History does not give the answer to how KI will change working life

Merete Nygaard warns in DN June 17 that KI can make an entire generation unemployed. The warning deserves to be taken seriously, not dismissed as evangelism.

The response from KI researchers and economists has provided few reasons to dismiss the warning. Torfinn Harding claims that “history shows that new technology has driven enormous wealth gains.” However, no one doubts this. The big question is whether KI, a technology that has the potential to perform large amounts of cognitive work, differs significantly from previous technologies.

Nor is there much comfort to be found in Inga Strümke and Anders Løland's point that there is still a need for “experienced programmers” in “quality-assuring Ki-generated code.” That is true, of course, but it is of little help to those graduates who have not had time to become experienced programmers, editors, bureaucrats.

Research shows that those who enter the labour market in recession get a more difficult path through working life, with lower wages and poorer work.

It's already signs of that the graduates are struggling in the competition with KI. But the important question is what KI models we will get in a few years and what consequences it will have for the labor market, not what they are getting at right now.

We have already seen tremendous progress in models' capabilities and development shows no sign of stopping. Investments are sitting loose and competition is fierce. Meta is rumored to be luring leading developers at its competitors with one billion kroner! According to thousands of experts on KI is it more likely than not, that by 2047 we will have KI systems that perform all work tasks better and cheaper than humans.

Nygaard focuses on the situation of the graduates and there is good reason for that. But that particular issue is well known. The problem she points to is that it is in the collective interest of all companies to train up experienced labour. At the same time, it is in the self-interest of every company to rather use KI than to train the newly graduated workers.

We already know this collective action problem from today's labor market. Economists have long told us that a free market will lead to less investment in workers' skills than is optimal. Every business would rather let other businesses do the training than do it themselves. Solutions to this problem are for the state to fund much of the education and require businesses to have to take in apprentices.

The question we need to ask ourselves is whether this model can be scaled to the new challenge from KI. Is the answer even more education than today's young people already receive, or are there other ways to distribute the burden of building young people's skills in the world of work?

The KI challenge probably requires continuous skill building, not a one-off education. To solve this problem, we need new institutions and funding models that can address the need for lifelong learning at a time of exponential technological change.

In a situation where more workers are being outcompeted, it may be tempting to put in place a citizen's wage scheme, since the current welfare state is not adapted to permanent unemployment. But public wages are both too much and too little at the same time. It is too little for the person who loses his job and not strictly necessary for the one who is still working.

Basic salary can be the solution when many people have already lost their job, but not on the way there. The scheme is hugely expensive scheme and little targeted to cover up the loss of income for ever larger parts of the economy. We risk building up unsustainable public spending before we have secured the revenue base from the KI revolution.

The time to prepare is now -- before “generation unemployed” becomes a reality that we are not equipped to deal with.

‍