We shouldn’t want AIs to have a human-like reality
From Will Mayall, Santa Cruz, California, US
Your article on “world models”, which would let AIs understand cause and effect in the real world, largely assumes that the goal is to give them something resembling the human experience of reality: objects, motion, cause and effect, and physical interaction at the scale of everyday life (8 August, p 30).
But humans have only the faintest sensory access to reality. We perceive a narrow band of electromagnetic radiation, a limited range of sound and a small set of chemical and mechanical signals. We don’t directly experience the molecular world, radio waves, magnetic fields, geological time, atomic processes or the scale of galaxies. Even our familiar senses fail to present reality itself – they provide highly filtered biological signals from which our brains construct useful models.
Why assume AIs should be limited in the same way? They could instead receive information from all spectra of light, molecular sensors, astronomical instruments, seismic networks and countless other sources. With that flood of information, an AI’s “world model” might be nothing like ours.
World models may indeed be important to advanced AI. But defining the “world” by what humans happen to sense is an unnecessarily narrow place to start.
An alternative to topical flea treatments for pets
From Jens Christian Jensenius, Odense, Denmark
There are interesting and worrisome elements in your report about the environmental safety of flea treatments for pets. However, it focuses entirely on treatments applied to animals’ skin. Chewable tablets are given orally, so they seem unlikely to be detrimental to the environment. I guess that, to be sure, you should collect your pet’s faeces for a couple of days after treatment to be disposed of as burnable rubbish (15 August, p 14).
Quantum maths solution isn’t in the spirit of things
From Dyane Silvester, Arnside, Cumbria, UK
I am unconvinced that adding a quantum element (which I presume Catherine the Great didn’t mention in her original challenge) counts as “solving” the puzzle she set mathematician Leonhard Euler. Surely this actually amounts to changing the rules, so that it becomes a different puzzle, which has a solution?
More work needed to refute Boltzmann brains
From Toby Pereira, Rayne, Essex, UK
In his interview with Jacklin Kwan, Sean Carroll doesn’t fully address the problem of Boltzmann brains. He has designed a universe where the issue is mitigated, but has offered no explanation as to how or why this universe might have become the universe (22 August, p 38).
If other universes exist that do have a proliferation of Boltzmann brains, we can still wonder why we exist as a mind from a brain in this universe, rather than some other. For Carroll’s theory to work, the proportion of minds that come from Boltzmann brains needs to be sufficiently small across all of existence, meaning his universe must be the only one, or at least statistically typical. It is hard to reconcile his universe with being statistically typical, as it relies on setting things up perfectly in order for entropy to spontaneously start decreasing until it reaches the same starting point.
Heatwaves reveal a flaw in push for nuclear power
From Sam Edge, Ringwood, Hampshire, UK
The shutdown of nuclear power plants across Europe due to low water levels from recent heatwaves is alarming. It demonstrates that building large nuclear fission plants, small modular reactors or even the mythical sustainable fusion reactor is never going to be a viable part of the fossil fuel transition (15 August, p 9).
Wind, solar and tidal power with electrochemical battery or thermal back-ups are clearly the only solution. Fortunately, those technologies already exist and are constantly becoming cheaper and more efficient.
Could we recover a literary great’s microbial traces?
From Stephen Durnford, Saltdean, East Sussex, UK
Matthew Ponsford writes about microbiome forensics in police investigations and testing for our microbial “fingerprints”. I have been wondering about the persistence of fingerprints on paper myself. During the 1870s, one of my great-grandfathers was a clerical colleague of Charles Dodgson, who wrote as Lewis Carroll and took a photograph of my great-uncle Cecil Crofts and two of his sisters. My great-aunts also received some of Dodgson’s letters and, in 1943, Cecil gave his print of the photograph to my mother, who later gave it to me (8 August, p 36).
Few other people are likely to have handled it, and Dodgson’s fingerprints may be among those present. Perhaps forensic analysis could even recover his DNA?
Anglo-Saxons may have seen a huge solar storm
From Mark Spinney, Haslemere, Surrey, UK
I was very interested in your recent article on solar storms. It struck me that an incident 10 times stronger than the 1859 Carrington event, estimated to have occurred around AD 774, might have been noticed and recorded at the time. Looking in the Anglo-Saxon Chronicle, a set of annals in Old English that was first compiled around AD 890, I see that the entry for AD 776 says: “Men saw Christ’s red cross in the heavens after sunset” (15 August, p 34). Could these observations have been a result of a solar event?
Unexpected implications of red-light research
From Richard Brown, Huntly, Aberdeenshire, UK
I note from Graham Lawton’s article “Red-light starvation” that many of us are deficient in exposure to infrared light. Bill Gifford’s earlier piece “Heat therapy” (18 July, p 26) highlighted the possible benefits of saunas (15 August, p 30). Does this not suggest that a prime area of immediate research should be the indoor personal saunas that use infrared to heat the body?
From Bob Foster, Leek, Staffordshire, UK
I have often wondered why dogs love to lie out in the sun. After reading your article, I perhaps have the answer: they are soaking up red and infrared light.
Want to get in touch?
Send letters to letters@newscientist.com; see terms at
Letters sent to Â鶹´«Ã½, 9 Derry Street, London, W8 5HY will be delayed