Introducing ChatGPT and Whisper APIs • Hacker News Board

Re: Introducing ChatGPT and Whisper APIs

#671

Post by sebzim4500 » Thu, Mar 02, 2023, 9:37 PM UTC

motoxpro wrote: post #664

sebzim4500 wrote: post #433

Earlier quoted context omitted.

Hard to see how businesses would become dependent on it, if it takes off then someone will run similarly capable models at a reasonable profit margin.

Didn't happen with Google maps. Mapbox is definitely not 1400% cheaper. And many many people use Google maps before and after the price change. So I would disagree on both points.

Google maps has a bigger network effect and has a way bigger barrier to entry. You can train a new LLM for a few million. Good luck collecting map data on the entire world for that much.

Re: Introducing ChatGPT and Whisper APIs

#672

Post by kordlessagain » Thu, Mar 02, 2023, 9:53 PM UTC

eitland wrote: post #561

kordlessagain wrote: post #387

Earlier quoted context omitted.

I use it with documents I shred into sentences and embed with ada-002. This makes it spot on when talking about a given document. https://mitta.us/

This is seriously cool. Bots for the people!

Thanks! :) Been working on it a LONG time. Getting ready to launch...

Re: Introducing ChatGPT and Whisper APIs

#673

Post by lee101 » Thu, Mar 02, 2023, 10:00 PM UTC

TechBro8615 wrote: post #26

polygamous_bat wrote: post #4

Earlier quoted context omitted.

> I have no idea how OpenAI can make money on this. This has to be a loss-leader to lock out competitors before they even get off the ground. The worst thing that can happen to OpenAI+ChatGPT right now is what happened to DallE 2, a competitor comes up with an alternative (even worse if it's free/open like Stable Diffusion) and completely undercuts them. Especially with Meta's new Llama models outperforming GPT-3, it…

Someone can still undercut them by offering an uncensored version.

[dead]

Re: Introducing ChatGPT and Whisper APIs

#674

Post by purplecats » Thu, Mar 02, 2023, 10:03 PM UTC

leobg wrote: post #604

cjbprime wrote: post #506

Has anyone made a ChatGPT-like app which both accepts and responds with voice input yet?

I have, but haven’t bothered to publish it. POC to use by me and my daughter. Works like Siri. Fun to play with. But the hallucinated responses made us stop using it pretty quickly. Needs more work to be of any use (integrate search engines, etc.).

> Works like Siri

you can also make it work _with_ siri. i get around this by proxying it through a sms service which is integrated with my bot via webhook. then use elevenlabs for TTS. sample siri query "hey siri tell leobgAi to check my finances"

Re: Introducing ChatGPT and Whisper APIs

#675

Post by Liutprand » Thu, Mar 02, 2023, 10:48 PM UTC

The pricing is insanely cheap. Why anyone would pay $20/month for ChatGPT Plus when we are going to have dozens of chat applications based on the APIs?

It took me just a little more then one hour to create a basic cli chat application (https://github.com/marcolardera/chatgpt-cli). In the next days/weeks/months I think we will see an explosion of ChatGPT based applications...

Re: Introducing ChatGPT and Whisper APIs

#676

Post by kkielhofner » Thu, Mar 02, 2023, 11:58 PM UTC

vishal0123 wrote: post #472

kkielhofner wrote: post #430

Earlier quoted context omitted.

Hopefully they’re doing plenty of batching - you don’t even need to roll your own as you’re describing. Inference servers like Triton will dynamically batch requests with SLA params for max response time (for example). That said I don’t think anyone anyone outside of OpenAI knows what’s going on operationally. Same goes for VRAM usage, potential batch sizes, etc. This is all wild speculation. Same goes for whatever t…

> will set you back $100k/yr This is 6 month of salary of one average developer's salary there. And BTW they are likely doing inference on 100s or 1000s of GPUs, not just 8.

Yes and a devops engineer to manage an even moderately complex cloud deployment is an average of an extra $150k/yr. I don't know where this "cloud labor skill, knowledge, experience, and time is free" thinking comes from.

8, 80k, or 800k GPUs depending on requirements and load - the point remains the same.

Re: Introducing ChatGPT and Whisper APIs

#677

Post by ChickenNugger » Fri, Mar 03, 2023, 12:59 AM UTC

gridlocdev wrote: post #499

ChickenNugger wrote: post #429

FYI: you can't sign up without an email address (which you have to verify) and then a phone number (which can't be a google voice / voip number). So count me out.

Those restrictions seem intended to prevent people from creating a bunch of bot accounts that exploit the free tier. (Which is a common problem among free tiers for CI services and free online games) I’m curious, what are some better alternative authentication methods to combat that problem than requiring a phone number?

Anything that isn't a massive privacy violation/security risk.

My phone number is none of their business.

Re: Introducing ChatGPT and Whisper APIs

#678

Post by graderjs » Fri, Mar 03, 2023, 1:51 AM UTC

iamflimflam1 wrote: post #643

graderjs wrote: post #524

Earlier quoted context omitted.

Why do you need diarization? That's attributing speech to different speakers, right? What sort of use cases?

Transcribing interviews, meetings etc…

Did some research and seems there’s no reliable diarization method right now. They all have error rates around like 20%.

Re: Introducing ChatGPT and Whisper APIs

#679

Post by nr2x » Fri, Mar 03, 2023, 2:27 AM UTC

Silverback_VII wrote: post #579

whatshisface wrote: post #284

Earlier quoted context omitted.

Their new AI safety strategy is to slow the development of the technology by dumping, to lower the price too much to fund bootstrapped competitors.

I highly doubt it. OpenAI, Google and Meta are not the only ones who can implement these systems. The race for AGI is one for power and power is survival.

LLM can do amazing things, but it’s a basically just an autocomplete system. It has the same potential to take over the world as your phones keyboard. It’s just a tool.

Re: Introducing ChatGPT and Whisper APIs

#680

Post by nr2x » Fri, Mar 03, 2023, 2:29 AM UTC

andrepd wrote: post #446

freeqaz wrote: post #236

Earlier quoted context omitted.

It's also worth mentioning that, because Microsoft is an investor, they're likely getting these at cost or subsidized. OpenAI doesn't have to make money right away. They can lose a small bit of money per API request in exchange for market share (preventing others from disrupting them). As the cost of GPUs goes down, or they develop at ASIC or more efficient model, they can keep their pricing the same and then make mo…

OpenAI doesn't have to make money right away. They can lose a small bit of money per API request in exchange for market share (preventing others from disrupting them). Maybe I'm just old but back in my day this would be called "dumping" or "anti-competitive" or "market distortion" or "unfair competition". Now it's just the standard way of doing things.

You kidding? What do you think a business loan is? Almost every business needs some form of subsidy to get off the ground.