More Data Needed for Artificial Intelligence Training

Many people now believe that artificial intelligence technology is completely different from what we see in science fiction movies, and there is nothing to be afraid of. However, if you know that artificial intelligence has already digested everything we have and still craves for more, you will feel its terrifying aspect. As artificial intelligence companies build bigger and better models, people are starting to worry that there may not be enough data on the internet to train them. Some companies are looking for alternative sources of training data, such as video transcriptions and "synthetic data." The last mentioned content is generated by artificial intelligence, and no one knows what will happen if it is essentially self-trained. Early research agrees that models trained on artificially generated data will eventually lead to "model collapse." Some companies claim that they can create higher quality synthetic data, but they have not revealed much about the actual output. Dataology, founded by former Meta and Google DeepMind researcher Ali Mokos, is one of the organizations trying to find ways to train larger and smarter models with less data. These data training methods are often controversial, such as transcribing public YouTube videos. Researchers have observed that artificial intelligence will eventually face data depletion. Pablo Villalobos estimates that artificial intelligence will run out of available data in the next year or two, but he doesn't seem worried. "The biggest uncertainty is what kind of breakthroughs you will see." Alternatively, companies can stop trying to create those bigger and better models due to issues with training data storage and other problems like excessive energy consumption.