Тёмный

ChatGPT Demo of Strawberry (model o1-preview) -- Hallucination Killer? 

Cassie Kozyrkov
Подписаться 73 тыс.
Просмотров 4,7 тыс.
50% 1

It's finally here! Join ex-Googler Cassie Kozyrkov for a walk-through of Strawberry, OpenAI's latest model o1-preview.
Can Strawberry "reason" and is it more hallucination-proof than other models? What even is an AI hallucination?
Learn what most people get wrong about AI hallucinations, what Strawberry is built for (and why it's expensive), what operationalization is, and more!
Plus a comparison of the difference between model o1-preview and model 4o in ChatGPT when we ask them to do the exact same task.
(Yes, the results are different!)
**********************************
Blog: / kozyrkov
Website: kozyr.com
Invite Cassie to speak at your event: kozyr.com/speaking

Опубликовано:

 

18 сен 2024

Поделиться:

Ссылка:

Скачать:

Готовим ссылку...

Добавить в:

Мой плейлист
Посмотреть позже
Комментарии : 13   
@nextrealm_ai
@nextrealm_ai 4 дня назад
It's definitely interesting, but baby steps into this new realm is recommended.
@SasskiaLudin
@SasskiaLudin 2 дня назад
I love your subtle but shining mind,,, Thank you for that refreshing review.
@soundformovement
@soundformovement 3 дня назад
Thank you so much for sharing your brilliance. I have enjoyed watching and learning from your amazing teaching. As a musician, it’s has been so helpful to have your perspective on LLM’s. The way you present ideas is super creative and fun. Incredibly grateful. 🙏✨
@zynga726
@zynga726 4 дня назад
I came to hear about strawberry and ended up also learning about hallucinations. Great review of Strawberry. Nice things is that we can use 4o for creative stuff and Strawberry for more factual stuff
@purplesheep99
@purplesheep99 3 дня назад
I wouldn't normally watch one of these demo style videos as it's with Cassie it has to be a mush watch.
@MustardGamings
@MustardGamings 3 дня назад
O1 isn't good for writing like you showed in the video it's good for advanced logic and reasoning. And most of writing is comparable to just 4o if you ready the paper open ai did about o1.
@LordOfRuin
@LordOfRuin 3 дня назад
"hallucination is a feature not a bug" that's the term I'm going to use whenever someone sees my code. Seriously though thanks for that, that's clarified for me hallucination, where its dependent on the context of what I'm requesting.
@sharif1306
@sharif1306 2 дня назад
I love this.
@RyanTeo
@RyanTeo 4 дня назад
I guess it is necessary for the marketing and finance folks to over-anthropologize AI so that the company valuation can skyrocket 😅
@PrasadJogalekar
@PrasadJogalekar 4 дня назад
Hallucination = an inferencing system where an error bound cannot be proven with respect to an evaluator (human or otherwise)
@iakobkv271
@iakobkv271 2 дня назад
EXACTLY!!!!! You nailed it!!!!
@kaykwanu
@kaykwanu 3 дня назад
Always a pleasure learning from you ma.
Далее
ХИТРАЯ БАБУЛЯ #shorts
00:20
Просмотров 1,2 млн
OpenAI Releases Smartest AI Ever & How To Use It
21:16
Просмотров 107 тыс.
AI can't cross this line and we don't know why.
24:07
Просмотров 586 тыс.
NEW ChatGPT o1 is INSANE: See the Full Demo! 🍓🚀
16:58
How AI 'Understands' Images (CLIP) - Computerphile
18:05
How To Make A VIRAL AI Music Video (For Free)
28:01
Просмотров 267 тыс.
Is o1-preview reasoning?
1:24:51
Просмотров 36 тыс.