Same is true of many other AI assistants. They’re really neat as a technical exercise, or for a bit of fun for 10-20 minutes. But when it comes to folding them into workflows, the utility is harder to grasp.
Couple that with the extreme energy requirements of these systems, the worries about where the training data comes from, plus the fact that it feels like every single corporation is just flailing around “AI” because they see dollar signs… I’m pretty over it.
Same is true of many other AI assistants. They’re really neat as a technical exercise, or for a bit of fun for 10-20 minutes. But when it comes to folding them into workflows, the utility is harder to grasp.
Couple that with the extreme energy requirements of these systems, the worries about where the training data comes from, plus the fact that it feels like every single corporation is just flailing around “AI” because they see dollar signs… I’m pretty over it.
Training GPT-3 (a model from 3 years ago) from scratch requires as much energy as is spent to raise two cows for meat.