Kevin Weil的动态

Testing on production already! Congrats!

The conversation then proceeded to inquiries about what sort of coworker you were. Don’t worry, I put in a good word for ya.

Honest question: if you as the CPO know that each product performs better for each use cases, why not have both under the same interface and decide for us which model is more adequate? Take away that cognitive load from us users - I don’t want to have to know the difference between all these models when we get to GPT 33

GPT-4.5 is making AI more conversational, emotionally aware, and adaptive a major leap for industries relying on communication (coaching, sales, HR). ?? If you’re using AI for personal branding, leadership insights, or customer engagement, now’s the time to test tone calibration in ChatGPT. ??Ask “How would a [tone: empathetic, authoritative, casual] expert in [industry] respond?” to fine-tune its style.

What unique benefits do you see going between o1 and o3? Why not all o3-mini?

How about... you use ChatGPT4o to analyze the prompt and just route it to the appropriate model. Dont get rid of the picker, but dont rely on us to track your slew of features and the latest synthetic benchmark report to know which latest version to pick? Also, this might save you some compute cost...

Can't wait for new creative usage for GPT-4.5! We're collecting some creative ways to use LLM's in our director https://ideakit.ai ??

Very interesting! Two things I hope to see more out of GPT now and in the future: 1. A way to transform soft skills into the most efficient transfer of knowledge that is possible. For example, I have been working on Feynman-like prompts lately, and have been getting ChatGPT to summarize data using mermaid.md prompts. But if I flat out ask it to create a chart of some sort summarizing the data, visually it is still very lacking. 2. A way to integrate better with quick video clips and external browser trees. It has remained a challenge generating designs and CSS styling with AI, especially since art is subjective. However, I attribute this to my prompts lacking. Still I have seen AI quickly forget that we are coding in Tailwind CSS and not vanilla CSS, or needlessly change CSS and HTML without being prompted, which is the other-edge of the sword when it comes to ChatGPT and related AI taking creative liberties based on interpretation (and probably to some extend, randomness). All in all, I hope 4.5 helps me interface with ChatGPT better. The more I can help it help me, the better. I think part of that involves its overall capabilities, so... between 4.5 and Sonnet 3.7, looking forward to checking out the new updates!

Can’t wait… Come share what you build and learn with us in the AI Agents group on linkedin: https://www.dhirubhai.net/groups/6672014

?? #GPT4.5 vs #DeepSeek R2: Who will rewrite the rules in the battle for AI hegemony? #AIRevolution #TechTrends #FutureOfWork ?? International Giants vs Chinese Dark Horses: ?? Data speak: GPT-4.5: inference speed ↑30%, but training cost ↑50% R2: multilingual reasoning accuracy ↑ 83%, image generation speed ↑ 40% ?? Industry earthquake: Claude, Qwen and other forced to accelerate the iteration, the global AI track “involution” upgrade! Enterprises are selected ** “international brands” or “domestic cost-effective” **? ?? soul torture: If R2 performance is true, China's AI can put an end to the “necklace” era? Is GPT-4.5's “last-generation non-chained model” a bold statement or a crisis? ?? See you in the comments section! Who are you betting on? #Leadership #Innovation #ArtificialIntelligence #EthicalAI There are no bystanders in this showdown! ??