Unsupervised learning with unstructured data, why it changes everything
Recently I red an interview with Demis Hassabis from Deepmind, a company that Google
acquired. The co-founder mentioned in that interview his main focus, "It’s also called unsupervised learning— you just give it data and it learns by itself what to do with it, what the structure is, what the insights are. We are only interested in that kind of AI.". The progress that has been made is published in Nature , read The Verge for the non-scientific version. The exciting thing as described in the mentioned articles is that the system does not "know" what a certain Atari game was. It let itself be trained by playing where a high(er) score was a reward. Just as with humans. In that way the system learned itself to play and win most old Atari games. With technology things becoming exciting when they become "general-purpose" , when you can use one system for different tasks and assignments. The PC is the ultimate example but also the smartphone. In software terms Deepmind/Google are trying to build a system that can learn, adapt and handle multiple situations. This first step is small, this application can only learn to play Atari games. But the next steps are more advanced selflearning systems in other domains like medicine, finance, logistics etc. After that systems will start combining their learning experience in various domains.
Google Now is an obvious candidate for this technology, it will become a system that learns to "play" us, let us buy things and let us behave in a way that Google wants us to. The most negative scenario. Formulated in positive terms it will be able to perfectly anticipate our behaviour, assist us in the best way possible, provide services which perfectly satisfy our needs. All of the above is not new for people who like to be informed about this subject. However for the majority of politicians, managers, citizens etc. a computer program is still something that can only do what it is programmed to do. It's behaviour and actions are predetermined, controllable and auditable.Our laws, checks and balances in society with regard to technology are based on that premiss. If that premiss is no longer true. It changes everything. There should be more discussion about this in society and politics. Instead people focus on the rather limited subject of privacy and personal data. That's the reason i keep stressing the importance of this and related subjects in my blogs. I am in good company with several technology guru's telling us the same thing. However in parliament and politics it's still too quiet.
acquired. The co-founder mentioned in that interview his main focus, "It’s also called unsupervised learning— you just give it data and it learns by itself what to do with it, what the structure is, what the insights are. We are only interested in that kind of AI.". The progress that has been made is published in Nature , read The Verge for the non-scientific version. The exciting thing as described in the mentioned articles is that the system does not "know" what a certain Atari game was. It let itself be trained by playing where a high(er) score was a reward. Just as with humans. In that way the system learned itself to play and win most old Atari games. With technology things becoming exciting when they become "general-purpose" , when you can use one system for different tasks and assignments. The PC is the ultimate example but also the smartphone. In software terms Deepmind/Google are trying to build a system that can learn, adapt and handle multiple situations. This first step is small, this application can only learn to play Atari games. But the next steps are more advanced selflearning systems in other domains like medicine, finance, logistics etc. After that systems will start combining their learning experience in various domains.
Google Now is an obvious candidate for this technology, it will become a system that learns to "play" us, let us buy things and let us behave in a way that Google wants us to. The most negative scenario. Formulated in positive terms it will be able to perfectly anticipate our behaviour, assist us in the best way possible, provide services which perfectly satisfy our needs. All of the above is not new for people who like to be informed about this subject. However for the majority of politicians, managers, citizens etc. a computer program is still something that can only do what it is programmed to do. It's behaviour and actions are predetermined, controllable and auditable.Our laws, checks and balances in society with regard to technology are based on that premiss. If that premiss is no longer true. It changes everything. There should be more discussion about this in society and politics. Instead people focus on the rather limited subject of privacy and personal data. That's the reason i keep stressing the importance of this and related subjects in my blogs. I am in good company with several technology guru's telling us the same thing. However in parliament and politics it's still too quiet.
Reacties