Nano AIs will make us healthier and smarter
Within 20 years tiny AIs, injected into our bodies, will be able seek out and repair faulty cells.
They will repair damaged tissue from the inside and even enhance us both physically and mentally.
This is the vision of John McNamara of IBM, given in written evidence to the UK House of Lords AI committee.
McNamara believes that these machines, a combination of nanotechnology and Artificial Intelligence, will help us fight disease and enhance our physical as well as cognitive abilities. They will also enable communication with external technology allowing us to control our environment solely with the power of thought or gestures.
But McNamara also has more down-to-earth ideas about the current uses of AI and Deep Learning.
He correctly points out that modern AI systems are often ‘black boxes’, we know what they are supposed to do but don’t know how they do it. These Machine Learning systems are so complex that they cannot be taken apart and analysed to see how they achieve their results. We cannot have absolutely certainty, then, that the results are, in fact, correct.
Vast amounts of data are used to train AIs and provide them with the information that they need to make decisions. But they might also hide misinformation that could affect the AIs performance. AIs cannot tell the difference between causation and correlation, so biases that exist in historical data may be (indeed, have been) incorporated into machine learning systems.
McNamara concludes that decisions that might affect a person’s health, well-being or legal status should be completely transparent and that even less critical decisions should not be left solely to a machine but need to be vetted by a human being.
How do you like your coffee?
In the next five years, we are told, we should expect to see AI in everyday objects and machines will learn from interacting with their users. A sad or happy face — or maybe an angry one — will give it feedback on whether it has done a good job, or not, and will allow it modify its future behaviour. We should expect a coffee maker to be able to recognise its customer and produce exactly the right brew for her. Transparency in this sort of machine decision is probably not so important. If you don’t like your coffee, just scowl at the machine and it will try harder next time.