- Introduction to Neurospark technology
- Training AI with brain cells
- Limitations and prospects
Swiss technology firm FinalSpark has launched a program allowing scientists to rent access to ‘biocomputers’ made from human brain cells for $500 a month. These biocomputers are designed to offer a more efficient and low-energy alternative for training artificial intelligence (AI) models.
Introduction to Neurospark technology
FinalSpark claims that using organic material, like human brain cells, to train AI could be up to 100,000 times more efficient than traditional silicon-based systems. The company's founder, Fred Jordan, says the technology, called Neurospark, is unique in offering biocomputers built from clusters of human brain cells, known as organoids. The program provides real-time online access to these living biocomputers, opening up new possibilities in AI development.
Training AI with brain cells
Training AI using these biocomputers involves different techniques than usual. Typically, AI models are trained using numerical rewards to guide them. However, when using brain cells, scientists rely on biological methods. They provide dopamine, a chemical linked to rewards in the brain, to encourage learning and use electrical signals as negative feedback. This process better reflects how real human brains learn and adapt.
Limitations and prospects
The only limitation is that these biocomputers have a short lifespan, usually lasting around 100 days. After that, the AI system needs to be transferred to a new cluster, or it will no longer function. While this limitation exists, it hints at future possibilities, such as creating AI with longer lifespans that might one day align more closely with human experiences.
FinalSpark's program launch opens up new horizons in AI training, offering more efficient and adaptive models. While there are certain limitations, the technology shows promise for further development and improvement.
