CC-BY-NC
Category Archives: Philosophy
Is it a crime to use AI for peer review?
I consult the almighty chatGPT frequently for additional information as this saves me hours of wading through my own database, Pubmed, Scholar and Goggle Hits.
But I have my own opinion, I never cut & paste as this is always running at risk (1) to plagiarize unknowingly and (2) to produce nonsense.
Miryam Naddaf has an article about this
In a survey of nearly 5,000 researchers, some 19% said they had already tried using LLMs to ‘increase the speed and ease’ of their review. But the survey, by publisher Wiley, headquartered in Hoboken, New Jersey, didn’t interrogate the balance between using LLMs to touch up prose, and relying on the AI to generate the review.
And well, maybe I am already sticking to the NEJM that said
Although human expert review should continue to be the foundation of the scientific process, LLM feedback could benefit researchers
The biggest problem of scientific integrity investigations
It is not the weaponization or the destructive vendetta of some investigators. The biggest problem is also not about the technical skills. It is something completely different —the intentional misuse of the internal scientific discussion (does something like this still exist?) by an political agenda. As devybee puts it forward in a book review about the swamp of science fraud
Szabo is impressed by the efforts of “data sleuths”, who perform post-publication peer review and report problems on the PubPeer website, but he regards this as unsustainable: and cleaning up the literature should not be a task for volunteers. …Szabo’s recommendations for change focus on funders, who have the power to deny funding to those who fail to take steps to ensure that their results are reliable. …This is a particularly difficult time to be conveying such a message. The only people who might be overjoyed to hear that a high proportion of published research is unreliable are politicians who are antagonistic to science and would like an excuse to defund it.
The great consciousness debate
The most recent discussion abut consciousness was going in circles, having now a dozen of competing theories as summarized by chatGPT and reworked by me for omissions. Not sure if all the summaries and references are good, but maybe this is a good starting point to understand the background of a new paper.
Continue reading The great consciousness debate
The trouble is, you think you have time
the title says it all
Similarity between false memory (of humans) and hallucination( of LLMs)
The common theme seems the low certainty about facts – a historical event that is wrongly memorized by a human or the Large Language Model that wrongly extrapolates from otherwise secure knowledge. But is there even more?
Yann Le Cun is being quoted at IEEE Spectrum
“Large language models have no idea of the underlying reality that language describes,” he said, adding that most human knowledge is nonlinguistic. “Those systems generate text that sounds fine, grammatically, semantically, but they don’t really have some sort of objective other than just satisfying statistical consistency with the prompt.”
Humans operate on a lot of knowledge that is never written down, such as customs, beliefs, or practices within a community that are acquired through observation or experience. And a skilled craftsperson may have tacit knowledge of their craft that is never written down.
I think “hallucination” is way too much an anthropomorphic concept – some LLM output is basically statistical nonsense (although I wouldn’t go as far as Michael Townsen Hicks…). Reasons for these kind of errors are manifold -reference divergence may be already in the data used for learning – data created by bots, conspiracy followers or even fraud science. The error may also originate from encoding or decoding routines.
I couldn’t find any further analogy with wrong human memory recall except the possibility that also human memory is influenced by probability as well. Otgar 2022 cites Calado 2020
The issue of whether repeated events can be implanted in memory has recently been addressed by Calado and colleagues (2020). In their experiment, they falsely told adult participants that they lost their cuddling toy several times while control participants were told that they only lost it once. Strikingly, they found that repeated false events were as easily inserted in memory as suggesting that the event happened once. So, this study not only showed that repeated events can be implanted, it raised doubts about the idea that repeated events might be harder to implant than single events
Das Gegenteil von “gut” ist “gut gemeint”
Podcast zur Wissenschaftsfreiheit mit Kritik an der Antisemitismus-Resolution des Bundestages – hörenswert!
Weitere Quellen
https://docs.google.com/forms/d/e/1FAIpQLScyErqrcDRrzrZ1EPuk6iX9x10g8JrwishN2rlnAhRyYZQwPg/viewform
Fake degrees from fake universities
Unfortunately there are so many lies, there is so much fraud in science. And science is a big business according to a new report
According to Hesselbäck, in Sweden it’s not illegal for a job applicant to submit qualifications from a fake university, although it is a crime to forge an academic transcript or degree from a legitimate university. The relevant laws vary by country.
Occasionally I had doubts on MD/PhD or professoral titles shown on badges while it is particular annoying if legal and illegal titles are even mixing…
In Norway, he adds, it is a criminal offence to submit credentials from a fake university, but the burden of proof has to be strong. At the federal level, the United States does not explicitly prohibit the issuing, holding or advertising of bogus degrees, although some states have laws banning them. In practice, individual holders of phoney medical degrees have surrendered their licences or been removed from their positions.
elife delisted
Elife is one of the most interesting scientific journals with a full history at Wikipedia. The Elife board introduced in 2022 that
From next year, eLife is eliminating accept/reject decisions after peer review, instead focusing on public reviews and assessments of preprint
with the unfortunate but foreseeable consequence that Elife now does not get anymore an impact factor
Clarivate, the provider of the Web of Science platform, said it would not provide impact factors to journals that publish papers not recommended for publication by reviewers.
I don’t care about impact factors. I also do not care about Clarivate or any other Private-Equity-company as we don’t need this kind of business in science. Elife however will loose it’s value in particular as system still has some flaws.
DeevyBee commented already about them a year ago
there is a fatal flaw in the new model, which is that it still relies on editors to decide which papers go forward to review, using a method that will do nothing to reduce the tendency to hype and the consequent publication bias that ensues. I blogged about this a year ago, and suggested a simple solution, which is for the editors to adopt ‘results-blind’ review when triaging papers. This is an idea that has been around at least since 1976 (Mahoney, 1976) which has had a resurgence in popularity in recent years, with growing awareness of the dangers of publication bias (Locasio, 2017). The idea is that editorial decisions should be made based on whether the authors had identified an interesting question and whether their methods were adequate to give a definitive answer to that question.
So the idea is that the editors get the title and a modified abstract with no author names and without results.
Scientific integrity is now included in the Helsinki Declaration
JAMA has a new revision of the Helsinki Declaration. Compared to the 2013 version there is now a new chapter on scientific integrity
Scientific integrity is essential in the conduct of medical research involving human participants. Involved individuals, teams, and organizations must never engage in research misconduct.
Additional details can be found in an Editor’s note and my comments are at Retraction Watch.
Hannah Arendt: Über Palästina. Zwei bisher unbekannte Texte von 1958.
(Gert Scobel ist im übrigen der einzige dem ich bisher einer Einladung zu einer TV Podiumsdiskussion gefolgt bin)
Comparatively trivial
Nature has a short report about historical peer reviews including a link to the Referee Report of Dorothy Crowfoot Hodgkin about the 1954 Watson & Crick complementary paper (not the 1953 Watson & Crick double helix paper).
And here is Fig 5 and Fig 6 of the paper under review. So did Watson & Crick follow her advice?
I don’t think so.
Risks you should be aware of as study participant
I wrote about this about this basically 15 years ago
Confidentiality has been seen in the past as a fundamental ethical principle in health care and breaching confidentiality is usually a reason for disciplinary action. It has been assigned such a great value because it directly originates from the patient’s autonomy to control his or her own life […] Two types of re-identification are possible: the “Netflix” type and the “profiling” type.
There is a new Cell paper that builds a “profiling” attack using even single-cell gene expression data only
we demonstrate that individuals in single-cell gene expression datasets are vulnerable to linking attacks, where attackers can infer their sensitive phenotypic information using publicly available tissue or cell-type-specific expression quantitative trait loci (eQTLs) information.
So this should be included in informed consent forms also.
AI hallucination
News article and paper showing
bigger AI chatbots more inclined to spew nonsense — and people don’t always realize.
and some solutions
various emerging techniques should help to create chatbots that bullshit less, or that can, at least, be prodded to disclose when they are not confident in their answers. But some hallucinatory behaviours might get worse before they get better.
Remarkable : I don’t want to be part of this scene anymore
From the creator of wordfreq
Generative AI has polluted the data
I don’t think anyone has reliable information about post-2021 language usage by humans.
The open Web (via OSCAR) was one of wordfreq’s data sources. Now the Web at large is full of slop generated by large language models, written by no one to communicate nothing. Including this slop in the data skews the word frequencies.