There is plenty of reason to worry about how AI is used
From Matthew Stevens, Sydney, Australia
Whether artificial intelligence will soon destroy the world is much in the news, but AI isn’t the problem; it’s how humans use it. The biggest threat to society at the moment is the naive use of AI to replace human labour, with no thought about what replacing workers with automated systems that simply follow instructions will lead to (3 October, p11 and p22).
Very soon, we face the threat of millions of people being put out of work in the interest of saving costs, for the benefit of the owners and shareholders, with no plan for the lives and livelihoods of the newly unemployed. Governments have two choices: ignore the coming labour crash or start taxing the financial beneficiaries of AI for the sake of everyone.
From Stephe Ellis, York, UK
A superintelligence that decides to eliminate us is an implausible distraction from more credible and immediate dangers. We should think about AI as a set of increasingly powerful and unpredictable tools that will inevitably become available to all.
Thirty years ago, the internet was new and we regarded it as benign. Seemingly nobody foresaw the multiple problems that many now attribute to this relatively simple technology. With AI, the situation is very different: a significant proportion of those who are intimately involved in its development are afraid of what they have had a hand in creating. When those “in the know” see the technology they are developing as potentially dangerous, we shouldn’t dismiss their fears.
From Richard Turner, Beverley, East Riding of Yorkshire, UK
The risk of machines becoming intelligent and seeking to destroy the society that builds and sustains them may be very small, but the risk of criminals using AI for their own purposes is evidently much higher. We are currently totally dependent on networked computer systems for all essential goods and services.
Even small acts of cybercrime could make some people very rich before it was even noticed, never mind major disruptions. Hackers have already shown how good they are at bypassing security measures to get inside computer systems to steal data or redirect payments. My guess is that, in terms of the number of attacks, we ain’t seen nothing yet. But the blame will lie with people, not machines.
True takeaway from study on dogs’ biased hearing
From Bob Ladd, University of Edinburgh, UK
Readers who were puzzled by your story about “consonant bias” in dogs listening to human speech would have been helped by a clearer account of what the dogs heard. The experiment fits into a line of research dating back to the early 1990s, in which listeners hear a long, boring series of short syllables lasting several minutes with no audible breaks. The string of syllables includes three-syllable sequences (“words”) like “tipola” or “maluvi” that recur periodically (3 October, p 12).
Such experiments have found human listeners, even babies, often notice these recurring sequences even though they aren’t audibly separated. It has also been shown that partial recurrences involving only consonants (e.g. tipola, tapele, tupalo, etc.) are easier to detect than those involving only vowels (e.g. maluvi, kaputi, savuri, etc.).
The dog study compared three patterns of such partial recurrences (consonant-based, vowel-based, neither) and found evidence that dogs, like humans (though to a lesser extent), detect recurrent patterns most readily in the consonants. Unfortunately, the story’s isolated examples of the three sequences (mozite, zobefu and fumoza) fail to make clear that the crucial difference lies not in the individual “words”, but in what recurs across the whole series.
The cosmos is beautiful because of our perspective
From Andrew Whiteley, Consett, Durham, UK
The universe is indeed awesome and beautiful, as Chanda Prescod-Weinstein says in her piece about the dangers of viewing the cosmos through the lens of humanity. However, it is only thus because we (and perhaps other conscious beings elsewhere) are here to see it that way. Without us, the universe is simply a conglomeration of forces and particles (19 September, p 19).
In a real sense, therefore, the universe points to us, not away from us, in a manner that isn’t mere hubris or narcissism on humanity’s part. Let us remember, too, that the vast scale and age of the cosmos tells us precisely nothing about our own importance or lack of it 鈥� unless, of course, we equate physical size with significance.
Causes of climate crisis begin to feel its effects
From Peter Jacobsen, Port Townsend, Washington, US
While Sam Edge correctly notes the vulnerability of nuclear power plants to drought, thermal and hydropower plants are also at risk. For instance, in Poland, two coal-fired thermal plants as their normal cooling-water requirements would have taken over half the Vistula river’s flow at that time. Paradoxically, the causes of our changing climate have become its victim (Letters, 12 September).
The first meeting of the simulation committee
From Graham Knight, Ware, Hertfordshire, UK
Reading the interview with Catherine Heymans, in which she discusses the possibility that we are living in a simulated reality, made me imagine the following scenario at the first meeting of the simulation committee following the funding decision:
“We’ll never be able to simulate the theory of everything on that budget (26 September, p 40).”
“No… but hey! Maybe we don’t need to! Remember all that quantum and curved space-time stuff we learned at primary school? We could do that easily enough.”
“But they’re fundamentally incompatible! The creatures we simulate will never understand what goes on inside black holes or before the inflationary phase of the universe.”
“So what? They’ll never get inside a black hole, and we can kick the whole thing off after inflation’s done. They’ll still have superconductivity and warp drives.”
“Yeah, I guess so… OK.”
For the record
The letter on quitting research on hissing cockroaches was written by Margaret Nelson (Letters, 15 August).
The clue to 8 down should read “Lacking covalent bonds” (October, p 45).