Two things are infinite: The universe and human stupidity; and I’m not sure about the universe.
Albert Einstein
Einstein’s genius was instrumental in the creation of the atomic bomb. When he became aware that his ideas were being used to develop the bomb, he warned FDR of the potential consequences[1]The Making of the Atomic Bomb. Einstein was right, and his predictions were prescient. His concern: What will happen when the weapon falls into the wrong hands? North Korea’s ambitions are proof.
Warnings of the dangers from artificial intelligence (AI) are being painted as posing a similar existential threat[2]AI ‘godfather’ Geoffrey Hinton warns of dangers as he quits Google.
There is a great deal of misunderstanding about AI. That’s to some degree because even the experts often don’t agree on what it really is.
AI in its truest sense is self learning. It identifies errors, and it fixes them, thus avoiding repetition. That’s good. If only humans would learn from their mistakes.
The problem now is the objectives that the AI is instructed to achieve, and what is identified as an error.
Let’s say the objective is maximum economic growth. The AI calculates that maximum economic growth is achievable with 7.8 billion people on the planet, so it puts into effect a way of killing 200 million. Not a desirable outcome. So we need to cover everything in the objectives, and set very clear contstraints. We tell the AI: “Killing people is not good”.
But what about income inequality.
One thing’s sure and nothing’s surer. The rich get richer and the poor get – children.
F. Scott Fitzgerald
The Great Hunger in Ireland is catastrophic proof of the lack of value placed on the life of poor people by the rich. That discrimination persists, albeit on a more subtle level. AI can eliminate it, if it is set as an objective.
If we understand the potential threats, then we are already better off. We define objectives, and we prioritise those that conflict with one another, with AI alerting us to objectives conflict.
There is of course the concern that AI will create massive unemployment, as mechanisation initially did during the Industrial Revolution. No doubt it will. We should look at which jobs can and will be replaced by AI to see how we will be better off.
Lawyers: at present, we need them to understand the law, produce contracts, and represent us in court. AI has the complete library of law, and the ability to draft contracts that meet the requirements of the parties. We won’t need them in court, because AI will replace the judges.
Judges: judges must be unbiased, have a comprehensive knowledge of the law, and must make decisions based on the evidence presented to them. AI has all of those attibutes, and since there is no entertainment or art in the presentation of the evidence, objectivity of the decision is guaranteed.
Doctors: we need doctors to help us get well, and to guide us to stay healthy. They need a comprehensive understanding of how the human body and mind work and an encyclopaedic knowledge of medical solutions. They also must have a good bedside manner. And then there is surgery.
AI has knowledge and understanding covered. It can evaluate the symptoms and calculate outcomes better than the leading experts. Surgery is a matter of building the robots. Bedside manner requires empathy, so perhaps there is still a little space for doctors, but the machines will be making the life and death decisions.
Power corrupts; absolute power corrupts absolutely
Lord Acton
Politicians: politicians create laws, and fulfil the needs of the electorate, and set policy to achieve the country’s goals. Laws are created in search of specific objectives, in the context of constraints.
Novel solutions require imagination, and this is something that AI lacks. But, that novelty can come from any source, and does not require politicians.
With AI there are no promises that may or may not become priorities. We will be far better off when AI replaces politicians.
The machines don’t hold power, and are not elected, so there is no risk of an AI coup d’état.
Setting policy is setting the rules given to bureaucrats to achieve the objectives. So, do we need bureaucrats?
Bureaucrats: AI will be one machine replacing another, without the needs for salary or annual leave. We’re better off again.
Plumbers: No, we will still need them.
Electricians: No we will still need them.
So that is the future, We will spend the leisure hours we’ve gained through artificial intelligence waiting for plumbers and electricians to keep their appointments.
References