Not a member yet? Register for full benefits!

Co-Processors for the Mind: New Directions in AI

The image above is from the film The Stepford Wives (2004). That film may be fictional, but it represents an angle that we are now pursuing in our world. Merging man and AI, or more specifically creating AI which can be bonded into the organic brain, such that the two work in harmony to produce something new. In many ways the plot of that film comes back to mind, as what is neing attempted, sans usage intent, is exactly what is portrayed in the film.

Headed by the US, the new direction in AI work has been awarded a $5,000,000 initial budget and a five year scope. The name of this new project is MMP, or the Mind Machine Project. An apt name really, since whilst mind machine interfaces are part of it, the true goal is to merge mind and machine together. This has a lot of useful applications of course, and al ot of truly terrifying ones as well. Thankfully the latter are limited by the primitive nature of our interfacing technologies.

Group members span all fifty years, and so have the wide variety of views and familiarity in working with AI through all half century of failure to create a true artificial general intelligence. Thus, the team knows what avenues failed in the past, intimately, whilst still having enough new blood to hopefully avoid being stuck in the same old ways. Researcher Newton Howard stands at the head of the pack, backed financially by Intel. Then there is world renowned AI researcher Marvin Minsky directly behind him, part of the field since day one. Ford Professor of Engineering Patrick Winston of the Computer Science and Artificial Intelligence Laboratory also weighs in, a researcher who started just ten years after Minsky. Gershenfeld started ten years later again. Ed Boyden, a Media Lab assistant professor and leader of the Synthetic Neurobiology Group, was a student of Gershenfeld, and the count rallies off with current students at MIT and a handful of other universities, just starting to work professionally themselves.

MMP is part of an attept to revitalise the AI field, and ignore the half decade of failed attempts by integrating the power of a human mind into the AI. It is a noble attempt, and is currently backed by a little over two dozen AI researchers, including some of the pioneers. They intend to create intelligent machines, no matter the means that has to be gone through to get there.

Neil Gershenfeld, one of the leaders of MMP and director of MIT?s Center for Bits and Atoms, spoke about the project. He says that part of MMP's purpose is to "revisit fundamental assumptions" in all of the areas encompassed by the field of AI, including the nature of the mind and of memory, and how intelligence can be manifested in physical form. He and his fellows intend to rewind the clock, going back 30 years to the initial glory days of the young field of AI, revisit ideas that were dismissed out of hand because they contradicted those fundamental assumptions. Utilising the breakthroughs in understanding the function of the human mind that have occured in those same 50 years, the plan is to integrate much if not all of the basic struccture of a human mind into the new AI models; to give something of an ecology, a mix of models based on the structure of one of nature's more complex minds, and integrate that at a basic fundamental level, into the AI model.

So, in many ways MMP seeks to bridge the ever-widening gulf between AI work and neurostudies; bringing both back almost under the same banner again, working the two fields off of one another holistically. Modelling thought, something which has traditionally been beyond AI researchers, is all but done already when you look at the human brain. So, take that model as-is, simplify it, and look for the basics. As a profound side effect of course, any successes there will tie directly back into neuroscience, helping us figure out what to look for in the human brain as well.

The team don't think small, in any shape or form. Minsky is even looking to supplantthe Turing test with one of his own creation, and one which he hopes MMP will best: A test of sophistication that would seek to assess not just whether a machine could be said to be intelligent and capable of holding a rational conversation - as the Turing test and Loebner prize test for, but going beyond that, to a machine that can understand, be conscious in it's own right.

This is where the human brain comes in, in a more literal sense than just modelling off of it. One of the key projects the group hope to produce is a form of neuroprosthetic, which harks straight back to that film image atthe start ofthis article. Exactly as portrayedthere, this would be a literal co-processor for the human brain. Linked into it, taking input from the brain, processing it internally and feeding the results back. Truly the nightmare scenario of every crank out there. The plan for this first generation is no small feat: They hope for an AI system on a network of chips, implanted into numerous areas of the brain, in order to monitor that person's activities.

By activities, meaning their body position, whether there is auditory information coming in, what the person appears to be trying to do - like stand, sit, or raise a hand to shake. Not activities such as where they go, who they meet with, what items they buy as the less technologically proficient might assume. That sort of thing is so far beyond our capabilities, it is laughaable in the extreme. No, the idea is an implant that will monitor what a person is trying to do, so that it can detect for example, if an attempt at standing is going to result in a fall, before the fall occurs. The obvious targets beingthose suffering from a form of cognitive damage, such as Alzheimers or a stroke.

The goal is noble, but on a five-year timescale? It seems doubtful at best. Still, with such minds on the case, it is at least within the realm of possibility, and the hardware side is within our current techniocal capability - just. Even if the attempt fails, there is no doubt the vdata collected will greatly aid attempts to try again, in the near future.



Dictionary: Artificial General Intelligence

Dictionary: Whole Brain Emulation

Podcast: The Brain Chip - A Prison For Your Mind

The Stepford Wives (2004)


The Mind Machine Project

Rethinking artificial intelligence

MIT Plans to Rebuild Artificial Intelligence from the Ground Up

Staff Comments


Untitled Document .