SAS Delphi results – Bio-augmented Machines

Schematic showing the electronic pathway for the electrochemical detection of the analyte via (A) mediator-assisted biosensors and (B) mediator-free biosensors. Credit: Dhand C et al. Dovepress

Area 4 – Bio augmented Machines

The use of bio (carbon based and living cells) has been on the researchers’ bench for a while with experiment on merging neurons on chip to leverage on qualitatively different sort of computation.

The evolution of bio-interfaces will support a variety of interactions potentially opening the way to synergies between machines and living beings, including humans.

Q 4.1

Considering the rapid evolution of machines, will bio-augmented machines in 2050 still make sense? In other words, would a machine still have any advantage by leveraging a biological brain?

The general consensus is that bio-augmentation will provide an edge to machine, even considering technology progress. There seem to be in the observed time frame still advantages in coupling bio with machines, particularly in the area of sensing. In the more specific case of leveraging a biological brain this is conditioned to the availability of effective CBI and BCI. Provided technology in those areas will be sufficiently developed, and consensus is lower, there may be benefit in leveraging on a biological brain.

Interaction with biological brain may have a value from a machine point of view in the sense of achieving more effective interaction. The progress of technology and the understanding of brain processes (and of the physical infrastructures supporting such processes) will indeed result in machines that are better than natural brains. What might be good in a natural brain would have been copied and injected into a machine. So in the second half of this century there won’t be any advantage for a machine to exploit a natural brain. However, connectivity with brain will still be important because it will provide an advantage to the brain.

Q 4.2

Brain-Computer Interfaces will clearly progress significantly in the next 30 years. Would a brain be able to contribute to the processing of a machine?

Unanimous consensus on the possibility of a brain to contribute to the processing of a machine. AS previously indicated there is not generalised consensus that the BCI – CBI will have reach a point sufficient for a real symbioses brain machine.

The understanding of brain processes and of the structures supporting them will provide new insights in syntheses, abstraction, conceptualisation, intelligence and free will. These will be used, sometimes mimicked, in a machine and in this sense I see the contribution of the brain. On the other hand, a real time contribution, like having shared processing between brain and machine does not seem realistic, not because it couldn’t be done, rather because it will not be effective.

Q 4.3

Would the interaction between brain and machine become so seamless to give rise to symbiotic processing?

Unanimous consensus that in the long term this will be achieved. However, in the medium term the symbioses will need to be mediated by senses, since a direct brain computer interaction is unlikely to reach the effectiveness required.

A direct seamless connection between brain and machine is not yet in sight, although there are several attempts to do so and results have been achieved. The crucial issues are seamless and the extent of the symbiotic processing. Seamless will remain a challenge for many decades. The extent of the symbiotic relation will be growing over time but it will take several decades before reaching  the point of symbioses that (sometimes) we experience between two persons knowing each other very well. On more limited extent, like a paralytic interacting with an exoskeleton to execute a a variety of actions, symbioses will be achieved in the next decade and it will keep expanding.

Q 4.4

In the case of a “brain” participating in a decision process with a machine, what accountability and responsibility issues would emerge?

The general consensus is that accountability will remain on the human side. However the scenario may get much more complicated considering the variety of human players involved in addition to the human in symbioses with the machine. The designer, developer, maintainer of the machine (and related software) will be involved in the sharing of responsibility.

The decision point in a symbiotic relationship cannot be tied to a single component as decision arises from the ensemble. Also, decision support fragments offered by the various components are likely to be heavily influenced by the other components and the ongoing interactions.

In general there will be a need to set up a new framework of accountability and responsibility for the whole ensemble, may be similarly to what has been addressed in the past as collective responsibility of a tribe, a nation.

Q 4.5

Machines mimicking life, and more specifically the brain, will become available in the next two decades (one aim of the Human Brain Project is to understand how the brain works in order to leverage that understanding by replicating it in machines)—and if so, will they be more performant?

The majority of experts does not consider likely that machines will reach the point of replicating a brain within the next 20 years (although a minority indicated that will be achieved).
There is a more broader consensus on the fact that processing performance au pair with a human brain might be achieved in the next twenty years, and possibly exceeded, however it seems unlikely that such a processing capacity will be possible within the energy budget of a human brain.

A brain in a purely technical sense is not very effective in many areas, although it is amazingly effective in taking what from a survivability standpoint is a working solution. Machine will keep increasing their analytical performance (already well beyond the human one) and will be copying, and refining, synthetic capability and guessing that are in general a trait of human brain. A single machine will probably not be structured in a way to take chances but clusters, swarms of machines and for sure symbiotic machines will take chances, and will become better in taking them.

Q 4.6

Life “information processing” may not be more “powerful” than silicon/quantum information processing, but it might remain more “energy efficient” remain more effective (an example is the flight control of flies that is based on some 5,000 neurons whilst a flight control of an airplane requires millions of code instructions). Could this be a reason to continue seeking for a bio-computer integration?

The experts were split almost evenly part in support to a bio-computer integration for energy budget reasons and part stating that will not be the main driver. Rather the possibility for a machine to crunch huge amount of data and for the brain to have a feeling on those data might be the main motivation.

AGI/ASI will narrow the gap, while fully-realized artificial brains could operate as a biological brain at much higher speeds and reliability.

Q 4.7

Would machines that have to interact in symbioses with humans, like a robotic exoskeleton, benefit from a symbiotic cooperation with the brain?

All experts agree that in this kind of situation a symbiotic relation brain-machine would be highly desirable. Notice that in these situations the symbioses can be restricted to certain aspects, hence might be more feasible than a more general symbioses.

A symbiotic relation involving the brain, like in the example of an exoskeleton supporting a paralytic person, can make the relation seamless and this is a great step forward.

Q 4.8

Would aspects like affection, emotion and feelings be better managed by machines interacting with humans rather than working on their own and simulating them?

The majority of experts indicated that the area of emotion in a broad sense is better managed by machine able to interact with humans rather than managed through simulation.
However, it has been noted that in practice the capability of humans to manage emotions, and related mental states, is not necessarily adequate in many situations, e.g. under stress, so that a machine can be more predictable in these areas, which in certain situations may yield a better result.

Most humans struggle to identify and respond to their own emotions in a rational manner, so it’s hard to see how they could convey useful information to a machine. Equally, as we have already seen with various infamous chatbots, given a sloppy set of human-driven rules, AI doesn’t always evolve very well and this should be taken into account.

Humans display a significant slate of emotions and in general I would say that a machine interacting with humans, being exposed and sensitive to those nuances and capable of learning will be better off that a machine modelling emotions.

About Roberto Saracco

Roberto Saracco fell in love with technology and its implications long time ago. His background is in math and computer science. Until April 2017 he led the EIT Digital Italian Node and then was head of the Industrial Doctoral School of EIT Digital up to September 2018. Previously, up to December 2011 he was the Director of the Telecom Italia Future Centre in Venice, looking at the interplay of technology evolution, economics and society. At the turn of the century he led a World Bank-Infodev project to stimulate entrepreneurship in Latin America. He is a senior member of IEEE where he leads the New Initiative Committee and co-chairs the Digital Reality Initiative. He is a member of the IEEE in 2050 Ad Hoc Committee. He teaches a Master course on Technology Forecasting and Market impact at the University of Trento. He has published over 100 papers in journals and magazines and 14 books.