The computer used a vast amount of data on past campaigns and calculated the winner of the 2020 election.
The company “Promobot” creates a new artificial intelligence Promobot V. 4 for use in robots. While the algorithm is being refined, it was tested on a set of data on the US presidential election and received the name of the most likely winner of the 2020 presidential race. Representatives of the company reported.
Neural networks analyzed an array of data with tens of thousands of social media posts, which included posts about two candidates-Donald Trump and Joe Biden. These were data on elections, results of open opinion polls, and news. All the information was compared with similar data on previous campaigns and received a forecast.
To solve the problem, the creators of artificial intelligence used multi-class classifiers based on a combination of various neural networks, including GPT-3, which was introduced a few months ago by OpenAI, created by Elon Musk. Classical machine learning algorithms processed tabular data, the developers explained.
“We have received a confident prediction that the current head of state, Donald Trump, will win the upcoming US presidential election in 2020”.
Promobot development Director Oleg Kivokurtsev noted that initially, the development of such a neural network was not planned. Still, when choosing a certain dataset, the developers managed to make a prediction. “Such a neural network could predict not only the results of elections but also waves of epidemics, such as the COVID-19 pandemic,” Kivokurtsev stressed.
Four years ago, during the 2016 US presidential race, Russian machine learning experts predicted that the new US President would be Donald Trump. Engineers used the model to recognize objects in the image — it assigned Trump the “President” and “ex-President” categories.