4 ms·

OpenAI made the first AI chat bot product that didn't suck. There is a critical mass of innovation and Open AI got their 9kg of Uranium-235 together first as re

by s_dev 3y ago

OpenAI made the first AI chat bot product that didn't suck. There is a critical mass of innovation and Open AI got their 9kg of Uranium-235 together first as reflected in the market. The real question isn't why did OpenAI get there first but why Google didn't, they had every advantage.

2OEH8eoCRo0 3y ago
> OpenAI made the first AI chat bot product that didn't suck. And let people try it for free. Non-techies were using it and the news was reporting on it. They nailed their launch.
- wongarsu 3y ago
  After horrible launches for GPT2, GPT3 and GPT3.5 where most hype and momentum was smothered by a long waitlist. ChatGPT was a 180 degree reversal from their usual release process and an instant hit.
- booleandilemma 3y ago
  I think this is the thing people like the guy linking to Google's paper on arxiv.org don't get. OpenAI actually delivered something.
andruby 3y ago
My armchair answer to that question is that Google isn’t a product company. They might have had similar or better technical AI capabilities, but they aren’t great at productizing it.
- sixothree 3y ago
  The one thing that most defines Google products is that each one must collect data from the user in a way that is unique to their other products. If it doesn’t do that it doesn’t get to exist.
- s_dev 3y ago
  >Google isn’t a product company. I disagree. Google does have great products, Google Maps, Search, GMail etc.
  - wongarsu 3y ago
    Survivorship bias. Any company that launches as many products as Google has to have a couple good ones by pure chance. And all products you mentioned were created two decades ago, when Google was a very different company.
  - osigurdson 3y ago
    Google products by launch date: Search: 1998 Gmail: 2004 Maps: 2005
  - moritonal 3y ago
    It has free Products. I would suggest that all three would struggle to act as individual paid products if they didn't cross-support each other.
fragmede 3y ago
Because keeping the user on Google, so they don't need to visit the actual page, means that people won't make web pages anymore. that was a real concern.
- s_dev 3y ago
  Steve Jobs always said in Apple if we don't cannibalize our own products someone else will eat them for us. So the answer is Google isn't willing to take risks like it used to.
  - wongarsu 3y ago
    > Google isn't willing to take risks like it used to Has Google ever taken a risk that endangered their ad business?
c0n5pir4cy 3y ago
Honestly it wasn't even having the first capable chat LLM. OpenAI released the first LLM that has interesting characteristics, based on an architecture they created to be good at this task (Generative Pre-trained Tranformers) and then iterated on improving those characteristics - It took them a while before they had a model capable of following instructions well. There was lots of interest in OpenAI when they released GPT-1 and much more when they released GPT-2. All the while this was going on Google slept - I think the reasons for this are varied and many, stretching into company culture, structure, processes etc. This was after researchers at Google created the Transformer architecture in the first place. Google seemed to panic suddenly when ChatGPT was released. I think this is when the people who could actually make change realised that ChatGPT could be a threat to Google in a number of ways - including for search market share.
benob 3y ago
Google had it in 2015 [1] [1] https://arxiv.org/abs/1506.05869 https://arxiv.org/abs/1506.05869

by s_dev 3y ago

2OEH8eoCRo0 3y ago

> OpenAI made the first AI chat bot product that didn't suck. And let people try it for free. Non-techies were using it and the news was reporting on it. They nailed their launch.

wongarsu 3y ago
After horrible launches for GPT2, GPT3 and GPT3.5 where most hype and momentum was smothered by a long waitlist. ChatGPT was a 180 degree reversal from their usual release process and an instant hit.
booleandilemma 3y ago
I think this is the thing people like the guy linking to Google's paper on arxiv.org don't get. OpenAI actually delivered something.

andruby 3y ago

My armchair answer to that question is that Google isn’t a product company. They might have had similar or better technical AI capabilities, but they aren’t great at productizing it.

sixothree 3y ago
The one thing that most defines Google products is that each one must collect data from the user in a way that is unique to their other products. If it doesn’t do that it doesn’t get to exist.
s_dev 3y ago
>Google isn’t a product company. I disagree. Google does have great products, Google Maps, Search, GMail etc.
- wongarsu 3y ago
  Survivorship bias. Any company that launches as many products as Google has to have a couple good ones by pure chance. And all products you mentioned were created two decades ago, when Google was a very different company.
- osigurdson 3y ago
  Google products by launch date: Search: 1998 Gmail: 2004 Maps: 2005
- moritonal 3y ago
  It has free Products. I would suggest that all three would struggle to act as individual paid products if they didn't cross-support each other.

fragmede 3y ago

Because keeping the user on Google, so they don't need to visit the actual page, means that people won't make web pages anymore. that was a real concern.

s_dev 3y ago
Steve Jobs always said in Apple if we don't cannibalize our own products someone else will eat them for us. So the answer is Google isn't willing to take risks like it used to.
- wongarsu 3y ago
  > Google isn't willing to take risks like it used to Has Google ever taken a risk that endangered their ad business?

c0n5pir4cy 3y ago

Honestly it wasn't even having the first capable chat LLM. OpenAI released the first LLM that has interesting characteristics, based on an architecture they created to be good at this task (Generative Pre-trained Tranformers) and then iterated on improving those characteristics - It took them a while before they had a model capable of following instructions well. There was lots of interest in OpenAI when they released GPT-1 and much more when they released GPT-2. All the while this was going on Google slept - I think the reasons for this are varied and many, stretching into company culture, structure, processes etc. This was after researchers at Google created the Transformer architecture in the first place. Google seemed to panic suddenly when ChatGPT was released. I think this is when the people who could actually make change realised that ChatGPT could be a threat to Google in a number of ways - including for search market share.

benob 3y ago

Google had it in 2015 [1] [1] https://arxiv.org/abs/1506.05869 https://arxiv.org/abs/1506.05869