[ExI] Convergence

[Anders Sandberg]

On 24/10/2012 11:04, BillK wrote:
> What I see as the main problem in forecasting is forecasting society 
> behaviour.

The problem is that societal behaviour is not really forecastable. There 
are too few constraints outside slow demographic and institutional 
change, and the system changes in response to what you do. You certainly 
ought to consider social and cultural factors, but it is not anything 
that can be reliably forecast.

My standard example is twitter. If a time traveller had explained the 
concept to me and my classmates back in 1991 I think we could have 
programmed the system without too much effort. However, we would not 
have been able to make any prediction about what it would be used for. 
In fact, it would likely have been fairly useless given the small number 
of Internet users back then. Twitter gets its functionality from what 
social functionality people invent, and for that you need 1) lots of 
users, 2) some experience with social media, and 3) creativity. So the 
current role of twitter was entirely impossible to forecast, although 
*maybe* we would have been able to figure out the above 1-3 requirements 
and hence deduce that if it ever got big it would get big in the future 
when there would be more internet users. But most likely we would have 
just concluded it was a useless system.

> So, after the technical phase of planning, I would like to see a phase 
> of 'reality' planning added on. People asking whether the project is 
> politically feasible, supported by enough backing, not opposed by 
> strong opponents, and likely to receive continued support and funding 
> for the duration, taking into account possible wars, disasters, 
> famines, etc.

And then what? Many very successful systems have been successful without 
such planning (consider the Internet, where again as in the twitter 
example, the actual role became known only as it was invented). And if 
you conclude that your system would be opposed by group X, is that an 
argument against the system or just an argument for a marketing campaign 
aimed at group X?

In general humans are bad at long-term forecasting and planning. Having 
foresight is useful, but one should not overestimate our ability to be 
accurate. Adaptivity is better than overconfidence in any plan.

-- 
Anders Sandberg,
Future of Humanity Institute
Oxford Martin School
Faculty of Philosophy
Oxford University