Cyborg Psychology 2
On the genesis of the psychological test. A workshop report
If ChatGPT is currently on everyone's lips, it's reminiscent of late 1980s debates when people were convinced artificial intelligence would do away with humans. And then, as now, an essential point is being ignored, namely the extent to which the new technologies open up the possibilities of new applications and how useful they can be.
We presented a small demonstration project of such possibilities in the form of a Cyborg Psychology testing application in our previous Ex Nihilo posting on February 14th. As a follow-up, it was written by my son Philipp Burckhardt and myself and is a product of our expertise. Philipp is the Director of e-Learning, Analytics, and Technology at Carnegie Mellon University, where he’s dedicated his expertise in artificial intelligence programming to developing educational applications for adult learners, primarily medical clinicians and researchers. With his Integrated Statistics Learning Environment (ISLE) program, he’s brought into the world a software for online teaching, which has found use at various American universities, but also in adult further education (such as in a large project with ModRNA).
The basic premise behind the construction of this little psycho-test, which is extremely cheerful, is that its planning wasn’t outsourced to artificial intelligence but is a combination of already available techniques. Structurally, that is to say, the creation of such a product primarily entails an effort of curation and planning — and then you hand the technical crafting of the application over to AI (which, de facto, is less an Artificial Intelligence than its an understanding of the world augmented via machine learning that is then transferred to the Machine).
For example, the avatar's voice is based on software from the company Murf AI. Here, the intervening parameters are still limited at the moment. First, you select a voice — and then you can introduce pauses and accentuations. The result is a relatively soberly delivered text that would do credit to a newscaster but which exorcises any emotional overtones, the uncertainties (the ahs and oohs) of a human speaker. And probably these practices (similar to the music industry, where the monotony of rhythm machines was addressed with a humanizer-function) will be the next challenge.
The avatar, in turn, is a product of the company D-ID — and its task is to give a body to the voice. Because there isn't a need to use a computer-generated voice here, the user would be free to use their own voice as well. Yes, you could even imagine a scenario where a person uses their own voice to set their avatar of themselves in motion. This is one of the services this company offers: the creation of doubles based on the user's natural person. As you can see from the avatar's amazing hand movements, the avatar's movements are based on human actors' movement studies — this is, in fact, what the 18th century produced with its tableaux vivants, an image that has come to life, so to speak. If you keep the quality of the avatar in mind, you can imagine that news or TV anchors being replaced by such avatars in the future — which already has a certain rationale because you can position any such avatar in front of a green screen and link it arbitrarily with all kinds of conceivable video material - in this case, a double of itself and an atmosphere of Storyblocks.
For the images of the psycho test, DALL-E 2 was used, which, like Chat GPT, goes back to the Open AI project.
Here, too, one does not automatically arrive at a result but is confronted with a certain amount of curation and post-processing. Because DALL-E allows certain stylistic specifications (here, a comic style), it is straightforward to come to a desirable result. Finally, Chat GPT 3 was also used — but only to vary the given questions of the psycho test. By doing this, the results of the psycho test are purely combinatorially individualized; less than a dozen duplicates will come out of a million tests.
Overall, it can be said that the work on this little psycho-test moved into areas that have comparatively little to do with artificial intelligence. Apart from combining available techniques and then employing them in the most intelligent way possible (much like a director directs his actors in front of the camera), the main task is to draft a script and implement it programmatically. Or, to put it another way, this is where imagination and programming come in.
Fields of application
Now, it may well be conceivable that such AI-augmented solutions in the future can replace certain activities — however, it must be added that this will be primarily those activities where the human behavioral repertoire will have predominantly become android replaceable. And because this topic requires a much more thorough treatment, I'll refrain from discussing this fundamental question here. Instead, we'll consider the possible applications of this kind of AI-augmented technology, accompanied by an increase in the services' quality. One application area that has engaged my son has been the possibility that students could be provided with such an assistant as they pursue highly individualized learning paths. Exemplar, this assistant could answer perenially recurring questions while acting as a motivational coach. As the psycho-test shows, a program interface recording the interactions of a user can already draw psychological conclusions from even a few interactions - which makes it possible to think of in terms of an individualized form of assistance that can also provide support over the long term (either by acting as an aide-mémoire or by responding to particular idiosyncrasies of the student). This is precisely where we can see the decisive thematic leitmotif already inherent in producing this kind of assistant for human beings. Because the knowledge of how to acquire knowledge requires knowledge of human nature — and a distinctive programmatic finesse. Neither is it, nor can it ever be the product of an artificial intelligence. In this sense, what customizing means today is effectively an open question — what would it be like to ask how and in what way a student can be motivated and elevated into a higher level of education? If the modern education system was based on a cohort system and a form of uniformity (the class), a movement of thought is needed now that focuses on the individual (in all their distinctive peculiarities). This consideration alone shows how out of place the speculation is that artificial intelligence could eventually supplant human beings. What can be replaced is machine-like android behavior, as knowledge of the psychological mechanisms becomes ever more precious.
If you want to do the psycho-test yourself, here is the Link