Hacker News Reader

Comment by nuz | Hacker News Reader

OpenAI are so good at fake leaking things to build hype

GPT-3 was released in 2020. GPT-3.5 in late 2022. GPT-4 in early 2023, and that's basically it for me. Granted, they introduced the 4o model later that has some nice features while being quite well behind their top model (now called by them "legacy"). In the meantime, Claude 3.5 became dangerously close while having a much better context window than what AI offers.

So since March 2023 we haven't seen any increase in quality on their part. They need to "leak" things to keep market interest in them.

4o is doing quite well in the benchmarks, especially after the latest updates.

In my experience (and in the benchmarks), gpt-4o is now way ahead of the older gpt-4 for most uses.

In my experience, GPT-4o is noticeably worse than GPT-4, to a degree where it is already frustrating. The only benefit is that it is faster.

And remember how a lot of people not even related to the field were predicting ChatGPT 16 by this time a year ago?

Never heard of that, the closest I did see was a meme of:

GPT-3.5: *small circle* 175 billion parameters

GPT-4: *huge circle* 175 trillion parameters

(I think the second number was constantly being increased with each day that passed, but perhaps I misremember)

you must have been living under a rock. The hype last year was unreasonable. Everyone thought AGI would be with us in a single digit year period

I guess I am reading HN too much :shrugs:

> So since March 2023 we haven't seen any increase in quality on their part

Huh? On typical benchmarks, the gap between current models and original GPT-4 is as high as original GPT-4 and GPT-3.5 at time of release

Reply on news.ycombinator.com