It would be great to have true freedom of information, but to do so requires understanding the difference between data, information, and knowledge--terms that are often used interchangeably. I believe that the best current view of the differences comes from the philosopher Luciano Floridi, who defines data as just a difference, semantic information as well formed data that has meaning that is true, and knowledge as information that a person believes and has a justification for believing.

The marks you are seeing on your screen as you read this are data, but because they are well-formed, and if we assume they are true, then they are also information. For this comment to be knowledge, you must believe it and also have some justification for believing it.

It is easy for our computers to manipulate data, but much harder for them to manipulate information, because to do so the manipulation must preserve the semantics of the data and the operations we wish to apply to them.

Knowledge is even harder because it has a personal psychological component.

Now imagine having true information freedom. For example, "Obama is a Muslim" is well-formed data with meaning, but that meaning is not true, so it is not information. True information freedom means that an information system would not contain such statements or would at least flag them as misinformation. Floridi points out that just as counterfeit money is not money, misinformation is likewise not information. That's hard to do, of course with our present technology. Such a system could also still contain valid information such as "Trump says that he believes that Obama is a Muslim" as long as Trump really said that and really believes it.

Finally, the definition of knowledge points out why it is hard to disseminate knowledge. Knowledge requires a person to believe and logically justify that belief. That is often difficult to do, since it must be done person by person.

By understanding these distinctions it is fairly clear that having lots of data is not enough. Instead we need to get better at making machines that help us manipulate information and also find social and cultural means to disseminate knowledge.

Note the current boom in deep learning has very little to do with information. Most of the deep nets are end to end systems, meaning data in and data out. Google Translate does a good job translating from one language to another, but it knows absolutely nothing about the meanings of either the input or the output.