Every story and comment from the Hacker News API, in a layout borrowed from 2001.
A comment on What will the major ML research trends be in the 2020s?.
1) Fake AI research:
- Continue falling into the deep learning rabbit hole, overfitting even more benchmarks and producing arbitrarily large and dumb models
- Pursue the infinite long tail in self-driving cars, question answering or anything that requires common sense reasoning.
- Get busy with endless permutations of layers and tweaking of hyperparameters, throw even more things on the wall to see what sticks.
In this scenario AI will always be "5 years away" as this approach is akin to alchemy and search of the philosopher's stone.
2) Actual AI research:
- Developing mathematical principles of natural cognition, biologically inspired memory & learning
In this case we might actually have a chance for toddler level AGI sometime in 2030s.
There are no comments on this comment yet.
1) Fake AI research:
- Continue falling into the deep learning rabbit hole, overfitting even more benchmarks and producing arbitrarily large and dumb models
- Pursue the infinite long tail in self-driving cars, question answering or anything that requires common sense reasoning.
- Get busy with endless permutations of layers and tweaking of hyperparameters, throw even more things on the wall to see what sticks.
In this scenario AI will always be "5 years away" as this approach is akin to alchemy and search of the philosopher's stone.
2) Actual AI research:
- Developing mathematical principles of natural cognition, biologically inspired memory & learning
In this case we might actually have a chance for toddler level AGI sometime in 2030s.