{"id":17771,"date":"2023-05-06T15:11:53","date_gmt":"2023-05-06T15:11:53","guid":{"rendered":"https:\/\/www.searchenginejournal.com\/leaked-google-memo-admits-defeat-by-open-source-ai\/486290\/"},"modified":"2023-05-06T15:11:53","modified_gmt":"2023-05-06T15:11:53","slug":"leaked-google-memo-admits-defeat-by-open-source-ai-via-sejournal-martinibuster","status":"publish","type":"post","link":"https:\/\/marketingnewsbox.com\/?p=17771","title":{"rendered":"Leaked Google Memo Admits Defeat By Open Source AI via @sejournal, @martinibuster"},"content":{"rendered":"

A leaked Google memo offers a point by point summary of why Google is losing to open source AI and suggests a path back to dominance and owning the platform.<\/p>\n

The memo opens by acknowledging their competitor was never OpenAI and was always going to be Open Source.<\/p>\n

Cannot Compete Against Open Source<\/h2>\n

Further, they admit that they are not positioned in any way to compete against open source, acknowledging that they have already lost the struggle for AI dominance.<\/p>\n

They wrote:<\/strong><\/p>\n

\n

\u201cWe\u2019ve done a lot of looking over our shoulders at OpenAI. Who will cross the next milestone? What will the next move be?<\/p>\n

But the uncomfortable truth is, we aren\u2019t positioned to win this arms race and neither is OpenAI. While we\u2019ve been squabbling, a third faction has been quietly eating our lunch.<\/p>\n

I\u2019m talking, of course, about open source.<\/p>\n

Plainly put, they are lapping us. Things we consider \u201cmajor open problems\u201d are solved and in people\u2019s hands today.\u201d<\/p>\n<\/blockquote>\n

The bulk of the memo is spent describing how Google is outplayed by open source.<\/p>\n

And even though Google has a slight advantage over open source, the author of the memo acknowledges that it is slipping away and will never return.<\/p>\n

The self-analysis of the metaphoric cards they\u2019ve dealt themselves is considerably downbeat:<\/p>\n

\n

\u201cWhile our models still hold a slight edge in terms of quality, the gap is closing astonishingly quickly.<\/p>\n

Open-source models are faster, more customizable, more private, and pound-for-pound more capable.<\/p>\n

They are doing things with $100 and 13B params that we struggle with at $10M and 540B.<\/p>\n

And they are doing so in weeks, not months.\u201d<\/p>\n<\/blockquote>\n

Large Language Model Size is Not an Advantage<\/h2>\n

Perhaps the most chilling realization expressed in the memo is Google\u2019s size is no longer an advantage.<\/p>\n

The outlandishly large size of their models are now seen as disadvantages and not in any way the insurmountable advantage they thought them to be.<\/p>\n

The leaked memo lists a series of events that signal Google\u2019s (and OpenAI\u2019s) control of AI may rapidly be over.<\/p>\n

It recounts that barely a month ago, in March 2023, the open source community obtained a leaked open source model large language model developed by Meta called LLaMA.<\/p>\n

Within days and weeks the global open source community developed all the building parts necessary to create Bard and ChatGPT clones.<\/p>\n

Sophisticated steps such as instruction tuning and reinforcement learning from human feedback (RLHF) were quickly replicated by the global open source community, on the cheap no less.<\/p>\n