Thread by @yleflour

🏁 Here we go with Part 3

While I was writing Part 2 I finally got my #gpt4 API access 🥳

Using the API #chatgpt, it answers at least 3x faster, cool

So we're back on track from the @openapi playground while literally siphoning my 💵

Some thoughts about that...

1/11

#gpt4 is ~20x more expensive than #gpt3

After 10 minutes I ended up maxing the context (what's taken into account for the response) and I was already at $2.18

May seem small but it really depends on your situation 🌎

2/11

When you figure out the Purchasing Power Parity between countries, or even within the same country, those $2 for 20 minutes of #gpt4 can represent a lot and is out of reach for some

Hopefully there are 4 things that'll likely happen 👇

3/11

1️⃣ It'll get cheaper and fast

#gpt4 is in preview mode, heavy demand and the infrastructure costs must be wild

#gpt3 already got 10x cheaper and it's faster

The same will likely apply, although some are less sure about scale

4/11

https://twitter.com/WilliamAEden/status/1630690009354502145

2️⃣ More companies will foot the bill

@Microsoft already integrated #gpt4 in Bing a few will likely follow

Only a few because you really need deep pockets for this right now 🏦

Sure you're likely still selling out your privacy, but at least it's free for everyone

5/11

3️⃣ Your company will foot the bill

When you can x3 your team's productivity you'll be paying that 💵 to get the edge on others

If you don't you'll be x3 slower than the competition

But in the long run this x3 will be the new norm, so what becomes of this extra cost? 🤔

6/11

I'm way too far from my domain of expertise to answer such question...

But I know someone who isn't 🤖

Seems right. But do I agree ?

Well, sure, but the implied impact is kind of bleak 😞

7/11

#chatgpt is already impacting entry-level jobs

Sure, it may boost wealth creation for companies and wealthy individuals, but don't expect it to trickle ⬇️

It's more than likely that it'll only exacerbate income disparity

BTW the labor impact report dropped and it's grim

8/11

https://twitter.com/rubinovitz/status/1637651591191842816

4️⃣ Open source models will play catch up

Stable diffusion and more are already available to anyone and can run on some home computers

But because of openAI's move this might get a push back from companies actually paying to train those models

vice.com/en/article/5d3…

9/11

Still, the clear winner here is obviously OpenAI

I don't know the exact power required to run it, but it's not even close to running locally anytime soon

There are 100 new tools built on top of it everyday. Be the gatekeeper in such ecosystem and you're bound to win big

10/11

Done with the rambling on cost and all

Took quite the detour 🔀 here so let's close up Part 3

If you liked this thread, you can like/retweet the first tweet

You can also hit the 🔔 to get notified of Part 4. Promise I won't spam ☝️

11/11

https://twitter.com/yleflour/status/1637786190169636865