familyandtrends 143

Who creates and who captures value in artificial intelligence?

Adobestock

4' min read

Translated by AI
Versione italiana

4' min read

Translated by AI
Versione italiana

Since March this year, companies have been seeing Excel files with far more worksheets, complex formulas and colours, and PowerPoint presentations with more slides, lots of rectangles and a great deal of text – often in unbearably small font sizes. Today, the most advanced AI models can work directly within files on PCs and create ‘artifacts’ in a matter of minutes, and so intellectual output is proliferating. In March, an issue of *Family and Trends* stated: “The impact of AI on intellectual work will be similar to that of factory robotisation on manual labour: greater standardisation of quality, less scope for craftsmanship.”

Will productivity grow as much as the enormous technological and financial enthusiasm suggests? On the financial front, Professor Damodaran recently stated: “I have noticed that investment banks estimate xAI’s addressable market at 22 trillion, which I think is more of a delusion than an estimate”. As for potential profitability, he drew an interesting comparison: “The success of streaming companies has not only been at the expense of traditional entertainment firms, but, on the whole, has made the sector less profitable.”

Loading...

From a technological perspective, analyses and documents will certainly be produced more quickly, and quality will become standardised. Some intellectual jobs will disappear, just as they did with the advent of office automation and the end of typewriters, ledgers, desktop calculators, hand-drawn charts, transparencies, photocopiers and all the people whose expertise was based on using these tools. Parkinson’s Law, i.e. ‘in intellectual activity, work expands to fill the time available for its completion’, has, however, allowed many activities to evolve whilst keeping working hours unchanged, if not increased. With the arrival of Excel, for example, those working in administration, finance and control seem to spend more time at the office. Excel has also introduced many levels of control, verification and analysis that were unthinkable when working solely with a calculator.

For the time being, since we’ve been using AI to analyse and prepare documents, it seems we’re spending more time interacting with the AI, and that’s a good thing. However, it also seems we’re spending more time reading longer and less clear documents, and that’s a bad thing. It is becoming clear that improving intellectual work requires a combination of human and AI. It is not enough simply to hand our work over to AI. This is an advantage for young people entering the world of work, just as it was in the 1990s for those who knew how to use Excel and its macros, compared to those who merely ‘ticked the boxes’ when doing sums.

Productivity may well increase. But who is already reaping the benefits of the value created by AI? For now, it is those who provide the inputs needed to develop it. A previous familyandtrends article compared AI to a diamond: in the 1940s, the stone became a symbol of eternal love and an essential element for proposing to one’s beloved. Since then, it has been De Beers that has truly profited, because it owned the mines. In the AI sector, the cash flow of semiconductor suppliers has risen from less than 50 billion in ’22 to more than 400 in ’26; that of the hyperscalers (Amazon, Google, Meta, etc.) has fallen from more than 250 to… zero.

These companies, alongside pure-play AI firms, are investing heavily to win a race in which they believe two rules apply: ‘first-mover advantage’ and ‘the winner takes it all’. It doesn’t usually work out that way. With the internet, for example, it took years for champions based on the new technology to emerge. It is no coincidence that one of the most innovative and long-standing companies in the tech world, Apple, has not altered its annual investment flow. For now, everything is going well because the investments are being capitalised and are not spoiling the golden age of profits. And when costs are capitalised, it is easy to be generous. When depreciation begins to weigh on the income statements and competition has run its course, we will see slashed profits, widespread impairment and disillusionment.

The race for computing power is based on a heuristic made famous by Richard Sutton in his 2019 essay *The Bitter Lesson*: “[In seeking to] make an improvement [to AI] that makes a difference in the short term, researchers tend to draw on their own human knowledge of the domain, but the only thing that matters in the long run is the exploitation of computing power”. Just think of chess. Computers kept losing as long as people tried to teach them moves and strategies, such as controlling the centre of the board and not weakening the pawn structure. In 1997, Deep Blue defeated Kasparov thanks to its enormous computational power. It didn’t play chess any better, but it could calculate 200 million positions per second and look 10–12 moves ahead, exploring a vast number of combinations. Today, when we ask it a question, AI performs billions of probabilistic matrix calculations to decide which word comes next. Yet it doesn’t have the faintest idea what it’s saying. Doing so requires a great deal of computing power, which costs billions of dollars. The question remains the same: how will consumers’ willingness to pay for this expensive computing power increase? For now, let’s just enjoy more complex and colourful spreadsheets and longer presentations with smaller font sizes…

(*) Lecturer in Family Business Strategy – University of Turin – bernardo.bertoldi@unito.it

Copyright reserved ©
Loading...

Brand connect

Loading...

Newsletter

Notizie e approfondimenti sugli avvenimenti politici, economici e finanziari.

Iscriviti