Diary 2024-05-14
overlast Well, this is cool. This is a great step up in the minimum for speech recognition, text-to-speech, and voice chat services. It won't be long before we will be able to do this in any language and voice quality we want, and even outside of OpenAI. At the same time, expectations for humans will also increase. In the future, humans will be required to be very flexible. ImAI_Eruel OpenAI, Google's Gemini, first surprised the world with a demo video, but "it ended up being a fake combination of various things". Google's Gemini first surprised the world with a demo video, but "it ended up being a fake combination of various things" and became a hot topic. https://pbs.twimg.com/media/GNeZ2mqaMAAd28p?format=jpg&name=medium#.png
ImAI_Eruel However, I think that Google will come to this rather soon (I think Google will win in the video-related market since YouTube, a major source of data, is under its umbrella). I believe that Google will win in the video-related market since it is under the umbrella of YouTube, the main source of data). LiamFedus GPT-4o is our new state-of-the-art frontier model. We’ve been testing a version on the LMSys arena as im-also-a-good-gpt2-chatbot . Here’s how it’s been doing. GPT-4o is our new state-of-the-art frontier model, and we are testing a version as im-also-a-good-gpt2-chatbot in the LMSys arena. Here's a look at how it's doing.
https://gyazo.com/27fa3e87f18fef390bd98ea9dda0ac25
hiro_gamo We all tried so hard to catch up with GPT-4, but it's too ruthless. And at half the price and twice the speed. As I mentioned before, OpenAI is one company whose evolution is discontinuous....
nishio OpenAI's launch of a desktop app is a strategic move to compete with Google's vast repository of 'videos taken by people to show to others' on YouTube. They aim to gather data from 'screens people watch while working' every day, gaining an edge in the data war. The launch of OpenAI's desktop app is a strategic move to compete with Google's vast repository of "videos for others to see" on YouTube. They aim to gain an edge in the data wars by collecting data from the "screens people see at work" every day.
Sonnet
Input: $3 / MTok
Output: $15 / MTok
Opus
Input: $15 / MTok
Output: $75 / MTok
gpt-4o
Input: $5 / 1M tokens
Output: $15 / 1M tokens
Gemini 1.5 Pro
Input: $7 / 1M tokens
Output: $21 / / 1M tokens
gpt-4-turbo
$10.00 / 1M tokens
$30.00 / 1M tokens
gpt-4
$30.00 / 1M tokens
$60.00 / 1M tokens
gpt-4-32k
$60.00 / 1M tokens
$120.00 / 1M tokens
Even though Claude's context is 200K context, which is longer than GPT's 128K context, it's still 3x cheaper than Opus, and when you get a 1.3x buff with the Japanese tokenizer...
GPT-4o
The message you submitted was too long, please reload the conversation and submit something shorter.
87K token
---
This page is auto-translated from /nishio/日記2024-05-14 using DeepL. If you looks something interesting but the auto-translated English is not good enough to understand it, feel free to let me know at @nishio_en. I'm very happy to spread my thought to non-Japanese readers.