Why we may elect our new AI overlords
If you follow conversations around AI ethics these days, you may be forgiven for believing that either:
- Some sort of AI dictator may be imposed by some shadowy new-world order type organisation, or
- Due to improper human-compatible programming, an advanced artificial general intelligence will turn us all into paperclips for the paperclip god.
Whilst I can’t fully refute either scenario, I can speculate on the potential for near-future political applications of AI which have the potential to be both disruptive and unintuitively desirable.
Automatic, near-instant, ubiquitous fact checking
UK fact-checking organisation Full Fact recently released a report that covered the state of various fact checking technologies in use and in development. It covered areas like overall existing semantic data standards in use and their challenges, and trends seem to be that natural language and contextual language recognition is going from strength to strength.
The techniques in play include quickly responding to previously debunked claims, data mining statistics on demand and drawing confidence-based correlations from less reliable but plentiful data like news and social media reports. The convergent goal of these various projects appears to be moving fact checking from a post-hoc arm-chair analyst world, to a force that for example, in the middle of a debate or news report, is avalible to instantly challenge, rate or question any given factual statement within seconds of it being said. This is a game changer.
We’re not there yet, but we’re not too far away either. In a future where people can run their fact check bot (or preferred bot as the case may be) along side every news report, political statement or even entertainment article they consume, and it immediately and reliably gives feedback as they go along, this becomes like having an expert critic reading along with you on any given subject. In the short term I would imagine it would increase existing trends such as complaints to the broadcasters or publishers about inaccuracies and cause some media to lose and others to gain followers. But more importantly, as this power asserted itself, many people will choose to run these bots for the convenience and assurances they give, and in time what gets said said will start to change. Politicians may be forced to ‘fact-check proof’ their speeches to a much higher degree, or where detailed figures were not avalible, to better qualify statements with ‘around such and such’ or ‘I believe so and so’ in order to avoid being instantly called out.
I’m well aware that a single self-selected arbiter of the truth is dangerous, as a recent conversation with an acquaintance and arch-Libertarian Mike Lorry showed on the matter. Persuading him of the utility of holding public officials accountable through such technology, rather than exclusive use by the existing incumbents to control discourse , can be done via mixing of data sources, preferences with confidence weightings and I even managed to bring an initial sceptic around to this point of view
Gamble on the future, and win
Prediction markets have been significantly in vogue in rationalist circles for a while now, the idea that stock market-like betting pools for a whole range of ideas could not only incentivise empiricism in traditionally loose predictions, but also create significant financial motivations and secondary markets in this space. Let me explain. Currently a politician can be elected on the completely nebulous assertions that they will ‘institute fairness’ or ‘take back control’ . When said politicians manages to get elected on the basis of this rhetoric, they have barely any accountability due to the insubstantial nature of their ‘prediction’. Sure, some political commentators may call them out this vagueness, but often to their supporters, happy feelings were promised, happy feelings were (at least initially) delivered. What if politicians could actually put their money where their mouth was more often with predictions? A number of things begin to change…
“A Labour government would be a disaster for the economy” becomes “I predict that if Labour win the general election with a majority in the next 4 years, UK GDP will experience negative growth” If the politician or commentator is confident on this point, they could bet on it, and people who disagreed could bet against. Of course as people bet for and against, the overall odds emerge and people start changing their bets. Perhaps an economics think tank weighs in publicly and places a bet on one side, causing a number of other high profile people to change their bet, causing a cascade effect. Perhaps this is then overblown and a market correction follows.
Run something like this over 4 years and watch how it changes, and observe people try and game the system. Suddenly people are clamouring for the latest trade report to know where they should move their money, they are desperate to hear the upcoming manifesto scope to see if the policies are well researched or not and so on. You’ve gone and created a marketplace for political ideas.
Ideally you don’t run such a system over such a long period as 4 years, but rather over much shorter cycles with predictions like “the upcoming environment report will show decreased air purity by more than 5% in central london” or “the result of the local NHS restructure survey from Ipsos-MORI will show an overall increased level of patient satisfaction”. Bet now, bet fast, because if you’re so sure you know the answer, you’ll make out like bandits.
Over time, the idea is that people will tend to bet along side parties with the most proven track records, creating more stable ‘bookies favourites’ in traditional sports terms. But unlike the horse races where the bookies are always one step ahead and ensuring the house always wins, a very significant amount of people will vote for the outsiders due to their existing political allegiances, tribalistic tendencies and plain stubbornness. And rational and informed people will take their money most of the time.
Bringing it all together
I’ve outlined how we may welcome fact check bots (or at least the bots we like) into our media-consuming lives, and how markets for all kinds of ideas, even politics create powerful incentives for predictions and research. Now we have to turn the system up to 11. What if rather than just individuals and organisations playing the prediction markets, we allowed hive-minds to do so?
I recently encountered a sports prediction bot, entirely powered by what sports fans said, then reacted to, then changed their mind to, then decided a consensus on — or not. This type of complex crowd sourcing is far more sophisticated than simple direct democracy, it’s a highly quantified social discussion model which can beat sports and even film experts at all kinds of predictions.
“Ah ha!” you might say, “That’s not real AI, that’s artificial, artificial intelligence!”. Well you say, potato I said root vegetable. The analysis algorithm is based on hard statistics and commentator shifting opinion, but the confluence between the two and the evolution the discussion and analysis is truly an emergent phenomenon. So now we have AI’s plugged into data sources and hive minds which can be programmed to make predictions on the markets. They can differentiate public option from fact via the fact check bots, look for the safest bests where the current pundits have got it wrong and win everyone’s money, again and again and again…
And you know what? A bot that wins the market much more than anyone else is probably on to a good thing. Perhaps I want to vote with it more often than against so I stop wasting my investments here. Politicians may fire their fleshy human advisers and see why Socialism Bot 9000 is getting rich — or more likely why Utili-tron, the utilitarian AI is always the bookies favourite. And after a while, in a number of areas we may find the politicians simply doing what the AI suggest, because they are provably right far more of the time.
At this point why elect the politicians at all — why not vote for your favourite AI with the best track record of both accurate predictions and — maybe just a bit — favouring your preferred politics.
You know, representative democracy might not so bad after all. Now if you’ll excuse me, I need to plug myself back into the Matrix and go cast my vote with Utili-tron, he’s seeking re-election for 10,000th year running, and you wouldn’t want the wrong robot to get in…
Yours, a futurist, political pundit and frustrated ‘remain’ voter looking to change the world for the better
