An ocean of information

The internet is something, which throughout human history was not even conceivable: an ocean of information. Where “ocean” does not only refer to the vast amount of data and its global nature, but also to its fluidity: it can be accessed extremely easily (Big Data guys talk of “data lakes”). This ocean is the ecosystem, where large AI Models the like of GPT chat emerge, like the first simple forms of live emerged in the physical oceans. While Chat GPT was trained on large amounts of text in the web, future multi modal models will likely be trained also on video, sound and other data (e.g. structured sensor–data generated by cars, robots etc.). It seems no longer too far fetched, to think of models, that digest virtually the entire internet.

 

But as these models do not only digest the internet for training, but will also increasingly create additional data and add it to the internet, this might lead to highly dynamic evolutions.

 

Like in nature, it seems easier to start evolving in the fluid data-ocean of the internet. Stepping on land (robotics, hardware) is harder and comes later, but might one eventually lead further.