Artificial Memory

So fire up your neurotech engines, ladies and gentlemen. And as for the rest of us, presumably the contest winner will remember it for us wholesale.

H+: Let me start off with a broad general question. I interviewed Zack Lynch a few years ago – the executive director of the Neurosociety. He believes that neurological improvement and self-control will be the defining characteristic of human society in a decade or so, acing out even biotech. I wonder if you share this view. Will we see a neurological age?

PETE W. ESTEP: I absolutely share this view and Zack is a trustee of the Inner- Space Foundation to help make this vision a reality. But maximum benefit will only materialize on that timeline if we push hard on the accelerator. I started out in neuroscience research as an undergraduate (at Cornell) because I saw the importance and centrality of the field to both understanding and improving biology and behavior. I also sensed huge future potential for the integration of neuro with computer technology. When I moved on to get my Ph.D. [at Harvard] I was still excited about the prospects for a neurotech revolution a few years down the road, but I wanted to do more in silico biology and I sensed an impending revolution in genomics after I met and began to work with my doctoral adviser, George Church. So, I got into genome science because it was so hot and exciting and so many smart people from computer science, engineering, and various hard sciences were joining in, and George’s lab seemed like the place to be. I am still very excited about what is going on in genomics but I’ve segued back into neuro because I think the potential is even greater — probably far greater, especially for people already alive. The Internet and electronic devices have become pervasive and indispensable, and interfaces between us and these outboard intelligences will become increasingly powerful and direct. I think these changes will come steadily and will profoundly transform our lives, but maximum impact will only come if we alter the current research and development dynamic to produce those technologies with the greatest potential.

H+: Your project, as I understand it, is offering awards for uploading information to the brain, and downloading information from the brain. And the idea is a device, whether external or implanted, that allows one to retrieve information by thinking about it. It sounds like a first step to the sort of mind uploading envisioned by people like Hans Moravec and much copied in various science fiction scenarios. I’m trying to envision what a prize-winning project would do. Would this be a first baby step toward these visionary ideas or a “great leap forward?”

PWE: The InnerSpace Foundation is concerned primarily with challenges that lie within the visible technology horizon, which is getting shorter in some ways. The challenges of improving natural mental functions are very daunting, so we have focused on establishing basic two-way communication between the brain and prototype devices. Interfacing with nonbiological electronic devices is important because they have many advantages over brains and neurons in terms of speed, accuracy, and durability. Input of information into the brain by electronic means rather than just through our normal sensory channels can be called learning, even though it is a non-traditional form of learning, and outputting existing memory information to a device for later access is potentially an extremely powerful way of augmenting memory because it has essentially unlimited capacity and high fidelity.

Since it is difficult for us to imagine exactly how these things might be done best in several years’ time, we have decided to set up a prize-based competition for rewarding one or more teams who produce the most compelling breakthroughs that most clearly satisfy the prize guidelines.

We already know we don’t have to destroy or dismantle the brain to get enormous quantities of information out of it; I think we simply need to push forward technologies that allow for maximum information flow to and from the brain in a non-destructive manner. Therefore, procedures like those suggested by Moravec that require the brain to be destroyed or dismantled and reconstructed don’t appeal to me. The IF is committed to technologies that will move essential information to and from the brain, and allow it to be stored and backed up, but I don’t want to speculate much on “mind uploading,” which implies dynamic reanimation of downloaded and stored information. Nevertheless, there are many very serious and respectable people who contemplate and seek the development of such technologies. The IF is trying to get the world’s leading neuroengineering talent to give us baby-step technologies toward what we currently regard as the future great leap of exceeding or transcending our unwanted evolved limitations — whatever they might be — and I am a very strong advocate of this bioprogressive view.

H+: Do you see this program of neural achievement as running in parallel to ideas of developing smart AIs, potentially of greater-than-human intelligence, and could this – in some sense – be a step toward fostering hybridization between humans and advanced AI?

PWE: A long-term goal of the IF is to allow the maximum possible degree of direct human control over powerful outboard intelligences. Many extremely bright people have argued that self-improving AI could have catastrophic consequences for humanity unless we are an indispensable part of the overall equation. My view on the AI developmental timeline is pretty conventional. I think AI of this level is some way off, and might even be dependent upon improved human intelligence, but I see the logic of their argument.

It is interesting to contemplate the interdependent hybrid human–AI intelligence scenario I just mentioned. It is entirely possible that naturally evolved human intelligence is incapable of producing catastrophically (for us) self-improving outboard intelligences, and that both natural human intelligence and AI are largely incapable of producing dramatic increases in human intelligence through purely biological manipulations because of the constraints of neurons and neuron-based storage and “computation.” However, when we consider that both abiotic and biotic storage and computational devices have their own strengths and weaknesses, it is easy to envision hybrids that tap the advantages of each and have characteristics superior to either alone. As one simple example of the comparative advantage of abiotic storage, my (inexpensive and old) 1 gigabyte keychain flash drive can store about a thousand 400-page books. And in less than a decade, a 1 terabyte (TB) keychain storage should be inexpensive and common. People will be able to store the equivalent of about a million books of text on their 1 TB keychain, and using standard and simple protocols retrieval is essentially error-free and extremely fast. 1 TB is also equivalent to about a million minutes of CD-quality music, a million photos from a typical 3 megapixel camera, or 140 days of continuous video (5 MB/minute bitrate, which is about YouTube or better).

Each of us should probably ask ourselves if we could store all information that is essential and important to us on a single such device how we might make real use of that potential. I think when we seriously reflect on such questions we begin to really see some of our inherent biological limitations. The harsh reality is this: the human brain is a magnificent and mysterious collection of abilities, but for fast and accurate storage and retrieval of important information, even a humble keychain flash drive has overtaken us. But I am extremely excited that — for the first time in history — we can envision using such technologies to augment the brain’s natural limitations.

H+: You’re focusing on memory rather than. say. perceptual intelligence or happiness, mainly because it’s measurable.

Welcome to the PRESENCCIA project E.U

This Integrated Project will undertake a Research Programme that has as its major goal the delivery of presence in wide area distributed mixed reality environments.

The environment will include a physical installation that people can visit both physically and virtually. The installation will be the embodiment of an artificial intelligent entity that understands and learns from its interaction with people. People who inhabit the installation will at any one time be physically there, virtually there but remote, or entirely virtual beings with their own goals and capabilities for interacting with one another and with embodiments of real people.

Specific subclasses of the installation will be used for the construction of a number of application scenarios, such as a persistent virtual community that embodies the project itself.

The core methodology will be to achieve this through the identification, understanding and exploitation of cerebral mechanisms for presence in conjunction with advances in the underlying technology for mixed reality display and interaction, with special attention to the interaction between people, and also between people and virtual people. Such cerebral mechanisms will be the basis for a core aspect of the IP which is the exploitation of brain-computer interfaces.

Processes within the environments adapt and correlate with the behaviour and state of people, and in addition people are able to effect changes within the environment through thought as well as through motor actions.

Magnus Olsson