Inducing chatbots to write nonsense

1 Reply

titanium dental implant face profile technical pictureThe chatbot, ChatGPT developed by OpenAI, has been in the news recently and is the subject of much discussion in universities primarily because of its potential use by students to complete their coursework assignments but also the positive uses to which it might be applied. After last week’s invitations to edit two special issues in different journals on cosmetic dentistry and wire arc additive manufacturing (WAAM) [‘Wire arc additive manufacturing and cosmetic dentistry?‘ on February 8th, 2023], I did a little research in the scientific literature to find out if anyone had published research on using WAAM to make parts for cosmetic dentistry but found nothing.  I was not surprised – the level of precision achievable with WAAM is about 1 millimetre which would be insufficient for most applications in cosmetic dentistry.  Then, I signed up for a free trial with ChatGPT and conducted an experiment by asking it to write about wire arc additive manufacturing and cosmetic dentistry.  The chatbot produced 128 words about how WAAM is becoming increasing popular in cosmetic dentistry because of its accuracy and precision also because a wide range of materials can be used allowing a match to the colour and texture of teeth.  I repeated the experiment and the chatbot produced 142 different words, again stating that dental prostheses can be produced using WAAM with high precision and accuracy to match a patient’s existing teeth in colour so that restorations appear natural and undetectable.  In each case the six or seven sentences were well-written and included some facts that were used to construct a set of false statements, which superficially appeared reasonable; however, only a modicum of knowledge would be required to identify the fallacious rationale.  Some of my colleagues are already exploring incorporating the chatbot into students’ coursework by asking students to use it to generate a description of a technical topic and then asking them to critique its output in order to assess their understanding of the topic.  I expect chatbots will improve rapidly but for the moment it is relatively easy to induce them to write nonsense.

Bibliography

Li Y, Su C, Zhu J. Comprehensive review of wire arc additive manufacturing: Hardware system, physical process, monitoring, property characterization, application and future prospects. Results in Engineering,100330, 2021.

Image: www.authoritydental.org CC BY 2.0 downloaded from https://www.flickr.com/photos/dental-photos/50730990757

Wire arc additive manufacturing applied to cosmetic dentistry?

1 Reply

photograph of a flower for decorative purposes onlyLast weekend I sat down at my laptop to write this week’s post with only a vague idea of a topic. When I opened my laptop I was surprised to see two emails from a supposedly reputable commercial publisher inviting me to be a guest editor for two special issues of two different journals.  For two decades, I served as editor-in-chief of two international journals consecutively with only a short overlap so I am well-qualified to act as a guest editor.  However, the invitations related to cosmetic dentistry and wire arc additive manufacturing.  I know almost nothing about these two subjects so why was I receiving invitations from the editors of two journals to be a guest editor.  In collaboration with colleagues, I have published some papers recently on another form of additive manufacturing [see ‘If you don’t succeed try and try again‘ on September 29th 2021].  My Google Scholar profile shows that my two most highly cited papers relate to work performed thirty years ago on osseointegrated dental implants [see ‘Turning the screw in dentistry‘ on September 30th, 2020]; although on closer examination it would also reveal that I have published nothing since then on this subject.  I suspect that a poorly programmed algorithm was fooled by my eclectic and long publication record into issuing poorly targeted invitations rather than the academic editors exercising poor judgment.  At least, I hope that is what happened since the alternative is that journal editors are no longer exercising academic judgment (though it is obvious this is also happening given the incoherent reviews of manuscripts that editors too frequently pass on to authors probably without reading them).  I will treat these invitations as spam; however, others may see them as opportunities to create or expand ‘peer-review’ rings and put more ‘Rotten eggs in the store‘ [see post on November 30th, 2022].  The peer-review and publication system for scientific papers is clearly broken and one part of the solution is to remove commercial interests from the process.

We are ecosystem engineers

Leave a reply

Decorative photograph of common cuscusHumans have been ecosystem engineers since the Pleistocene, more than 12,000 years ago.  There is evidence of a tree-dwelling possum, the common cuscus, being introduced to the Solomon Islands from New Guinea more than 20,000 years ago as a game species [1].  The ecosystem is a complex system and there are unintended consequences of our engineering.  For instance, the burning forests and grasslands about 8,000 years ago changed reflectivity and absorption of heat in parts of Eurasia which altered the pattern of monsoons in India and parts of South East Asia.  The palaeobiologist, Thomas Halliday has suggested that we are such effective ecosystem engineers that is impossible to think about a pristine Earth unaffected by human biology and culture [2].  The challenge now is to re-engineer the ecosystem so that it remains habitable.  This involves handling the complexities of  the ecosystem, human society and their interactions.  The philosopher, Nabil Ahmed has written, in the context of his native Bangladesh, that it is impossible to differentiate between land and rivers, human population, grains and forests, politics and markets because they all coalesce as a single entity resulting from the legacy of interaction between politics and natural actors [3].  Everything is interconnected – more than we realise.

References

[1] Abate RS & Kronk EA, Climate change and indigenous peoples: the search for legal remedies Cheltenham UK: Edward Elgar, 2012.

[2] Halliday T, Otherlands: A world in the making, London: Allen Lane, 2022.

[3] Ahmed N, Entangled Earth, Third Text, 27:44-53, 2013.

Image: Exhibit in the Museo Civico di Storia Naturale di Genova, Via Brigata Liguria, 9, 16121, Genoa, Italy; by Daderot, CCO 1.0 licence

Label-free real-time tracking of individual bacterium

4 Replies

Images from the optical microscope showing the tracks of bacteria interacting with a surfaceAntimicrobial resistant (AMR) infections are already the third leading cause of death in the USA and are predicted to kill 50 million people per year by 2050.  It is the next pandemic starting already.  We have been using our capability to track nanoparticles in an optical microscope [see ‘Slow moving nanoparticles‘ on December 13th, 2017 and ‘Nano biomechanical engineering of agent delivery to cells‘ on December 15th, 2021] to track individual bacterium as they interact with surfaces to form biofilms.  Bacterial biofilms are complex colonies of bacteria that are highly resistant to antimicrobial agents and can cause life-threatening infections.  We have used our label-free, real-time tracking capabilities to explore the dynamics and adhesion of bacteria to surfaces and found that viable bacteria adhered to the surface but continue to move with rotary or sliding motions depending on the mechanics of their attachment to the surface.  Bacteria that were killed by contact with the surface did not move once they were attached to the surface.  The image shows examples of these motions from our paper published last month.  Our ability to detect these differences in the dynamics of bacteria will allow us to detect the onset of the formation of biofilms and to quantify the efficacy of antimicrobial surfaces and coatings.

Image: Figure 4 – Tracks (yellow lines) of the sections (purple circles) of four E. coli bacteria experiencing: (a) random diffusion above the surface; (b) rotary attachment; (c) lateral attachment; (d) static attachment. The dynamics of the four bacteria was monitored for approximately 20 s. The length of the scale bars is 5 μm. From Scientific Reports, 12:18146, 2022.

Source:

Giorgi F, Curran JM & Patterson EA, Real-time monitoring of the dynamics and interactions of bacteria and the early-stage formation of biofilms, Scientific Reports, 12:18146, 2022.