Comments on: Does AI Need Guts to Get to the Singularity?

By: Malcolm MacIver

Malcolm MacIver — Mon, 21 Feb 2011 01:09:44 +0000

Brian – well put. The paper suggests there may be more too it though too. Messing with the development of the brain, in an area which sets how risky a behavior an animal picks, suggests it is not just the source of drive, but also part of the enormously complex task of adjusting phenotypic expression to match the unique environmental circumstances of each organism.

By: Brian Too

Brian Too — Sat, 05 Feb 2011 22:55:03 +0000

I would argue that the role of the gut in biological intelligence is as a source of a positive drive. Specifically, the need to eat, or acquire energy.

AI researchers attempting to produce machines with behaviour often find that they need a system of competing and interacting positive and negative drives. “Seek such-and-such, it’s good”. “Avoid so-and-so, it’s bad”.

If you do not do so, you wind up with an AI that quickly goes to one place or state and then simply stays there indefinitely. The machine immediately optimizes it’s state and, because it’s rule set is too simple, there’s nothing else to do.

By: Malcolm MacIver

Malcolm MacIver — Thu, 03 Feb 2011 14:59:47 +0000

They are zoomin’ each other. The original study is showing that how risky a mouse is in its environment is modulated by gut microbes. The general nature of the effect of environmental conditions on the development of nervous systems seems to be modulatory, establishing setpoints in behavioral reactivity that are well-tuned to the specific environmental conditions the animal is developing within. It’s tempting to speculate that the reason germ-free mice are more prone to risky behavior is that a reduced gut microbe load might arise due to scarcity of food supplies, so behaving in a more risky manner is necessary for survival.

For any given behaving agent with appetites, a proper way to set priority of pursuit of those appetites against the potential risk to survival is needed. If you buy the suggestion that the gut microbes are participating in establishing these priorities, we have this link at least. This would then input into planning systems that assess different potential futures to a goal, and pick one with the right mix of risk and payoff. At least prior to action, this would qualify as influencing an off-line aspect of intelligence.

By: Rob Wilson

Rob Wilson — Thu, 03 Feb 2011 13:33:06 +0000

I agree about the depth of the connection between the control of movement or action, braininess, and gutness. Adopting a view whereby perception and action are tied together has, I think, paid dividends for the understanding of vision, and the incorporation of the gut into the picture here is especially interesting. The same is likely true for intelligence more generally, especially in the realm of the dreams of AI. But since there are special reasons to think that perception and action are tied together because action is what perception is for, I’d be curious to know why you might expect this to carry over to so-called “off-line” aspects of intelligence, and whether there is anything out there on the gut in non-visual contexts.

So a brain is really the gut’s way to get fed. And a gut is the brain’s way to acquire its own kind of food via the blood supply. Who’s zoomin’ who?