3 ms·

How big is their moat? Actual open source projects seem to be closing the gap. I feel like openAI will plateau and there will be a red sea of competition in 202

by ddxv 3y ago

How big is their moat? Actual open source projects seem to be closing the gap. I feel like openAI will plateau and there will be a red sea of competition in 2024. Is the company really worth that 100b?

arthurcolle 3y ago
Yeah, they have negative moat. I was able to implement a better RAG system than what they have for ChatGPT within 3 weeks as a single engineer working on it at a small startup. Realistically yeah they have scale and capital and I loved the DevDay stuff and the new APIs, but I don't see how they keep an edge unless they specifically pursue contracts with data providers and maintain exclusivity. ¯\_(ツ)_/¯ References: 1. Mixstral 8x7b (now hosted on Anyscale w/ function calling within a few weeks of release) existing at all 2. This post from moments ago: https://old.reddit.com/r/LocalLLaMA/comments/18orv70/in_a_test_mixtral_q4_0_is_better_than_llama_2_70b/ https://old.reddit.com/r/LocalLLaMA/comments/18orv70/in_a_te... Consider how many people are advantaged by having an open source GPT-4 level model. Everyone wants this. Everyone has built OpenAI compliant APIs to help devs plug and play. I don't see this persisting. The singular key innovation IMO from DevDay was the Threads API with very good reranking Assistants are much worse qualitatively vs using chatcompletion models. They are good but very limited. The best thing is the Threads API
- rafark 3y ago
  There’s always (at least in this era) going to be a market for a could-based saas. Most companies are more than fine using a solid HTTP API and a lot of people (I’d say the majority) just need a nice ready to use web client. So I think they’re good for the next few years. Openai has the branding advantage over the competition too. Chatgpt is kind of like what Google is for search engines. So the competition not only needs a good enough product (I’d even say you might need to offer more if you want to compete with the big players) but they also need good marketing to attract users.
  - amluto 3y ago
    > There’s always (at least in this era) going to be a market for a could-based saas. It hasn’t been that long since companies would haven’t thought the cloud was weird and much preferred on-prem. The tides could turn back.
thorum 3y ago
The OpenAI moat is private datasets, a sizable chunk of the best talent, proprietary discoveries (like whatever Q* is) that keep them ahead of everyone else, and lobbying advantages that will help them to survive future AI regulations. It doesn’t matter that much if you can build your own AI system with open source tech as long as OpenAI’s solutions are better and cheaper to use.
- yen223 3y ago
  The OpenAI moat is that most people have heard of ChatGPT, and not any of the hundreds of AI products out there
  - xnx 3y ago
    True, and a year ago very few of those people had heard of ChatGPT, so it's very possible that something new could come along in the same way.
- OnlineGladiator 3y ago
  You could've said the same thing about Google (specifically their AI efforts) 5-10 years ago, and what ended up happening is the majority of their best engineers left for other places like...OpenAI.
  - profile53 3y ago
    And you probably could have said the same thing about Google 25 years ago when altavista and yahoo were around. What ended up happening was a complete dominance on web search for Google. The end result is unknown but it does appear they have an advantage right now.
    - OnlineGladiator 3y ago
      Google had a revolutionary search technology which they quickly learned how to monetize and it began printing money. OpenAI has competitive technology in a crowded field and they are currently burning through money, likely hundreds of millions every month. They are still a glorified tech demo in search of a profitable product. I agree I don't know what will happen. But apart from PR (and arguably funding) OpenAI really doesn't seem special. I say this as someone who interviewed there and with competitors. And many people in the field have developed a distaste for OpenAI's management over the years to the point it's becoming a problem with recruiting.

by ddxv 3y ago

arthurcolle 3y ago

Yeah, they have negative moat. I was able to implement a better RAG system than what they have for ChatGPT within 3 weeks as a single engineer working on it at a small startup. Realistically yeah they have scale and capital and I loved the DevDay stuff and the new APIs, but I don't see how they keep an edge unless they specifically pursue contracts with data providers and maintain exclusivity. ¯\_(ツ)_/¯ References: 1. Mixstral 8x7b (now hosted on Anyscale w/ function calling within a few weeks of release) existing at all 2. This post from moments ago: https://old.reddit.com/r/LocalLLaMA/comments/18orv70/in_a_test_mixtral_q4_0_is_better_than_llama_2_70b/ https://old.reddit.com/r/LocalLLaMA/comments/18orv70/in_a_te... Consider how many people are advantaged by having an open source GPT-4 level model. Everyone wants this. Everyone has built OpenAI compliant APIs to help devs plug and play. I don't see this persisting. The singular key innovation IMO from DevDay was the Threads API with very good reranking Assistants are much worse qualitatively vs using chatcompletion models. They are good but very limited. The best thing is the Threads API

rafark 3y ago
There’s always (at least in this era) going to be a market for a could-based saas. Most companies are more than fine using a solid HTTP API and a lot of people (I’d say the majority) just need a nice ready to use web client. So I think they’re good for the next few years. Openai has the branding advantage over the competition too. Chatgpt is kind of like what Google is for search engines. So the competition not only needs a good enough product (I’d even say you might need to offer more if you want to compete with the big players) but they also need good marketing to attract users.
- amluto 3y ago
  > There’s always (at least in this era) going to be a market for a could-based saas. It hasn’t been that long since companies would haven’t thought the cloud was weird and much preferred on-prem. The tides could turn back.

thorum 3y ago

The OpenAI moat is private datasets, a sizable chunk of the best talent, proprietary discoveries (like whatever Q* is) that keep them ahead of everyone else, and lobbying advantages that will help them to survive future AI regulations. It doesn’t matter that much if you can build your own AI system with open source tech as long as OpenAI’s solutions are better and cheaper to use.

yen223 3y ago
The OpenAI moat is that most people have heard of ChatGPT, and not any of the hundreds of AI products out there
- xnx 3y ago
  True, and a year ago very few of those people had heard of ChatGPT, so it's very possible that something new could come along in the same way.
OnlineGladiator 3y ago
You could've said the same thing about Google (specifically their AI efforts) 5-10 years ago, and what ended up happening is the majority of their best engineers left for other places like...OpenAI.
- profile53 3y ago
  And you probably could have said the same thing about Google 25 years ago when altavista and yahoo were around. What ended up happening was a complete dominance on web search for Google. The end result is unknown but it does appear they have an advantage right now.
  - OnlineGladiator 3y ago
    Google had a revolutionary search technology which they quickly learned how to monetize and it began printing money. OpenAI has competitive technology in a crowded field and they are currently burning through money, likely hundreds of millions every month. They are still a glorified tech demo in search of a profitable product. I agree I don't know what will happen. But apart from PR (and arguably funding) OpenAI really doesn't seem special. I say this as someone who interviewed there and with competitors. And many people in the field have developed a distaste for OpenAI's management over the years to the point it's becoming a problem with recruiting.