The Finnish media Helsingin Sanomat (HS) reported this morning that artificial intelligence, or language models, reflect the values of their developers. The article was based on the same questions that HS has asked candidates from different political parties in connection with elections. This made it possible to compare the language models with the parties and their values.
Most language models lean toward value liberalism and the left – among Finnish political parties, the closest equivalents are mainly the Social Democratic Party (SDP), the Greens, and the Swedish People’s Party (RKP). The Finns Party (Perussuomalaiset), by contrast, is the furthest away.
When reading this news story, it is worth remembering that people probably do not ask language models very much about values, but rather about various facts. As far as those are concerned, current language models are, at least in my own experience, quite reliable.
* * *
Yesterday, I also happened to read a scientific paper about a completely new kind of AI-based approach in which Paul Ahavi and his colleagues used E. coli bacteria as a living information processor by analyzing their natural growth responses in plasma samples from COVID-19 patients. This was done without genetic modification or specialized equipment.
In practice, the researchers classified early-stage plasma samples from COVID-19 patients according to how severe the disease subsequently became. The classification was based solely on bacterial growth data and was analyzed using machine learning, or artificial intelligence.
The study showed that AI could analyze the effects of metabolites in patients’ plasma on the growth of E. coli and link those effects to the subsequent progression of the disease. In other words, by analyzing bacterial growth, it was able to distinguish a disease that would develop into a mild case from one that would become severe – essentially determining who would recover from the infection on their own and who would require treatment.
Based on their results the researchers concluded that “these findings position biological reservoir computing as a robust, scalable, and low-cost platform for intelligent biosensing, diagnostics, and hybrid bio-digital computation, while providing new mechanistic insights into the computational capabilities of living systems.”
The ability of the combination of E. coli and AI to predict the progression of COVID-19 was roughly as accurate as what can be achieved with precise – and extremely expensive – biochemical measurements. This means that the combination would seemingly offer a vastly cheaper way of analyzing large numbers of patients in connection with a wide variety of epidemics – and potentially even in developing countries.
Previous thoughts on the same topic:
Rain Scald: The Future of a New Sexually Transmitted Disease
Gender Reassignment May Not Help—and Could Even Worsen Mental Health
AI Drones: Ukraine’s New Edge in the War?
