Michael Recce, Department of Anatomy, UCL

How does the brain represent space
Computer with radio transmitter on top. Mirror on side shinging a laser across a camera image which reads the distance from the wall calculating wheter to turn or not.

Temporal lobe hippocamus supposed to be involved in memory.

Middle ages and the brain - cerebral spinal fluid. Hollows full of fluid which was how brain worked through transfer. Leonardo - visualised

DEscartes - inspired by water driven robots in gardens in Paris. Pineal gland moves liquid back and forth hydraullically. Brain as gland 1662.

History of brain - is there any part which has a local function.
Phineas Gage railway work - rod entered through rim of his eye and out through head. Nothing happened except that his personality changed.

Different parts of the brain have different structure and function. Signals across brain eg to produce speech. Scanners can be used to detect activity dependent upon function. Reductionist idea of Central Nervous System through to molecules in neuroscience 40 years ago. Sensory processing detectable, but consciousness is not. If one loses some braincells people generally do not forget a particular memory. Cannot reduce things to cells but can to locality - each cell is not really doing one thing. Like a hologram - if you cut it in half you get the whole image but it is less resolved.

Penfield 1963 tried to establish activity in different parts of brain eg epilepsy brain parts have to be removed. You can feel no pain in the brain - poking around in the brain triggers memory.

HM had hippocampus removed on both sides. IQ unaffected but could not form new memories. Could not recognise new doctor, can do same crossword everyday.
Motor neurones not affected.
Primed memory - tell you to do something it will be remembered.
Had problems negotiating space. Putting objects into imagined space was like writings.

Different ideas about processing space in the brain 1) maps -Stimulus - response. Does not work if off Route ) Notion of space through maps but change the routeor methodsomething changesO'Keefe. Amoebae can follow gradients but they can't read maps. Pathinagrating if lights go out by keeping track of distance from door. Robots only pathinagrate.
Humans swap back and forth using pathinagration or cue based localisation and maps. Hippocampus only involved in maps.
Priota Cortex (?) Used to build egocentric map. fovia you see moon in 60 pixels but from side only a third of accuity. There is no feeling of end. Have a sense of what wis behind even without direct sensory data - ego centric map.
Robots made using this notion of egocentric map. Sonar sensor on arneid the robot attempts to create these egocentric maps. Speculative reflection - sonar sensor beam does no t come straight back. Unaware of certain parts where the beam does not go. We can estimate how the robot and humans will function. Robot and we have to keep building new maps. The brain takes partial information and cna reconstruct the rest of the scene of film or environment.

Want the associative memory to fundction with robot. Use virtual reality how to deal with this. Left - lingusitic. Right - spatial. HM has no obvious linguistic problems. Amigdula at front of hippocampus - controls emotional response to stimuli. Japan amigdulepomy if responses out of control.

Shell Game Environment - subjects with part of hippocampus removed asked to find blue cube in an environment with red and green shells/spheres. Right hippocampus removed find it difficult to detect.

Red sign which keeps disappearing further into the woods. Can these patients pathinagrate. Then one has to find an object where the features of the wood change mapping.

New works on how robots move. Do not need lights to show robots cleaning a windscreen. Humans do need lights. Humans are sensor driven. Building robot arms which have vision and other sensors. Robots expensive because of machine parts. Made of rubber with hollow tubes filled with transmission fuludi. The rubber can bend in segments.
Rubbery rollers which emulate our sensors. They will stretch. Stretch receptors know the distance and can figure where the arms are in space. Humans can have their stretech receptors inhibited so that they do not know where theire arms are.

Humans use trial and error with distance to figure where they are. Inexpensive robots to pick oranges.

Neural networks. We can figure out what individual neurons can do. Modelling of neurons. Can use parallel computers. Machines faster than humans eg eye like film. We parallel process - machines look singly.

Consciousness and emergent behavious using parallel processing computers.

Imprecise robots - no feedback in engineering. Sonar sensor constantly changing unlike usual robots which equate move then equate. Combination of ever shifting information.

Penrose, quantum physice and consciousness. AI - symbolism, logic then function. That is his problem because things have to be symbolised.

neural approach to do with parameters and experimentation. Neurons work like that. Computers recognise patterns. Ted Honderich - how can the brain follow the laws of logic if it has to first follow the laws of maths physics and chemistry.