ThirdAI, a Houston-based company that develops tools for deep learning technology, raised $6 million in seed financing.The investment was led by Neotribe Ventures. Cervin Ventures. and Firebolt Ventures. It will be used to hire additional staff and invest in computing resources. Anshumali Shrivastava (third AI co-founder, CEO) told TechCrunch.Shrivastava, a mathematician, was always curious about artificial intelligence and machine-learning. He especially wanted to rethink how AI could be made more efficient. He was studying at Rice University when he began to explore deep learning. ThirdAI was founded in April by him and some Rice graduate students.Shrivastava stated that ThirdAIs technology is a smarter way to deep learn. It uses its algorithm and software innovations, to make general-purpose central processor units (CPU) more efficient than graphics processing units to train large neural networks. Many companies abandoned CPUs many years ago to use graphics processing units that can render high-resolution video and images simultaneously. He said that the downside to graphics processing units is their limited memory, which can cause users to hit a bottleneck when trying develop AI.Deep learning was a landscape that looked back at technology from the 1980s. A majority of the market (some 80%) used graphics processing units. They also invested in expensive hardware and engineers, then waited for the magic of AI.His team and he looked at the future of AI and sought to find a cost-saving solution to GPUs. Instead, their algorithm uses CPUs that don't require special acceleration hardware.Swaroop Kittu Koluri, founder and managing director of Neotribe, stated that this type technology is still in its infancy. He said that current methods are slow, costly, and laborious. For example, if a company runs language models that use more memory, they will have problems.ThirdAI is where it all comes together, Kolluri stated. This is why we decided to invest. ThirdAI will allow anyone to do this, which is going be a game-changer. There is no limit to the technology available in deep learning.AI is at an early stage and can solve many of the most difficult problems like seismic processing. However, he also notes that there are concerns about the climate impacts of AI models.Shrivastava stated that deep learning models can cost more than five cars over a lifetime. These are important considerations as we work to scale AI.
