Emerald AI Personal Prediction Neural Network

18:02
13.04.2025
mclander
22

In 2023, an international group of students and postgraduates at the Technical University named after George Asachi in Iași began researching the potential of personal predictive systems based on neural networks. Initially, the group's goal was to prove that reliable predictions of environmental behavior are impossible. However, two out of five models used in the initial experiment produced unexpected results of 67.1% and 74.5%, respectively. The models were able to predict exam scores, personal impressions of trips, visits to the cinema, interview results, personal relationships, and even responses to grant requests. Predictions were made in a closed video form for the participant, and the result was considered after some time to prevent feedback.

For collecting information for training and functioning of the models, data was used from social networks, university information systems, corporate and scientific group websites, and news outlets, both online and print. The neural network estimated the probability of receiving grants to continue research at 80-90%. And grants were indeed obtained. Moreover, participants from the EU, primarily from Portugal and the United Kingdom, joined the research.

In the February 2025 issue of the leading Portuguese academic journal "La sistemoj neŭronaj," an article was published about the experiment's achievements. The network, named Emerald AI, managed to achieve nearly 80% successful predictions related to the experiment's participants. In reality, it reached 77.3%, which is still commendable.

Of course, this is provided the question was correctly formulated. Questions like "Should I take an umbrella?" showed a correlation close to weather forecasts (41.3%), which rather speaks in favor of the neural network and its methods of information gathering. Questions like "Will I be happy when I find out the result of the Steaua-Borussia game?" showed that Emerald predicts sports events moderately (53.3%). Nevertheless, financial advice provided by the AI turned out to be correct in 82.8% of cases.

Currently, researchers have proposed offering test access to the neural network to students and university staff from countries participating in the experiment. To do so, it is necessary to:

to be a student, graduate student, teacher, or researcher at an accredited university in one of the EU countries or the UK
to have an active profile on Facebook*, the profile must have been registered no later than January 1, 2021, and have at least 10 posts per year throughout the entire time
to have an active profile on LinkedIn** registered no later than 2022

After moderation, participants gain access to the Telegram bot and a token for access. Communication and the bot are only possible in one of the official languages of the country where the participant's university is located.

To participate, you need to send a request to the email emerald-beta.3@asashi.iasa.edu indicating in free form:

your university
full name
age
faculty
position/year of study
country of the university
language (one of the official languages for the country where the university is accredited)
FB profile
LinkedIn profile

A photo of the student ID must be attached to the email.

I managed to get only the Romanian bot. In my opinion, the predictions are quite vague. It is quite difficult to get a definite answer of "yes" or "no".

Regarding the salary increase, the neural network gave a positive forecast within a year. On relationships, it answered vaguely, which is not surprising – I have been married for 11 years. On the question about crypto, it suggested investing in agricultural startups.

Below is the dialogue in Romanian, with an approximate translation: