- Also known as deep neural networks, the AI involves throwing massive amounts of data at a neural network to train the system to understand things like speech and images.
- Gamalon, a new startup coming out of stealth on Tuesday, claims it’s developed an AI technique that requires training the system on only a few pieces data while achieving the same levels of accuracy as the exhaustively-trained neural networks.
- By reducing the amount of training data, this also reduces the ridiculous amounts of computing power required by deep learning systems.
- “It’s going to get worse training deep learning models,” Gamalon founder and CEO Ben Vigoda said.
- Gamalon has pitted its techniques in image recognition against Google’s TensorFlow using the tech giant’s deep learning-powered “Quick, Draw” program.
Deep learning, a branch of AI, has blown up in the tech industry, but new AI systems are coming online that promise big bumps in efficiency.
@Forbes: This startup claims its artificial intelligence is 100x more efficient than Google’s
Quote of the Day
It’s better to make mistakes while trying something new than not try anything at all.