Open-access content
Monday 4th July 2016
—
updated 5.50pm, Wednesday 29th April 2020
TANC top tips on AI

On 24 May, TANC (The Actuarial Network at Cass) had the pleasure of welcoming Dr Stuart Armstrong, a research fellow at the Future of Humanity Institute at the University of Oxford to speak on 'Artificial intelligence: risk and risk control. This event was sponsored by the IFoA and well attended, with more than 125 guests.
In his presentation, Dr Armstrong explained that short-term artificial intelligence (AI) is easier to predict, but long-term AI predictions are trickier. Long-term AI is full of extreme uncertainties and could lead to severe outcomes in either direction, good or bad.
Additionally, creating an AI with sophisticated cognitive capabilities is not a simple task - nearly every word used in defining a simple task would require definition, with consideration given to the interaction and competing interests between concepts. One of the most challenging aspects is that we simply do not know the shape of the AI that will come.
This was evidenced by certain past predictions which turned out to be untrue, and although we can see now how these events occurred, it is hard to see how people at the time could have foreseen such outcomes.
The event concluded with a lively Q&A session, leading to further discussions during the networking session. TANC's next event 'Smart Beta, Scrabble & Simian Index Strategies' will be held on 15 September.
For details, visit www.tanc-cass.co.uk or email [email protected] to join the TANC mailing list.
In his presentation, Dr Armstrong explained that short-term artificial intelligence (AI) is easier to predict, but long-term AI predictions are trickier. Long-term AI is full of extreme uncertainties and could lead to severe outcomes in either direction, good or bad.
Additionally, creating an AI with sophisticated cognitive capabilities is not a simple task - nearly every word used in defining a simple task would require definition, with consideration given to the interaction and competing interests between concepts. One of the most challenging aspects is that we simply do not know the shape of the AI that will come.
This was evidenced by certain past predictions which turned out to be untrue, and although we can see now how these events occurred, it is hard to see how people at the time could have foreseen such outcomes.
The event concluded with a lively Q&A session, leading to further discussions during the networking session. TANC's next event 'Smart Beta, Scrabble & Simian Index Strategies' will be held on 15 September.
For details, visit www.tanc-cass.co.uk or email [email protected] to join the TANC mailing list.