4 ms·
Frankly, AI is going to have a hard time being open for several reasons: - The computing resources required to train models are not distributed - Training dat
by smoldesu 5y ago
Frankly, AI is going to have a hard time being open for several reasons:
- The computing resources required to train models are not distributed
- Training data will often contain licensed material
- Our digital concept of 'open' revolves around transparency, which is not readily available with conventional AI
That's not to say that we should give up on open efforts in the field, but we're still deep in the experimentation/research phase of artificial intelligence. Copilot has been an excellent demonstration of how poorly suited ML is for prime-time use, even by developers.
- MichaelZuo 5y agoIndeed given these fundamental constraints, OpenAI is about as open as it can be short of simply giving away everything for free.
- shmageggy 5y ago> simply giving away everything for free Which is what DeepMind has done with the AlphaFold code (Apache licensed https://github.com/deepmind/alphafold https://github.com/deepmind/alphafold) and published model predictions (CC licensed at https://alphafold.ebi.ac.uk/ https://alphafold.ebi.ac.uk/). I guess they could publish the weights but that would probably be useless since nobody else would be running the exact same hardware.
- nmfisher 5y agoExcept, dictating who can and can’t use their models, or what you can and can’t say if they deign to give you access https://twitter.com/mtrc/status/1363893222993498117?s=20 https://twitter.com/mtrc/status/1363893222993498117?s=20
- toiletaccount 5y agodoes training data need a permissive license, or is sharing the training data the issue? honestly, im surprised there hasnt been a bigger fuss over data and who really owns it. maybe in 25-50 years we'll collectively bemoan how everyone was taken for a ride.