracking) and be able to self-evaluate performance on the sub-tasks used to get food.

Solomonoff induction captures the algorithm, but it cannot be implemented. Perhaps that means the mind is non algorithmic.

Biological brains try to predict the immediate future: what they will sense next. That's the ground truth.

I wager it has something to do with prediction and rules

Actually:

>>Most of the time for biological intelligence there isn't an oracle describing the truth at every moment of life

This is false for most of learning of human intelligence. Most of schooled human intelligence looks like supervised learning.

Innate intelligence and other animal intelligence looks like pre-trained models and some reinforcement learning.

"Schooled human intelligence"? What does that even mean?

I think you are severely underestimating humans, and probably animals too.

People have multiple modes of learning. You can try to get an A in school ("optimize" for the objective function), but you also might be skeptical of what the teacher tells you. Likewise you might be skeptical of what you read in a book.

You can also correct the teacher based on common sense and what you already know.

Currently, a computer can do no such thing. I think he's right that there is a fundamental difference.

Humans also learn from extremely small amounts of data. (Sometimes they learn badly, but they can change their minds, or not if they still manage to navigate the world.)

How >>Humans also learn from extremely small amounts of data. (Sometimes they learn badly, but they can change their minds, or not if they still manage to navigate the world.) ??

What does this mean? That 18 years of schooling is a "small amount of data"? Do you think you show 2 pictures of an airplane to a child and they know what an airplane is? Do you even have kids?

Except that by the age you reach school you already know amazing wealth of knowledge about the world and how it works, particularly the kind of commonsense knowledge that AI has always had enormous trouble to acquire. See Moravec's paradox.

5 year olds "know amazing wealth of knowledge about the world and how it works"?? Que? How many 5 year olds do you know or interact with?

Your statement is exactly the incarnation of Moravec's paradox. You don't see the amazing things that a 5yr old (even a 2yr old) can do, exactly because these things are completely obvious to us. They become much less obvious once you want your robot to be able to do these same things. At that point these things become horribly difficult.

You're severely underestimating children (or over-estimating the current state of AI). 5 year-olds have better natural language skills than any AI in existence, better motor and navigation skills than the most advanced DARPA robot, better mood recognition capabilities, etc.

You appear to be thinking of just facts and figures. AI is not that.

When did I mention AI?

How many robots have you interacted with?

5-year old humans have picked up a lot. For instance, I knew about the dangers that hot surfaces, sharp objects, angry cats, and big sisters represented.

I had learned to walk, run, swim, ride a bicycle, and catch a ball.

Ahh yes who doesn't remember their days in school doing gradient descent and backprop on millions of samples of text describing the solar system.

Can we go with 'incomplete' rather than 'false'? :) One of the most interesting areas of study for me is early childhood motor learning which is a fascinating combination of unsupervised, reinforced, and imitation learning. There are even significant intrinsic knowledge (instinctual/reflexive/morphological) effects. Also I really like to encompass non-human learning when talking about this stuff to remind myself of all the work nature did before we got to cultural transmission of knowledge.

Please describe what "unsupervised learning" a child may be doing.

Learning to crawl. Learning to perceive depth. Learning to walk. Learning to pick objects. Learning to hold balance. Learning to hold balance while picking an object. Learning to handle liquids (such as food or drink). Learning to avoid dangerous situations. Learning to ride on a bike. Learning to see. Learning to interact with other agents. And so on and so on.

The list is almost endless. In all of those cases, if there is any supervision it is minimal.

Even before: learning to swallow in the womb. There's definitely no reinforcement from other humans involved in that, and it's fundamental that they learn it in the womb because it would be too late to learn it with milk.

That's been passed down over 3.5 billion years to form the DNA that creates that part of the brain in an infant. AKA pre-trained.

Not really, they develop it by sucking on their thumbs, drinking amniotic fluid, etc. You can see it develop, and you can see how premature babies have issues with it. Adding "breathe" to sucking and swallowing is even harder for preemies.

Starting to suck when they feel something on their lips _is_ a reflex though, it isn't learnt.

reflex = pre-trained?

More or less, yes. They usually disappear before the first year. Many later reappear in a more refined and voluntary way, which is learned rather than pre-trained.

All those things you described can be explained by having a pre-trained model built from DNA. Babies aren't a blank slate. Why would you even think they are? Do you think a baby learns to breath from blank-slate intelligence?

If what you are saying was true, then why all these skills require time to develop? If they are pre-programmed in the DNA, why are they not enabled on the day of birth? Clearly from evolutionary point of view an infant that is already able to walk and perceive depth would have vastly superior chances of survival? Why all the brain development, and much enhanced plasticity in the early stage of development? Why do neural receptive fields change rapidly in the first months of life and do depend a lot on the stimulus received?

Clearly there is stuff pre-programmed in the DNA and some animals e.g. are able to hold balance right after birth. But apparently most of this stuff in case of humans is instructions how to build a learning system, rather then encoding the complexity of the world in a strand of DNA. I grant you, the evolution did not have the time to encode ways to handle lego bricks and yet 2yr olds spontaneously acquire ability to manipulate these objects way beyond what any contemporary robot can do.

I hope it's clear that the answer lies somewhere between these two extremes. Most likely, there are instilled a number of prototypical reflexes that quickly make available the "right" decision for a given situation. Consider it that we don't know the objective function per se, but that we start in a place and with a gradient vector already pointed in such a way that we quickly reach the optimal location in action space with relatively few exploratory trials.

Yes, the walking movement is one of a newborn's 9 primitive reflexes. [1] However, developing the fine control skills required to balance while walking is a skill that babies learn, and it takes a long time to do so.

[1] https://en.m.wikipedia.org/wiki/Primitive_reflexes

> If what you are saying was true, then why all these skills require time to develop? If they are pre-programmed in the DNA, why are they not enabled on the day of birth?

I'd argue that we see clear examples of such in animals that do require those for immediate survival. We ought to think about it as a consequence of natural selection rather than a 'necessity' that is somehow intelligently pre-programmed for survival. It's far more likely that path to early survival traits were mutational (accidental) rather than supervised.

Language acquisition. Sure, parents will point at an object and say, "Ball", "Chair", "Food", etc. But they don't do this for the conjunctions and interjections and prepositions. Nor for most of the 10,000 words a child will learn in the first few years of life.

Children learn the rules of grammar long before they realize there are any rules at all. Tenses and plurals, etc.

It's amazing how little people on this site know or understand intelligence, artificial and human.

You would be surprised how much infants learn from listening around them and how long it takes to learn that language, and how a brain interprets language was formed through so many years of evolution to be highly tuned to learn language fast.

We're arguing past each other then. I have no doubt that evolution is the essential thing that made us capable of unsupervised learning. That's not the central point of the debate for me. What I'm talking about is the problem of replicating that ability (unsupervised learning) in AI systems. We don't have the luxury of billions of years of evolution to do it. So we need to engineer another way.

I understood your previous comments as diminishing the sheer amount of unsupervised learning we do as humans.

I think innate intelligence/ animal intelligence is developed by evolutionary algorithm. You don't need to do back prop for it. If the network is wrong, the animal just dies.

I would agree with this approach. In evolution, it would be the equivalent to a conditional lethal mutation. In humans and even non-humans, may behaviors are learned through a form of adaptive behavior that oftentimes becomes a form of abductive reasoning. What is learned is "good enough" to serve as ground truth until there is evidence to contradict.

What you call "evolutionary algorithm" is just a pre-trained model your DNA created. That took 3.5 billion years to get to this point.