Back to Stories
STORIES

"We need to seriously think about the total cost of digitization..."

13 June 2024
Portrait image of Sarah Spiekermann, Professor of Information Systems, Vienna University of Economics and Business
Sarah Spiekermann, Professor of Information Systems, Vienna University of Economics and Business

We talk to Sarah Spiekermann, Professor of Information Systems, Vienna University of Economics and Business (WU Vienna) in our series of interviews with leading industry experts who have contributed to our forthcoming news report: Trusted Journalism in the Age of Generative AI 
 
Lead author and interviewer is Dr Alexandra Borchardt
 
The EBU News Report 2024 will be available to download from June  

Is generative AI a game-changer in artificial intelligence?

It is certainly a game-changer, because it is a new way of how people communicate with the digital. The second reason, it is multimodal, so we don’t only have the text and speech interfaces, but also video, music. Literally, it changes the way we interact online. 

At the DLD conference 2024, you gave quite a downbeat presentation on AI. Was that to provoke a response or are you genuinely pessimistic about it?

I would call myself a realist. I don’t get carried away. I don’t believe in superintelligence. We are seeing a new cycle of productivity gain. It is not the solution to reach true progress or even solve our true current IT problems. 

Most experts predict that generative AI will transform pretty much everything. What are the true problems you are talking about?

The cost of IT is becoming unbearably higher all the time. Not only do companies need to invest in hardware and software that is constantly seeing updated product releases and that needs to be administered. There are also increasing legal and governance costs, security and privacy costs. Cost to be compliant with the AI Act, DMA, DSA, Cyber Resilience Act, etc.. Electricity costs are exploding. And on top of this, complexity adds ever more new unknowns.

What about environmental costs?

Right. IT companies often claim they are CO2 neutral, because cloud centres are placed next to hydrogen power plants. But our ecological balance is not positive. You have to start with the resources, the minerals. For sourcing one ton of rare earth, you create 75,000 litres of acidic water, for example. Such effects need to be included in a truly meaningful ecological balance sheet. Such a balance sheet must include globally distributed mining, shipping, manufacturing, and assembling across the thousands of resource and parts suppliers of IT components, and then add to that the ecological cost of service provision and consumption. Anything else is just kidding oneself about the true ecological cost.

Is it then realistic to invest so much in AI and perpetuate the digital transformation?

I am not sure. I increasingly doubt the sustainability of digital transformation at the scale we’re pursuing it right now. Another reason for this is we face increasingly more supply chain difficulties in our tumultuous world. If you need to build a machine with over 430,000 components, that you need for instance to produce machines that produce GPUs [graphics processing units], you have the most complex supply chains ever created by humanity. 

Is there a solution for that?

Peace. The West cannot afford to have too many geopolitical enemies that hold the resources for our innovation cycles. How can we have green transitions if solar panels come from China? We are already facing a dramatic shortage of IT supply. We are talking about sophisticated AI solutions, and you cannot even get an Xbox in time for Christmas. We need more discussions about these hands-on challenges in the media. Journalists love to speculate about superintelligence, they should report on something that is relevant now. The motor of innovation is chip technology. Chips are much more complex than oil or gas. There is not going to be a solution. There is going to be hopefully an understanding that we need global cooperation. 

Aside from this, what kind of potential do you see with generative AI?

There are interesting potentials in this new way of interacting with machines that have access to information patterns that we don’t even know about ourselves, not even in science. There is tremendous potential for learning, science, it can save massive amounts of time with dull documentation stuff and international cross-language communication. But if it is not built in a reliable, ethically responsible way, in the end what will happen is the same as what happened with social networks: the gains will be traded into net zero or even negative, because we don’t consider the drawbacks and social cost of the technology. With social networks this has been truth, manipulations, lack of transparency and struggling democracies. These are no small challenges. 

So, would you recommend we slow down?

Particularly if you are using these technologies in services that are vital for citizens, like food, electricity, telecommunications. Because what will you do if you cannot maintain this in 10 years because there are no chips? 

What about regulation, do you think those involved are up to the task?

There are very intelligent people in charge, but many of them are exhausted, they have little bandwidth to live up to these challenges. What makes it harder is the general zeitgeist: if you criticize in a period of hype, you are immediately marked as a pessimist, you are cornered. We should really stop this. We need a more realistic debate because then responsible actors are enabled and empowered.

Does the media live up to the task?

My impression is that there are two tribes in the media: sceptics and optimists. But optimism without realism is naive thinking. Politicians and executives are in their own filter bubbles. 

What’s your recommendation?

We need to seriously think about the total cost of digitization. We need to work towards an ecological global balance sheet. We need to have strategic research on geopolitical effects on the IT industry, because of the risk of running out of chips. Finally, we have to get a more holistic understanding of human capabilities. We need values-based engineering of AI. Building machines in a way that they foster goodness and virtue in human decisions, not the virtue of just another machine to increase profitability. 
 

Relevant links and documents

Contact


John O'Callaghan

Head of Content Communications

ocallaghan@ebu.ch