Naked and Afraid AI Test

Posted on December 21, 2023 by James Cross

If you haven’t seen the TV show Naked and Afraid it puts two naked individuals, man and woman usually, in a challenging wilderness situation – savannah of Africa, Ecuadorian rain forest, for example – with a map and, a few basics like fire starter and machete. They have to figure out how to survive 21 days. The majority make it but in a decent number of cases one or both of contestants drop out. Depending on the environment, they could endure incessant rain, wild animals, thorns (they’re barefoot remember), incessant heat, freezing cold (even in the rain forest – I know this first hand)., and insects. They have to build shelter, hunt for food, and move to an extraction point on the last day.

That is TV, of course, but there have been a remarkable number of real life survival situations too. One hiker cut off his arm, I think, because it became caught under a boulder. Some young children have even survived in difficult conditions for days.

Over on Scott Aaronson’s blog there is an interesting conversation about predicting AI growth. In a minor part of the discussion I proposed a test for human-level AI.

Let’s put AI on the savannah of Africa with a robotic body and see if it can survive longer than 21 days (naked and afraid so to speak). It does have to do things that humans might do like navigate a tough terrain, find water, procure resources, etc with no oil changes or extra charge-ups. Its robotic body should have roughly the same physical capabilities and deficiencies as humans. It can’t have a shell that protects it when the rhino gores it, for example. But it can move just as fast as a human – in other words slower than almost every other animal on the savannah.

I would expect an AI approaching human level intelligence would survive at least as often as humans can survive the challenge.

Somebody objected that few humans can do this challenge. True, many on the TV show are trained survivalists. I also stipulate the AI can also be trained.

Another variation that occurred to me was to have the robot AI run off batteries and the AI was tasked with finding battery caches in order to survive. This is akin to hunting for food.

The idea is that the AI will have approximately or equivalently human challenges and vulnerabilities.

Twenty one days in the savannah tests vision, hearing object recognition, reasoning in unpredictable situations, ability to optimize with scarce resources, and more. It would be a test in a real environment similar to which we think humans evolved. The Turing test in contrast only requires a way with words, a Wikipedia knowledge of the world, and a facility for logic and math.

What do you think?

Posted in AI, Human Evolution | 6 Comments

Do We Misunderstand the Hard Problem?

Posted on December 11, 2023 by James Cross

Or, maybe the question I should be asking: have I been misunderstanding the hard problem?

I had occasion to go back and look at Chalmers’ famous paper. Here’s one place where he talks about the hard problem:

At the end of the day, the same criticism applies to any purely physical account of consciousness. For any physical process we specify there will be an unanswered question: Why should this process give rise to experience? Given any such process, it is conceptually coherent that it could be instantiated in the absence of experience. It follows that no mere account of the physical process will tell us why experience arises. The emergence of experience goes beyond what can be derived from physical theory.

https://consc.net/papers/facing.pdf

I think the way the majority of people, with have any understanding of the hard problem, understand it is much as I have understood it. It is the problem of explaining how consciousness arises from physicality. Typically joined to this understanding are discussions of qualia, the question of how physical processes can generate them, and some general “mystery” of consciousness.

Take a look at the last line again of the Chalmers’ quote. It seems to me that an understanding of why experience emerges by physical theory is not all that intractable. In fact, a reasonable explanation could be arrived at without explaining qualia at all. It is all a question of what would constitute physical theory explanation for the emergence of experience.Explaining why and how consciousness has emerged shouldn’t require a complete explanation for how it works; however, a more complete explanation of how it works might fall into Chalmers’ category of easy problems: ability to discriminate, categorize, and react to environmental stimuli; the integration of information by a cognitive system; the focus of attention, the deliberate control of behavior; and others.

A few observations:

1- Consciousness has emerged in biological organisms

2- Consciousness, if not a spandrel, performs some function of evolutionary value or it would not have evolved.

3- Intelligent motility is a good candidate for why consciousness might provide evolutionary value.

4- Intelligent motility requires memory, learning, and ability to predict based on memory and learning.

If we look at the complex of learning, intelligent motility, memory, and prediction (LIMMP), we have nearly a complete evolutionary explanation for why consciousness has emerged in physical organisms. Of course, there is one relatively large thing missing – the direct physical links between consciousness and the LIMMP complex – but we already have good scientific evidence that these links exist. No organism moves with intelligence while unconscious. No memories form or complex learning take place in an unconscious person. Decisions and predictions can be traced to brain processes that involve conscious experience in one or another.

Some will say that the missing physical links is exactly the problem needing explanation in the hard problem. But the problem of missing links is completely researchable by science. It is so researchable that it could actually be placed with the easy problems.To be clear, we don’t completely understand memory and learning. But we already have good reasons to understand these to involve physical processes in the brain since synaptic connectivity patterns change when learning takes. Fundamentally, we need to link the physical processes identified in prediction, learning, and memory to conscious experience.

In other words, what we have to discover is what the brain is doing physically to itself when it has conscious experience.We could derive the why and how consciousness from physical theory by successfully correlating physical change in the brain with conscious experience.No explanation of qualia is required and no explanation of the “mystery” is needed. The research topics are all closely related to Chalmers’ easy problems.

Some might say a robot or zombie could also learn, have memory, and exhibit intelligent motility; hence, consciousness isn’t needed and the hard problem remains. Yes, a robot could do all of those things but the point is irrelevant. The questions are why and how living organisms predict, learn, remember, and move intelligently. It is not any or all of those things are consciousness by themselves. It is that those abilities require consciousness in living organisms. That means that consciousness is itself physical and can modify the brain. We only need to discover how and when we do, since the brain is physical, we would have a physical theory.

So, tell me. Was I misunderstanding the hard problem? Or, am I misunderstanding it now?

Posted in Consciousness, Human Evolution, Intelligence | 21 Comments

Ends and Odds 2

Posted on November 17, 2023 by James Cross

This is the return of ends and odds.

New McFadden Site

Philosopher Eric, who has been a frequent commenter on this blog, has created a McFadden oriented website.

Consciousness as an Electromagnetic Field A site for exploring Johnjoe McFadden’s cemi field theory

He already has some posts up.

If you are unfamiliar with McFadden, you should know his core argument is effectively that the electromagnetic field generated by neurons is consciousness.

The more I read about neurons the more I discover new (to me) ways that they communicate. I read recently an article that says neurons emit photons and that possibly the photons could transmit information to other neurons. We have the well-known electrochemical communication. Then there is the possibilities that neurons also used electromagnetic fields to communicate with each other. If nature finds something useful, it will tend to use it, so I wouldn’t be surprised that neurons would use electromagnetic fields. Likely any or all of these methods could be related to consciousness, but it isn’t as clear whether one or a combination is more greatly involved.

Imaginative Rats

Rats are more imaginative than you might believe. Through some clever experimental methods researchers have found that rats navigate toward a goal only in brain activity. In other words, they are imagining themselves taking a path to a location different from where they are at to get a reward.

There are a lot of articles on the internet about this. Here is an explanation from one:

To explore whether rats can do the same, and unpick a possible mechanism, researchers employed a brain-machine interface in which electrodes were surgically implanted into the rats’ brains. The rats were then placed on a treadmill ball within a 360-degree immersive virtual reality (VR) arena, and presented with an on-screen goal to run towards.

As the rats moved, and the treadmill ball turned, the animal’s apparent location within the VR environment updated on screen – as if the rat was running through a real environment. When the rats reached the goal, they received a treat and the goal was moved within the VR environment. The process was then repeated.

During this initial phase, the team recorded the activity within the animals’ hippocampus. They then used a computer system to translate this neural activity to specific locations within the VR environment.

In the next step, the researchers decoupled the treadmill from the VR system. This meant the rats could not reach the goal by running on the treadmill. Instead, they could only use their brain activity to navigate through the VR environment.

By analysing the activity in rats’ hippocampus in real time during the task, the team were able to update the screen every 100 milliseconds with the animals’ current location in the VR environment, based on what was happening in their brain.

The results reveal the rats could indeed navigate to the goal using just their brain activity.

https://www.theguardian.com/science/2023/nov/02/rats-may-have-power-imagination-research

Perhaps this isn’t surprising. If one of the main roles of the brain is prediction, then the ability to imagine alternative outcomes and different ways to accomplish a task might be an inevitable consequence.This would extend to imagining steps involved in a task such as navigating to a reward.

Other research has show navigation in rats, perhaps most mammals, may be accomplished in grid cells in the hippocampus.

Mammals are able to navigate to hidden goal locations by direct routes that may traverse previously unvisited terrain. Empirical evidence suggests that this “vector navigation” relies on an internal representation of space provided by the hippocampal formation. The periodic spatial firing patterns of grid cells in the hippocampal formation offer a compact combinatorial code for location within large-scale space.

https://www.ncbi.nlm.nih.gov/pmc/articles/PMC4534384/

This is evidence that, in some brain functions, the geometric arrangement of neurons may be critical to conscious content. We might speculate that, if each neuron had its own unique electromagnetic signature, the geometric arrangement of multiple neurons might create an unique combined “picture” composed of the individual signature of each neuron.

Bats Too

Bats have grid cells too and may use the same grid cells for spatial and social navigation.

In short, the bats’ navigation system appeared to do double duty as a social map. The mammals weren’t just moving around their home — they were also using the exact same brain cells to track who was on the premises.

In that sense, Forli said, the hippocampus may be like a powerful graphics card in a computer, which can have many uses, from rendering graphics for video games to performing machine learning computations. The hippocampus may be great at particular kinds of computations and may have the ability to be modified or programmed by evolution.

https://www.quantamagazine.org/bats-use-the-same-brain-cells-to-map-physical-and-social-worlds-20231031

I’m not especially fond of the “computation” metaphor in the quote, because it becomes associated with digital computation in the mind of many. Whatever form of “computation” is occurring hardly seems like digital computation, especially if the geometric arrangement of computing elements is part of the “computation.”

Grid Cells, Memory, Hexagonal Patterns

Brain’s Positioning System Linked to Memory is a fascinating article in Quanta from a number of years ago. It probably would be a great one to update.

The article is about special neurons called place cells, which fire whenever an animal is in a certain location and grid cells, which are like a dead-reckoning system, telling the animal its location independent of external cues.

Apparently these type of neurons and patterns are widespread in mammalian brains, including those of humans. The firing patterns for grid cells seems to have a hexagonal geometric pattern.

The hexagonal pattern of grid cell activity is repeated all over nature, from honeycombs to the benzene ring to a tightly packed crate of oranges. It’s a highly efficient arrangement; bees use hexagons in their combs to minimize their use of wax. In the grid-cell system, the hexagon isn’t a physical object. Rather, it’s the organization of space that encodes information most efficiently. “It’s the most efficient way to compress data,” said Marianne Hafting Fyhn, a neuroscientist at the University of Oslo in Norway and a former student of the Mosers. Researchers aren’t sure why grid cells use hexagons, but the hexagonal organizing principle has attracted attention from computational biologists, who are trying to figure out how the grid is generated.

Apparently it is not simply physical space and location that is maintained with grid cells. The grid contains abstract properties such as the social map mentioned about bats earlier.

The article concludes:

Many researchers believe that memory and space are even more intricately linked. In a popular trick for remembering speeches, dating back to ancient Greece, the orator calls to mind a familiar path through a city and attaches a segment of the speech to each location along the path. This mnemonic may unwittingly exploit the fact that the hippocampus encodes both location information and autobiographical memories. “It just happens that space is a good way of organizing experiences,” Wilson said.

Posted in Consciousness, Electromagnetism | 6 Comments