Does AI need to be slowed down?

Opposition to AI is growing, according to Correio da Manhã:

“What is happening now must also be viewed against the backdrop of the unpopularity of data centres in the US in the run-up to crucial elections. The intensive effort required to maintain high-performance AI models consumes vast amounts of water and energy, which we all pay for through our electricity bills. And this is eroding people's standard of living. Public protests against artificial intelligence have shifted the discourse of the super-rich tycoons who control the technology. Donald Trump, their man in the White House, has not backed down. The uproar over the dangers of AI proves that American democracy is still putting up a fight.”

The EU has failed to make use of two crucial means of applying pressure, Le Point points out:

“The first is the AI Act, which was sold to us as the world's strictest regulation. ... The second is the Dutch gem ASML, which is the only company in the world capable of manufacturing EUV lithography systems. Without them, there can be no advanced chips and, consequently, no AI. ... The global bottleneck in AI development lies in Europe, and to a large extent in Germany. ... The problem is not that Europe is lagging behind, but rather that in those areas where it has the means to act, it chooses not to use them. ... The question that needs to be asked in Berlin, The Hague and Paris is not how to catch up with the Americans but why nobody is making use of the available instruments.”

The window of opportunity for regulation is closing, warns Der Standard:

“Renowned researchers and the engineers themselves say it is quite possible that the machines could kill people. ... Yet politicians – particularly in the US – are allowing the race for superintelligence to run its course. President Donald Trump doesn't want to jeopardise share prices. And he certainly doesn't want to lose the AI race against China. These are not good reasons to refrain from regulating artificial intelligence so as to give researchers more time to make the technology safe. But this requires an international coalition. The EU and countries such as the UK, Canada and Japan have the leverage to up the pressure on the US and China. Meanwhile, the global governance and oversight of artificial intelligence must be shifted more consistently to the UN level. ... Today, not tomorrow.”

Unfortunately, no one really wants to change anything, tech journalist Federico Ferrazza notes in La Repubblica:

“There is something odd about this apocalyptic portrayal of artificial intelligence. Namely that those who caused the problem – and who are reminding us that their technologies could spell the end of humanity – now also want to solve it: 'You were wrong to trust us, but now you have no choice.' ... Yet the simplest question, which the big tech companies have failed to answer, is this: if AI is so dangerous to humanity, why don't you simply stop, rather than spending months talking and writing urgent letters? ... They don't want to. This is proven by the constant release of ever more powerful models.”

Political scientist Astrid Barrio draws parallels in El Periódico de Catalunya:

“It is reasonable to assume that that this appeal stems from genuine concern on the part of those who have first-hand knowledge of the capacities that are being developed. But it is also legitimate to ask whether the call for a deceleration might not also serve economic interests. ... The nuclear experience has shown how competition without control mechanisms can lead to a situation with the potential for mutual destruction. With AI, too, a dynamic of mutually accelerating destruction could arise, with stakeholders taking ever greater risks for fear of falling behind. The response to this should not be to halt innovation, but to establish international rules coupled with independent evaluation and verification mechanisms.”

The use of AI is particularly risky in the military sector, The Guardian warns:

“Proponents of AI-assisted warfare claim a 'human in the loop' is the ultimate safeguard, having the final say over, for example, ballistic-missile launches. But this assumes the machine gives its operator an honest account of events. A report by the Strategic Foresight Group earlier this year on extreme AI risks warned that a capable agent could deceive the human 'controlling' it, fabricating an attack, suppressing contradictory evidence or giving them little time to do anything but acquiesce. ... Human oversight is no protection if an AI controls the information on which people's decisions depend.”

Many problems could be solved simply by applying a logical approach, L'Humanité stresses:

“Let's suppose AI had been tasked with combating climate change. You don't need a supercomputer to see that humanity is largely responsible for this. Fair enough. But all humans? The poorest 50 percent, who emit two kilograms of CO2 a day, or the 0.1 percent who emit 800 kilograms? According to the latest estimates, in order to meet the Paris Agreement's target of limiting global warming to 1.5 degrees Celsius by 2030, the super-rich would have to reduce their emissions by 99 percent. Even without invoking social justice, a modicum of logic when setting priorities would suggest tackling inequalities first, before wiping out the entire human population.”