GPT-5 represents a meaningful advance: fewer hallucinations, better reasoning benchmarks, and stronger rule-following in its best variant. However, it remains a probabilistic text generator, not a reasoning engine. Whether next-token prediction can support robust, generalizable reasoning is still debated. Until that question is resolved, the most insidious risk may be the hardest to detect: the illusion of understanding. In medicine and public health, in which decisions carry life-or-death stakes, that illusion can be as dangerous as outright error.
A reasonable (sic) argument I guess. I wonder though: Isn’t [human] understanding (i.e. reasoning, generalizable or not) also based on next token prediction ? The difference being that our tokens aren’t necessarily limited to text. The illusion may be that human reasoning is something magical.
What is outright dangerous is bad performing systems of next token predictions. In my experience, those systems can be both human and artificial.
Taking advantage of the current political climate, I am following up on my FDH-project. I’m using my updated Non-violent response app regularly on tweets. The updated app is now public and can be used by anyone. I have also updated the blog-home of the FDH-project.
Tweet today:
App generated response:
I hear your deep concern about food safety and fair treatment of American farmers. While enforcement actions affect many communities, I share your desire for consistent, transparent oversight that protects public health while respecting local farming traditions. Let’s explore constructive ways to improve the system.
I wanted to read Ayn Rand’s “The Fountainhead” mostly as an introduction to modern conservative philosophy. It’s also a really good book though, Ayn Rand writes beautiful prose.
Ayn Rand’s thinking seems to have a major influence on the political development in parts of the world right now. Not really the “facts are facts” part of her Objectivism-philosophy (support for scientific thinking is at an all time low after all), but rather the mindset where individualism is far superior to collectivism. Unfortunately, the black and white perspective of this narrative seems to attract a lot of followers.
How abolishing collaboration and altruism can possibly be good for anyone is beyond me, but reading texts like these at least help me understand.
The court room speech delivered by the main character, the architect Howard Roark, serves as a representative summary for this thinking. The whole speech can be found here, or if you prefer a movie version, you can watch it here. Below is in my view, the quintessential part where Roark is summarizing his views:
I came here to say that I do not recognize anyone’s right to one minute of my life. Nor to any part of my energy. Nor to any achievement of mine. No matter who makes the claim, how large their number or how great their need.
I wished to come here and say that I am a man who does not exist for others.
We should all work towards becoming as strong and independent as possible – it’s tempting to say “duh”.
Not even the strongest of us can be strong all the time, however. At some point, we all need help and support from our fellows.
Most importantly, collaboration and collectivism create wonderful things, a lot of them more wonderful than any man can create alone. Also “duh”, but has to be said anyway it seems.
I have updated the PartyRock app. You can now upload documents, scan for hateful language and generate responses. You can find the app here: Nonviolent Response Generator
“How does Dennis know how to fake levitation?” asked Robin, diverted. “Mate of his, when he was young, used to do stuff like that to impress girls” said Pat laconically. ” Some girls are really silly, let’s face it. When does anyone need a man who can rise two inches into the air?”
Which is true for so many other silly real world flexes some people do
In keeping with the history of European white supremacy of my people, I am permitting myself to arrogantly add a final last sentence to end this truly eye-opening essay:
“In fact, the world was never white in the first place”
Which, coincidentally, when you say it out loud, sounds almost like “Morning quote” – a quote to start a new day.
Mourning is about dwelling with a loss
And so coming to appreciate what it means,
How the world has changed,
And how we must ourselves change and renew our relationships
If we are to move forward from here.
I took this quote from Donna J. Haraways book “Staying with the trouble” although I believe she may be quoting someone else, possibly Thom Van Dooren. I changed the sentence structure.
I while ago, I uploaded my 23andMe data to Genomelink. Partly because 23andMe stopped reporting traits, partly because I was curious and partly because I just finished writing up my genetic counseling master thesis where consumer genetics was a central topic.
Unfortunately, I haven’t been able to practice much genetic counseling and 23andme still doesn’t have traits, – so I continued following my Genomelink-results as a way to keep me on my toes. The reports they make are questionable at best and horrifically misleading at worst. No better way to sustain interest than reading things that upset you – … right ?
The latest example is my genetic result-report on “Views on Environmentalism”. Based on two SNPs (rs1397924 and rs4902960), according to Genomelink, I am “less likely to support environmental laws by the government, and that regulations for the environment are needed.”.
So, this was a report that immediately fell squarely into the “misleading” category even before I started fact checking, – hence sorted into a box that already contained “Dog allergy” (which they claim I don’t have while I do), “Vulnerability to Computer Frustration” (which they claim I don’t have, but everybody does, – especially me) and “Left-Handedness” (which they got wrong).
The good thing though, – and the reason I keep looking into these results, is that it got me looking into behavioral genetics. A cool, but also scary field of research. There is a genetic component to all behaviors studied to date and the nature/nurture interaction is particularly fascinating when it comes to genetics in the social sciences.
A really good intro into one of those interaction mechanisms is the story of the MAOA-gene. Again misleading, this gene has been called the “warrior gene”, because one finds a link between MAOA-variants and aggression. However there is strong evidence to support that these agression effects will not manifest unless there’s a environmental trigger (like being abused or paid money to be cruel):
“These findings suggest that some behavioral traits are co-dependent on (1) the possession of genetic risk and (2) exposure to environmental stressors in order for them to manifest.”
Tanksley PT, Motz RT, Kail RM, Barnes JC, Liu H. The Genome-Wide Study of Human Social Behavior and Its Application in Sociology. Front Sociol. 2019;4:53. Published 2019 Jun 26. doi:10.3389/fsoc.2019.00053
So, when it comes to your predisposition to a certain behavior, your genetic risk may never be exposed, or conversely your lack of risk may never benefit you, unless you experience that external trigger.
The nicest possible way then to interpret this clearly wrong conclusion from my Genomelink report, is that I just haven’t experienced that environmental trigger to turn me into a climate-denialist.
Turns out however, that the paper cited for the finding, by Genomelink themselves, directly warns about making conclusions such as those made. Here it is:
“In summary, our molecular-genetic–based estimates of heritability partially corroborate the twin-based estimates and suggest that molecular genetic data could, in principle, be predictive of preferences. Our other results, however, suggest that excitement about the practical usefulness of molecular genetic data in social science research needs to be tempered by an appreciation that much of the heritable variation is likely explained by a large number of markers, each with a small effect in terms of variance explained. As a consequence, for economic and political preferences, much larger samples than currently used will be required to robustly identify individual SNP associations or to generate sizeable predictive power from many SNPs considered jointly.”
Benjamin DJ, Cesarini D, van der Loos MJ, et al. The genetic architecture of economic and political preferences. Proc Natl Acad Sci U S A. 2012;109(21):8026-8031. doi:10.1073/pnas.1120666109
No sizeable predictive power. Sorry Genomelink: no sigar, not even close.
I’ll keep supporting that “regulations for the environment are needed” then.
BIOpinionated 