Singapore's central bank warns AI isn't ready to handle monetary policy

A senior policymaker at Singapore's central bank, the Monetary Authority, has suggested that artificial intelligence technology is not yet suitable to inform its policy development work. "Over the past year, central banks have had to answer difficult questions about our collective failure to foresee the persistence of …

[Rikki Tikki]

Laws of economics

First law of economics: for every economist, there exists an equal and opposite economist

Second law of economics: they're both wrong

Third law of economics: if you replace one economist with an LLM, they'll still both be wrong.

[amanfromMars 1]

Upton Sinclair* and Mandy Rice-Davies Applies**... and in Spades too***

Well he would [say all of that], wouldn’t he.** It is difficult to get a man to understand something, when his salary depends on his not understanding it.*

And aint all of that the honest gospel truth with particular and peculiarly specific regard to emerged and submerged and converged AI/LLM matters ..... which you do best to accept, because it surely destined to be obviously increasingly more evident, is the Surprising Uprising of AIMachines is so deeply embedded and entangled and vital within and across human SCADASystems as to render such operations practically effectively captured for otherworldly remote virtual command and control in the hands, hearts and minds of its IT Controllers and their Universal Drivers and Global Operating Devices.

Do yourselves a great favour and start to realise that you do inglorious and bankrupting battle in vain against future leading technologies which be designedly quite alien in concept and nature for all manner of very sound reasons and best practices protecting and securing creative core source and raw kernel ore. To do otherwise has one fated to wander and wallow and perish in petrified worlds intent on delivering stagnant growth and terrifying resulting product.

* ... https://www.goodreads.com/quotes/21810-it-is-difficult-to-get-a-man-to-understand-something

** ..... https://en.wikipedia.org/wiki/Well_he_would,_wouldn%27t_he%3F

*** ...... https://grammarist.com/idiom/in-spades/

[Neil Barnes]

Magic 8-ball

Reply hazy, try again later.

[Bebu]

Very good sir.

"Do yourselves a great favour and start to realise that you do inglorious and bankrupting battle in vain against future leading technologies which be designedly quite alien in concept and nature for all manner of very sound reasons and best practices protecting and securing creative core source and raw kernel ore. To do otherwise has one fated to wander and wallow and perish in petrified worlds intent on delivering stagnant growth and terrifying resulting product."

I very nearly understood this twaddle.

[Bebu]

Augury?

"generate dynamic measures of inflation expectations using social media posts,"

Consulting chicken entrails might well be as useful?

[Bebu]

No change here then?

"they [LLMs] are not yet capable of providing credible explanations for their own predictions."

Not so very different from the dismal array of politicians, bankers and economists.

[Howard Sway]

artificial intelligence technology is not yet suitable to inform its policy development

In the same way that my dog is not yet suitable to run the ministry of finance.

[Graham Cobb]

Surely this isn't the sort of things (today's) AI can do?

I'm no expert on AI - and I look forward to being educated. But, to me, it appears that today's AI tools (particularly LLMs) are very much geared to searching data, and looking for statistical correlations. As far as I can see, they do not claim to do any reasoning.

To take a simple example, simple reasoning such as "if I reduce the price of my goods, demand will increase" is not within their scope. They are, of course, able to use lots of data to see that cases where the price of goods have gone down are heavily correlated with demand increasing. But this is just correlation - they have no idea which effect is causing the other, and no way to apply the reasoning to cases where there is little data (for example, are prices falling correlated with meteorites falling? no answer because too few meteorites fall to have any data).

So, it would seem to me that monetary policy, which is a complex area involving a lot of understanding of the behaviour of markets and people is one of the last areas to be able to be analyzed by AI. Of course, they may be very useful in finding and crunching the data needed by the human analysts, but they are not going to be making predictions, let alone "credible explanations". A different sort of technology is going to be needed for that, presumably.

[amanfromMars 1]

Re: Surely this isn't the sort of things (today's) AI can do?

As far as I can see, they do not claim to do any reasoning. ..... Graham Cobb

Quite so, Graham Cobb. It is humans making all of the claims about what AI may or may not be able to do. And if the truth be told, they haven't a clue about how disruptive and destructive or creative and constructive it is sure to be ..... although of course that will not stop them spinning all sorts of yarns and tall tales revealing all of their hopes and wishes or fears and worries about something relatively novel and physically unknown .... and about which they in truth know very little, if anything at all ..... leading command and exercising remote virtual future control.

Que Sera, Sera.