Carboncopies–Realistic Routes to Substrate-Independent Minds
August 9, 2010 by Randal Koene, Suzanne Gildert
What might brains and minds look like in the future? It can be difficult to manage and organize ideas from many highly specialized fields of expertise that must necessarily converge to answer this intriguing question. Not only must one consider the areas of brain imaging, neuroscience, and cognitive psychology, but also artificial intelligence, nanotechnology, biotechnology, computational hardware architectures, and philosophy.
In the past, the transferal of minds into computer-based systems has been rather vaguely referred to as “uploading.” However, those hoping to advance this multidisciplinary field of research prefer to use the term Advancing Substrate Independent Minds (ASIM), to emphasize a more scientific, and less science-fiction approach to creating emulations of human brains in non-biological substrates. The term ASIM captures the fact that there are several ways in which hardware and software may be used to run algorithms that mimic the human brain, and that there are many different approaches that can be used to realize this end goal.
On May 22, 2010, carboncopies was born in an effort to unite the disparate areas of research contributing to ASIM. In the few months since the organization’s conception, we have opened our doors to welcome collaborating persons and organizations, and embarked upon a series of events and outreach activities to promote the growing field of ASIM. Carboncopies offers a networking platform and hub around which experts in the individual fields relevant to ASIM can gather and exchange ideas.
It also promotes these ideas and explains the motivation behind ASIM to a wider audience through events, and through the publication of literature both for a wide audience of interested non-specialists and by actively seeking specialist scientific peer review. Although historically there has been much philosophical and metaphysical debate about the subjects that underlie ASIM, one of the main goals of carboncopies is to prioritize action over speculation. Projects that take practical steps to discovering what may or may not work are therefore highly promoted, such as those highlighted above.
Advancing Substrate-Independent Minds 2010 (ASIM-2010) to be held after Singularity Summit
In addition to the virtual events, carboncopies is holding an official launching conference in real life: Advancing Substrate-Independent Minds (ASIM-2010) will take place at the Hyatt Regency in San Francisco on August 16 & 17, 2010, immediately after the Singularity Summit. A virtual link will also be provided for those who wish to attend remotely.
Please join us in person, or through virtual presence in Teleplace on August 16 and 17, 2010, for an action-oriented event aimed at advancing substrate-independent minds. Contact us to find out more about each option. Regarding the ASIM-2010 conference, all are welcome and there is no registration fee — although donations to offset our expenses are greatly appreciated. Please RSVP the organizers to secure your seat.
What is meant by “substrate-independence”?
The concept is nicely encapsulated by the Church-Turing thesis, which can be summarized in the following way:
Everything that is computable can be computed by a Turing machine.
When it comes to the brain and the mind, the strong neuroscientific consensus is that behavior and experience, phenomena correlated with what we consider the mind, emerge from biophysical functions that are adequately described in terms of classical physics. These processes (and in fact, even quantum physical processes) are computable. It follows that the mind is computable; our brains are machines. The Church-Turing thesis implies that one Turing machine can implement another.
Of course, this has already been implicitly assumed true by all those who work on the development of neural prostheses. Once the functionality of the original system carried out in one substrate can be emulated in the computational hardware of another, substrate-independence is achieved for those functions.
Remarkable as our brains are, it is clear that there are many limitations. We experience limited working memory, finite and unreliable long-term memory, and the inability to multitask effectively, and our sensory experience and comprehension are a small subset of what may be possible. Substrate-independence enables us to exceed those limitations and suitably adapt to novel environments, to explore other ways of thinking, and to experience virtual environments from a truly first-person perspective.
The conservative technological approach to accomplishing substrate-independence for the functions of the mind is known as whole brain emulation (WBE). With that approach, high-fidelity of emulation is based on the careful reimplementation of the structural connectivity and of the biophysical functions of every high-resolution component on a scale that comprises the whole brain. A mechanistic reimplementation of this structure at sufficient resolution will enable predictive computation of successive mental and behavioral states. A successful reimplementation of such mind functions may also be a useful step along the way to the development of artificial general intelligence (AGI).
Advances in neuroinformatics and neuroprostheses
Recent advances in the fields of neuroinformatics and neuroprostheses are of immediate importance to ASIM. Significant examples from neuroinformatics include the high-resolution, large-scale evidence-based modeling carried out by Henry Markram’s team in the Blue-Brain Project, and with Eugene Izhikevich’s large-scale model of the mammalian thalamo-cortical system. Meanwhile, Ted Berger has been pioneering in the field of cognitive neuroprosthetics, by developing a replacement chip for a section of hippocampal brain region CA3.
One of the most important fundamental questions in ASIM is that of how we identify the minimum set of data that needs to be obtained from a biological brain in order to maintain an adequate representation of a mind in an alternate substrate. In practice, this means that we need to select suitable spatial and temporal resolutions necessary for the data-acquisition, or “scanning” technique. In some cases, it is the averaged activity of large groups of neurons that provides relevant information. This is the case when we consider the motor control output of the brain.
But in other types of nervous system response, for example when carrying out certain visual and auditory sensory tasks, it is essential to identify precise spike timing of individual neurons. In fact, to identify and re-implement brain plasticity (the modification of functions of the mind through experience and learning), evidence suggests that it may even be necessary to register chemical processes at individual synaptic sites. This is a hotly debated area of research. However, successful models of human cognition will likely include algorithms that operate at many different scales.
There are at present several experimental approaches that emphasize structural analysis. An example would be the structural scanning carried out by Kenneth Hayworth using an “Automatic Tape-Collecting Lathe Ultramicrotome” (ATLUM), in the Lichtman Lab at Harvard. In this work, a brain sample is encapsulated in resin and sliced very thinly such that the individual neurons and synapses can be observed, and the 3D structure recovered using an automated microscope.
The complementary technique of functional analysis is also being explored, such as the investigation of multi-neuronal activity by Anil K. Seth and Peter Passaro. Here, the researchers correlate behavior and action of an organism with signals detected after placing its dissected but living nervous system on an array of electrodes. This work is currently being carried out using small batches of neurons from Lymnaea (snails).
This work and similar projects will hopefully find ways to answer fundamental questions regarding the necessary information needed to emulate small organisms. The work can then be extended by scaling the projects to include the collection of data from larger brains and nervous systems. Progress in these areas is likely to depend heavily on the development of new tools for high-resolution, large-scale neural recording. It is a daunting task when one considers that there are around 100 billion neurons in a human brain, and it may well be the case that we need entirely new technological advances to work on these scales.
However, there are promising ideas on the horizon to aid in such an undertaking. There have been recent advances in the field of optogenetic approaches to the analysis and stimulation of neural circuits. Additionally, the possibilities offered by synthetic biology and nanotechnology, while not yet the main focus of ASIM research, are however likely to play an important role in the next couple of decades.
With so many disciplines converging around the area of ASIM, it can be difficult for researchers to keep track of advancements outside of narrow fields, and for news about the potential breakthroughs surrounding these exciting technologies to be made accessible to everyone.
Carboncopies hopes to help keep track.