Uhhhh, so Sam Altman says here in this clip that OpenAI already had GPT4 internally in Nov22, when chatgpt (3.5) got released.
So that means we're now in a solid plateau of LLM capabilities for the last 18 months now. The idea that LLMs will get significantly better is just plain wrong.
OpenAI already tried twice to go to a GPT5, but failed both times with 4turbo and 4o. Just announced they'll try again lol
https://www.nytimes.com/2024/05/28/technology/openai-gpt4-new-model.html?smid=tw-share
https://x.com/bilawalsidhu/status/1796968949201969430
Laurens Hof /
npub1cf…4lnhq
2024-06-02 08:46:01
Author Public Key
npub1cf27656assh3ykhn0ruxspw8xvpx4c9xk0f8sasfsqk2qf8tauss64lnhqPublished at
2024-06-02 08:46:01Event JSON
{
"id": "2250093d6557b1bf1438a535108ae74904fbc69c557323eedfe9480956c83d08",
"pubkey": "c255ed535d842f125af378f86805c733026ae0a6b3d2787609802ca024ebef21",
"created_at": 1717317961,
"kind": 1,
"tags": [
[
"proxy",
"https://indieweb.social/users/laurenshof/statuses/112546149906963128",
"activitypub"
]
],
"content": "Uhhhh, so Sam Altman says here in this clip that OpenAI already had GPT4 internally in Nov22, when chatgpt (3.5) got released.\n\nSo that means we're now in a solid plateau of LLM capabilities for the last 18 months now. The idea that LLMs will get significantly better is just plain wrong.\n\nOpenAI already tried twice to go to a GPT5, but failed both times with 4turbo and 4o. Just announced they'll try again lol\n\nhttps://www.nytimes.com/2024/05/28/technology/openai-gpt4-new-model.html?smid=tw-share\nhttps://x.com/bilawalsidhu/status/1796968949201969430",
"sig": "3e9f11995fb6bdc04b0504d4a757444aec00b765a41ded4f2250970e7c0b5e3e5d3852044a8aa6e3002eb42dbc989efa409047d65497e59fe56da12d8806567f"
}