Replying to @ronawang
I knew him before his pod cast and met him in one of the data science conference. Really enjoyed his YouTube videos and he is super humble and nice in person. A lot of people dunk on him for no real reason.
Karthik Ramasamy
3,480 posts
Karthik Ramasamy
@_cartick
Building agent platform for knowledge work, checkout myaidrive.com . Previously ML at Google, Uber and LinkedIn.
San Francisco, CA
Joined September 2009
- Replying to @OfficialLoganK and @OpenAII really like you folks but your message is not consistent with 70 percentage of company that is threatening to quit. As someone who wholly depends on your API for our product/company the recent set of events is hard to digest.
- Replying to @NickADobosI also thought it was strange but then it seemed to make sense. This page design is very important from a security perspective, people are used to it. When you change it without warning many will think it is a phishing page.
- Google's new generative AI search is amazing! RIP StackOverflow :( I find it easier to use than ChatGPT as it is readily available in Chrome search tab and I can always go to search results if I want.
- Replying to @amix3kDid you try Gemini 1.5 pro flash or the OSS vision model, both were available immediately? I tried flash model and it looks pretty good so far. No multi modal model comes close when it comes to context size.
- Bard is improving really well in code generation. It is amazingly fast compared to GPT4. In one instance, I preferred its output to GPT4. I am going to use it for code generation from now on.
- Replying to @keerthanpgSame reason we have different tiers of storage from L1 cache to blob storage. We deal with data that is much more than 10M tokens. Also 10M tokens doesn’t come free, there is a lot of cost and latency associated with it.
- Replying to @DrJimFanI don't think they are processing video but the model takes a screenshot of the video when it wants. You can see this from how it behaves.
- Replying to @bindureddy@bindureddy Great points. I am the creator of Ai PDF GPT along with @vicentes . Currently our GPT is #2 in popularity. We provide practically unlimited file uploads in free tier and in pro tier offer folder search for searching across 1000s of files, automatic OCR (quality is
- Replying to @Grady_BoochStrong disagree, our users come all over the world, from different domains, from different age groups. Most of them are very productive in their work after using LLMs. Don't come to conclusions without talking to actual users of GenAI products.
- Replying to @ClementDelangue and @hardmaruWe don’t want one entity (company or government) tell which models can be downloaded.
- Replying to @sundarpichaiGemini 1.5 Pro Flash is a game changer for document use cases with 1M context and soon 2M context. The cost makes it very useable. It unlocks lots of use cases that is only possible with long context. We just added the support for it in our product and waiting to see how it is
