According to experts, AI will only have an impact on the labor market in years to come
Artificial intelligence is currently catching on like wildfire, both privately and in companies. The tipping point was programs such as ChatGPT, which can independently create texts, images and now also videos at a good and rapidly improving level. Thomas Mück, President of the Austrian Computer Society (OGC), is convinced that no company has yet laid off employees as a result. "But in the five to ten-year horizon, it will of course affect some professional groups," he says.
On the one hand, the boom in AI could partially compensate for the baby boomers leaving the labor market, but at the same time it will make it more difficult for young people to enter sectors such as the creative industries.
According to a conservative estimate, the programs that now run under generative artificial intelligence (AI), i.e. create new content, could save around 1.5 hours of net time in a 40-hour week, estimates Mück. Although the first drafts of the programs cannot be used without human control and revision, they are faster "than starting from scratch".
AI like a bank account: First free, then chargeable
So anyone who can use AI has a competitive advantage, emphasizes Mück, who was head of the Institute for Computer Science and Business Informatics at the University of Vienna until 2001 and has been running the non-profit association Österreichische Computer Gesellschaft OCG for almost a year. Its 1200 members include IT companies, ministries, research and educational institutes. The aim is to "promote computer science and communication technology".
The fact that generative AI is available free of charge for private individuals and at low cost for companies has contributed to its current triumph. However, Mück assumes that this will soon be a thing of the past or will only apply to basic models. It's like current accounts, which were initially provided free of charge by the banks, but "since you can no longer exist properly without a current account, they are no longer free".
Warning against AI monopolies
AI models have to be trained with enormous amounts of data and very high computing power. The more data is fed in, the better the models become - and the more they are used, the greater their lead. Only a few large companies have the necessary resources for this. The German AI Association already warned in a 2021 analysis that monopolies or oligopolies could emerge for AI models - and that Europe is not yet at the forefront.
Amazon has now invested four billion dollars in the AI start-up Anthropic, Microsoft has invested ten billion dollars in OpenAI and has announced AI investments of around three billion euros in the UK and Germany. Google (Bard) and Facebook, as well as Chinese tech giant Baidu, are also investing billions in this area.
Europe is being left behind
Europe cannot keep up financially. Mück mentions two European models as relevant. The French "Mistral AI", which has so far raised around half a billion euros in funding, and the Trust LLM at Linköping University in Sweden, which has so far been financed primarily with European research funds. The German AI company Aleph Alpha is also financed with around half a billion euros.
Science can no longer keep up financially. Europe would be at the forefront of AI research, but putting it into practice is not working, says Mück. The majority of market-ready AI models currently in use are offered by US companies, while the enormous computing power required is concentrated in Asia, especially India and China. Europe is in danger of being left behind, as has already happened with electric cars and photovoltaics. In the end, only a few AI models will remain worldwide, Mück expects.
Data protection as a "club"
On the other hand, Mück has no data protection concerns. Data protection is important and would be manageable from a factual point of view, but is repeatedly "instrumentalized as a cudgel" when lobby groups want to derail projects: "If I don't think something is politically opportune in the broadest sense, the easiest thing to do in Europe is to come up with data protection," complains Mück.
The strict laws in Europe are also not the reason why projects fail on this topic, but rather interest groups that use the topic to bring down things they don't like. Of course, this could also happen with AI, he admits. On the other hand, it is clear that in Europe we have to deal with people's "basic fear" of being overrun by IT. "As a result of digitalization, there are an extremely large number of people for whom their own present is becoming alien, who can no longer orient themselves." This needs to be taken into account.
Development too fast for legislators
Mück therefore also welcomes the recently adopted regulation of AI at EU level, the "AI Act". This regulates AI models to varying degrees depending on the "risk" to humans. However, "I wouldn't bet on whether the AI Act is the last word in wisdom". After all, developments are moving too fast for legislators to keep up. According to Mück, the drafting of the rules of the game was "not ideal, but that doesn't matter, it's a good attempt". In a way, EU regulation is symbolic policy, but it is important to signal to people: "We don't want to be passengers." He does not fear any competitive disadvantages as a result of the regulation.
Not everyone is aware that modern AI models represent a "neural network" and ultimately only estimate the probability of certain word sequences - factual truth or accuracy is not a criterion. AI can therefore also "hallucinate", i.e. invent new facts. Mück makes no secret of the fact that he would prefer an AI that is based on logical-mathematical laws. However, this approach has not prevailed. "The sentimental favorite, as is so often the case, didn't really get us anywhere."
Rule-based algorithms were probably not "the right form of representation", or only for very limited applications. The neural network, on the other hand, which has nothing to do with formal logic, has "worked in practice", even if we don't always know exactly why. Mück admits that "from the outside, it is certainly comparable to human thinking". In reality, however, it is not "elegant algorithms" that represent intelligence, but "only the enormous computing power", i.e. "brute force". (SDA)