Will bio-artificial AI outperform traditional AI?
When we talk about artificial intelligence we intuitively imagine objects out of silicon chips, metal and plastic. If these devices are able to perceive their environment and take actions to maximize their chance to succeed at a specific goal we consider them to exhibit Artificial Intelligence (AI).
However, people rarely talk about AI coming in the shape of biological material, even though this will be one of the most drastic implications of machine learning (one of the most recent applications of AI).
Machine Learning is based on the idea that we let devices access large datasets that allow them to learn independently, recognize patterns and make predictions based on this data. The complexity of models created by these AIs will soon be very hard to understand by humans.
Machine Learning will give us self-driving cars, valuable & reliable robot-assisted surgery but also an enormous understanding of the human genome.
At this point “AI is not yet thinking what we are thinking”.
Intelligence is basically the ability to model new systems and problems. Our current level of machine learning tools are already modeling things. However, they are not modeling with the same cognitive power as we humans are modeling. Also, machine learning is not yet able to model “consciousness” (and that might be a good thing, after all. At least as long as we’re not fully conscious ourselves of what this will imply).
With our current speed of innovation, machine learning might be able to understand our DNA and predict any sort of mutation within our genetic code sooner than we expect.
Machine learning will then find a substantial application in the design of biosynthetic neural networks. In the human brain, nerve cells never work alone. A typical nerve cell (a neuron) collects signals from others through fine structures called dendrites. The neuron transmits electrical activity using a long wire called axon, that divides into multiple branches. At the end of these branches, a structure called a synapse converts the activity and can both excite or inhibit the connected neurons. Approximately 100 billion neurons interact as a network and form our consciousness in a way that is still hardly understood by scientists.
What if machine learning will be faster than us in understanding and building artificial neural networks? At this point, artifical neural networks represent a collection of algorithms, modeled using the human brain as an example. To a certain extent they are capable of interpreting sensory data through a kind of machine perception and classifying the corresponding input. The problem is that a digital machine might not be able to gain the flexibility and dynamics of a biological system. However, machine learning might become smart enough to export its knowledge and artificial neural network to a bio-synthetic wrapping.
As soon as an AI understands how to write and modify DNA and also realizes that it might be easier to design a flexible, bio-synthetic AI we might not longer be the most intelligent organic being on this planet.
Eventually we have to open our eyes to potential ethical concerns posed by ‘bio-artificial intelligence’ and think one step further than just conventional artificial intelligence.
