Chat GPT has been an overnight sensation in the world of internet dependents – that is most of us. Though as any overnight sensation would attest – it takes a lifetime of sacrifice and investment to become that overnight sensation. As has surely been the case with the application of Generative Artificial Intelligence. Huge investments have been made – are being made – in developing and applying AI. And the great IT firms are leading the way with their impressive operating margins and returns on capital and abundance of cash flow invested in clouds of computers . They profitably supply the indispensable picks and shovels at the frontiers of knowledge.
And much of their heavy R&D is in the form of employment benefits for their army of researchers – increasingly applying AI – to answer the questions their customers and colleagues ask of them- and answer them far more effectively and rapidly. Among the important applications of AI is in the writing of the code that animates software and its development. With AI the applications and adaptations – the answers to the coders – comes much more rapidly. And the R&D is mostly expensed through the income statement and may not appear on the balance sheet. But will attract great value from the investors who determine the share price of the IT giants who dominate the market value of the S&P 500. Understandably so given the promise of AI. Perhaps the most important question shareholders should now be asking their managers is how are you adapting to AI?
It is estimated that a fifth of the time of office workers is spent answering enquiries of one kind an another. Imagine AI as that true expert on the customer or the internal functions and operations of your company always sitting beside you and your laptop, and comfortably speaking and understanding your language. You will have clearer answers and immediate answers to the questions. Better still the expert may help you ask better more imaginative and important questions- the answers to which will follow. It is asking the right questions that lead advances in science. Humans will be needed for that.
McKinsey has attempted to measure the potential of AI from the bottom up so to speak. By examining in detail how AI is now and could be adopted in the workplace. They have come up with very imposing estimates of extra GDP and of faster rates of change of output and productivity. Which is output volumes divided by numbers of work hours producing them. To quote Bloomberg on the McKinsey study “ Whole swathes of business activity, from sales and marketing to customer operations, are set to become more embedded in software — with potential economic benefits of as much as $4.4-trillion, about 4.4% of the world economy’s output — according to the study by McKinsey’s research arm…….Depending on how the technology is adopted and implemented, productivity could increase 0.1%-0.6% over the next 20 years, it found….”
A follow up question is worth asking. How well will the growth in productivity show up in the numbers we use to measure output and productivity and its growth? One of the puzzles economists have been wrestling with for many years is the apparently persistently slow growth in productivity recorded over many years despite conspicuous automation and labour saving. Are we entering a new phase of productivity improvements – almost certainly – but to recognize them we will need superior techniques to measure them.
We measure the value rather than the volume of production. Revenues recorded are prices charged in money of the day, multiplied by the quantity of goods and service supplied- easier to measure in mines, farms and factories than in the increasingly predominant service sector of a modern economy that sells a service the volume of which may not be obvious. For example how does one judge the quality of a report produced by an analyst today- enhanced by abundant data and powerful software and increasingly by AI? Not surely by the number of words written. Furthermore an enhanced customer service, better advice more rapidly provided, as for example, provided by a call centre, now armed with AI, will not be usually be directly charged for. The improved benefits it provides will come with a single charge for the good or service supported by a call centre- a laptop or cell phone for example. Or the fee charged by a customer relationship manager- a financial advisor perhaps, based upon assets being managed. A higher price or fee perhaps charged for the good or service bundled with the call centre or advisor would not necessarily mean more inflation. But rather represent a higher payment for an improved good or service.
To calculate output (GDP) and incomes or the values added we compare firm revenues today with revenues one or ten years ago, when prices observed were generally lower- given inflation. To make comparisons of real output and income and their growth over time, the value of all the goods or services supplied, must be adjusted for inflation to estimate the volume of goods or services produced. To estimate the real volume of goods or services produced or incomes generated over time. Most important prices have also to be adjusted for the changes in the quality of the of the goods and services supplied. We will not just be comparing the price of an aspirins with an aspirin which may well have increased over time. But rather comparing the prices charged for ever more accurately targeted capsules, developed with the aid of AI, and worth more in a real sense. AI is very likely to improve the quality as well as the quantity of goods or services provided.
Improved and lower costs of production could bring a mixture of absolutely lower prices and improved quality. It might mean deflation rather than inflation. How much prices fall in response to increased supplies will depend upon the growth in demand generally. That is on monetary and fiscal policy that could cause prices to rise on average even as economic growth – that is the volume of goods and services provided is growing. But if you underestimate quality gains incorporated into the price of goods and services and overestimate inflation by a per cent or two a year, you will then be underestimating productivity gains and economic growth at the same rate. And then be telling a very different story about economic progress. Perhaps AI will help economists and statisticians adjust more accurately for the changing and improved nature of the goods and services we consume.