Summarize this content to 2000 words in 6 paragraphs in Arabic Unlock the Editor’s Digest for freeRoula Khalaf, Editor of the FT, selects her favourite stories in this weekly newsletter.Greetings to all. Thanks to Tej Parikh for writing last week’s Free Lunch on the bearish case against the artificial intelligence hype. I’m dipping back in this week to contribute some accompanying thoughts on how the public policy thinking on AI is evolving. For the next three weeks, various FT colleagues will keep the newsletter going so you can keep up during what I hope will be restful and re-energising holidays for all Free Lunch readers.I found quite fascinating Tej’s analysis of the numbers you have to believe for expectations of an AI business boom to add up. I have no reason to challenge them; if anything they leave me a somewhat greater AI sceptic than I already was. But much as if you want peace, you should prepare for war, or you should be ready for pandemics while hoping they never come, we should also formulate policy for the eventuality that AI does become the hugely disruptive technology (financially and otherwise) many think it will. I wrote two columns on AI last year: one taking stock of the shock with which the world received ChatGPT, the other after the Bletchley Park AI summit. In both, I argued for making the AI debate a lot more boring. To handle this new technology responsibly, we should look at the mundane but real harms it is already triggering, rather than science fiction-style risks. The focus should be on technocrats not terminators, I wrote.So far, it appears I have had my wish granted. This summer, both the IMF and the Bank for International Settlements — two of the world’s most important technocratic economic institutions — have published reports on AI. And next week, the EU’s new Artificial Intelligence Act enters into force. Increasingly, such technical work is quietly crowding out the more sensationalist debate.The BIS chapter on AI from its annual economic report — which includes a handy primer for non-experts on how AI works — highlights the role AI can play in overcoming information bottlenecks in the financial system. For example, correspondent banking has been in decline because the costs of increasing informational demands from (much-needed) anti-money laundering rules are not always justified for what is a relatively low-margin activity. By significantly lowering the costs of know-your-customer checks and of assessing money-laundering risks, AI can, therefore, safeguard an important aspect of global financial connectivity. The BIS mentions lending, insurance and asset management as other examples where efficiency depends on affordable information processing.AI can help central banks do their job better, too, the BIS thinks, by improving cyber security and real-time analysis of the economy and financial stability risks. Both the BIS and the IMF give a good overview of the main macroeconomic issues. Sensibly, but unsurprisingly, the shared overall point they put across is that AI may be good for productivity, but (a) we have no idea by how much, and (b) it may well affect different tasks, skill levels and sectors in various ways. So there are likely to be winners and losers, with some earning more because they become more productive and others becoming obsolete in their old jobs. But again, we have little idea who will be affected and how. We should observe that this is not just a question of which workers will be made more productive by AI. It also depends on whether the lower effective cost of the tasks they perform means the use of those tasks will take off (so there is a need for more workers even if each is more productive) or simply spending on them will fall (and fewer, more productive, workers will be needed). There is a parallel here with how greater manufacturing productivity went hand-in-hand with rising western factory employment (in absolute numbers) in the three decades to the late 1970s, after which the opposite applied as ever fewer workers were needed to make the volumes of manufactured goods that markets could absorb (even before globalisation shifted some of the lowest-skilled jobs abroad).What to do, given how little we can at present predict? The IMF report has some good ideas. More generous unemployment insurance, especially a system that is tied to the overall employment situation, can have a big effect in giving superfluous workers in AI-affected jobs the time to find new and even better jobs elsewhere. I would add that high demand pressure is important — as we saw in the post-pandemic recovery.There is a theme here, if a sort of photonegative one, which is that these are serious but not highfalutin ideas. No terminators, singularities or godlike AI takeovers in the future, but concrete opportunities and risks in the here and now, and some sound advice about how to go about it.The same can be said for the EU’s AI Act, about which there seems to be a bit too much uninformed protest or suspicion. On the whole, it aims to do the sensible thing of categorising very real and present risks not so much of the technology itself but of the plausible uses of it, and of putting some restrictions on the riskier ones. It will, for example, ban dystopian applications such as subliminal manipulation of human users or China-style social credit systems. That is, surely, precisely the first step we would want regulators to take. (If only we had had such rules when online behavioural targeting first took off!)Of course, these more pedestrian exercises can also make mistakes, even if it’s not the mistake of letting science-fiction dangers draw attention away from real and present dangers. For example, I worry that the AI Act’s last-minute inclusion of regulation of foundation models, which unlike the rest of the law tries to regulate the technology itself rather than how it is used, may do more harm than good. I also worry about the IMF’s willingness to say that maybe we should tilt taxation towards taxing AI directly (rather than as just part of heavier general capital taxation) if the social costs of the labour market transition it portends are particularly high. That seems like an abdication to me. The best policy response to a disruptive but productivity-enhancing technological change cannot be to slow it down — that’s the stance that has led to the survival of wasteful and awfully paid manual jobs in the UK and US. Instead, it should be to double down on the policies that make companies elsewhere having to compete for workers looking for new jobs: high-demand pressure, active labour market policies and welfare schemes that minimise the cost of leaving a job to look for a better one.Mistakes and disagreements notwithstanding — they come with the democratic territory — these sorts of debates are much more down to earth than the first reactions to the recent breakthrough. That also makes them a lot more useful. More of this, please.Video: AI: a blessing or curse for humanity? | FT TechOther readablesAn FT editorial calls for western countries to welcome China’s electric vehicles as a contribution to their decarbonisation goals. In Beijing, advanced manufacturing capacity remains central to the government’s economic vision.The president of the Eurogroup of Eurozone finance ministers writes in an FT op-ed that Europe faces a budgetary inflection point.India has a new budget, which promises to rein in the deficit while satisfying the wishes of new government coalition partners and pushing ahead with infrastructure investment.Our Madrid correspondent explores Spain’s backlash against tourism.
