Major developments continue to take place in the artificial intelligence industry. Facebook’s Jerome Pesenti thinks the current model of deep learning is reaching its limits, however.
Dozens of companies are in the process of exploring the potential of artificial intelligence.
The Costs of Artificial Intelligence Research
Virtually all of these companies have scientists and engineers pushing the boundaries of deep learning.
The development of new algorithms has allowed for some intriguing insights and developments over the years.
Unfortunately it would appear that the current strategy involving deep learning may hit a glass ceiling sooner rather than later.
Those are the findings of Jerome Pesenti, head of artificial intelligence at Facebook.
In a recent interview with Wired, Pesenti acknowledges how deep learning and current artificial intelligence have severe limitations.
Achieving human intelligence, while still an attainable goal, will not happen any time soon.
Thankfully, there is still progress being made to address some limitations.
Taking into account how the artificial intelligence space is still evolving and growing in 2019, there are still millions of options left unexplored.
One aspect no one can ignore is how the compute power required to research advanced AI continues to increase twofold every three years or so.
Pesenti confirms this “problem’ exists, and highlights the need for scaling if any more progress is to be made.
At the same time, he is convinced the rate of progress for advanced artificial intelligence is not sustainable through this model.
Rising costs make it rather unattractive to conduct these levels of experiments today.