5 ms·

MS bet on OpenAI is one of the biggest tech success stories of this decade for sure. Now everyone wants a pie of it but it seems that they are way behind for so

MS bet on OpenAI is one of the biggest tech success stories of this decade for sure. Now everyone wants a pie of it but it seems that they are way behind for some reason (for example Google Bard is nothing compared to OpenAI implementation)

red_admiral 3y ago
Yes, it's quite ironic isn't it that the big AI progress is being made not by the cool kids at Alphabet and Meta and the like, but in a place backed by "boring" old Microsoft. Whatever next - Bing becoming actually useful, perhaps?
- threeseed 3y ago
  The fundamental research in the AI space (that OpenAI depends so much on) originally came from Alphabet and Meta. They just never thought to productise it.
  - theGnuMe 3y ago
    OpenAI has taken it to the next level though.
  - simpleladle 3y ago
    They did; perhaps not in ways that were immediately obvious to the general public. Back in 2014, they created attention to improve how RNNs handled long sequences, an effort driven by google translate. Same goes for when they created the transformer architecture. The first product to see an improvement was also translate. They simply never exposed the bare language models to the public a la OpenAI style with GPT3
salawat 3y ago
Microsoft bet on Github to get the requisite training corpus needed for OpenAI to even achieve relevancy. If there had been negotiation for permission to train on others code like there should have been, things would have gone differently.
- brookst 3y ago
  While GitHub copilot is a big deal, it is a tiny tiny tiny portion of the promise of LLMs. There will be far more economic disruption from LLMs as productivity tools than as coding tools.
threeseed 3y ago
> Google Bard is nothing compared to OpenAI implementation That's simply not true based on the comparison done by MKBHD [1]. Bard has the requirement that it needs to deliver an accurate and quality result above everything else. This makes it more useful but far less interesting and compelling than OpenAI. [1] https://www.youtube.com/watch?v=HKvo6acp5aI https://www.youtube.com/watch?v=HKvo6acp5aI
- int_19h 3y ago
  It's true based on its demonstrable lack of ability to solve tasks that even ChatGPT (3.5) can handle. Sure, you can cherry-pick some test suite to come up with a number that says otherwise - just like there are people claiming that LLaMA can match ChatGPT - but this is more of a "squinting at it right" kind of thing. I don't know of anyone using Bard over ChatGPT in any application where quality of the output matters.

red_admiral 3y ago

Yes, it's quite ironic isn't it that the big AI progress is being made not by the cool kids at Alphabet and Meta and the like, but in a place backed by "boring" old Microsoft. Whatever next - Bing becoming actually useful, perhaps?

threeseed 3y ago
The fundamental research in the AI space (that OpenAI depends so much on) originally came from Alphabet and Meta. They just never thought to productise it.
- theGnuMe 3y ago
  OpenAI has taken it to the next level though.
- simpleladle 3y ago
  They did; perhaps not in ways that were immediately obvious to the general public. Back in 2014, they created attention to improve how RNNs handled long sequences, an effort driven by google translate. Same goes for when they created the transformer architecture. The first product to see an improvement was also translate. They simply never exposed the bare language models to the public a la OpenAI style with GPT3

salawat 3y ago

Microsoft bet on Github to get the requisite training corpus needed for OpenAI to even achieve relevancy. If there had been negotiation for permission to train on others code like there should have been, things would have gone differently.

brookst 3y ago
While GitHub copilot is a big deal, it is a tiny tiny tiny portion of the promise of LLMs. There will be far more economic disruption from LLMs as productivity tools than as coding tools.

threeseed 3y ago

> Google Bard is nothing compared to OpenAI implementation That's simply not true based on the comparison done by MKBHD [1]. Bard has the requirement that it needs to deliver an accurate and quality result above everything else. This makes it more useful but far less interesting and compelling than OpenAI. [1] https://www.youtube.com/watch?v=HKvo6acp5aI https://www.youtube.com/watch?v=HKvo6acp5aI

int_19h 3y ago
It's true based on its demonstrable lack of ability to solve tasks that even ChatGPT (3.5) can handle. Sure, you can cherry-pick some test suite to come up with a number that says otherwise - just like there are people claiming that LLaMA can match ChatGPT - but this is more of a "squinting at it right" kind of thing. I don't know of anyone using Bard over ChatGPT in any application where quality of the output matters.