Listen now | You probably were wondering how long into the new year before a bunch of focus and attention were placed at the efforts of Hugging Face. You won’t have to wait any longer as this missive will dig into BigCode and some other efforts to use AI to build out code generating systems [
My sense, although based upon a much smaller dataset than yours is that AlphaCode (which I have evaluated previously) is remarkably impressive. While I carry it as a bias, it seems to me, that almost without exception, the Google approach to systematically automate is still the leading edge. I used to provide a training primer in an old job to explain the path from GFS to what became Hadoop. Some of the folks using Hadoop behaved as if it were some amazing insight. Teaching people it was merely a gift from Google GFS version 1.0 was sobering for many of them. Many organizations still have legacy ties to hierarchical data structures (mostly mainframe) and of course a dizzying set of relational models (Oracle, SQL Server, etal). Concepts in AI seem to me to REQUIRE a different scale and a different, more parallel way of thinking.
I believe the action to leave behind hierarchical and relational data systems and move headlong into a different organizational pattern is the secret sauce that makes the slow but steady breakthroughs we will continue to see with DeepMind in general and AlphaFold & AlphaCode almost inevitable in my opinion.
I would be interested in your thoughts about where these other systems fit in comparison. I am sure I would learn something I have not considered to date.
Hi Nels,
My sense, although based upon a much smaller dataset than yours is that AlphaCode (which I have evaluated previously) is remarkably impressive. While I carry it as a bias, it seems to me, that almost without exception, the Google approach to systematically automate is still the leading edge. I used to provide a training primer in an old job to explain the path from GFS to what became Hadoop. Some of the folks using Hadoop behaved as if it were some amazing insight. Teaching people it was merely a gift from Google GFS version 1.0 was sobering for many of them. Many organizations still have legacy ties to hierarchical data structures (mostly mainframe) and of course a dizzying set of relational models (Oracle, SQL Server, etal). Concepts in AI seem to me to REQUIRE a different scale and a different, more parallel way of thinking.
I believe the action to leave behind hierarchical and relational data systems and move headlong into a different organizational pattern is the secret sauce that makes the slow but steady breakthroughs we will continue to see with DeepMind in general and AlphaFold & AlphaCode almost inevitable in my opinion.
I would be interested in your thoughts about where these other systems fit in comparison. I am sure I would learn something I have not considered to date.