“We automated 150 tasks with AI Agents, just copy us” - Microsoft AI
ฝัง
- เผยแพร่เมื่อ 10 ต.ค. 2024
- Want to build an AI Agent Startup? Go here: www.skool.com/...
Research Paper: arxiv.org/abs/...
Follow Francesco: x.com/francedot
Follow Dan:
Windows Agent Arena: microsoft.gith...
GitHub: github.com/mic...
Follow me on Instagram - / thedavit
Follow me on Twitter - x.com/DavidOnd...
Subscribe if you're serious about AI.
David Ondrej Podcast - Episode 22
Want to build an AI Agent Startup? Go here: www.skool.com/new-society
Quote of the Day: "You cannot improve what you cannot measure."
It's a quote given in many management books.
If I could screen-record what I do for work and use that as AI training data... sooner or later I'd have an agent that could do my job, right?!!
Yes, it’s like writing a macro script for your ai agent.
Then playing it back, but it learns and may do it differently and better over time
@@x1k790 i could sure use this. Sit back and watch it do my work for me!!!
NotebookLM is a good example, with the podcast functionality
So much work on this, but it’s not usable since it does not work with the web, a better thing to do would have been to take autogen and then make an offshoot that works locally on windows, that way one framework would have best of both worlds, there wasn’t a need to build a separate local only, even if benchmarks were the reason.
This ISNT autogen, just to be clear. They said kind of the same as you, autogen isnt made for this so these geniuses whipped up a proprietary agent that scans the screen.
@@MrWizardGG I understand, but my point was why make something new, instead just add these screen scanning abilities to autogen
@@deepak12428 i so wish it was added to autogen as well
“We automated 150 tasks with AI Agents".... yeah..... anyone can do that when the tasks are so incredibly simple.
Can u explain the title??
Were the agents trained in those tasks?? How did it play out?
@@arslaanmania1309 They are changing the profile icon using AI. These were AI Automation not Exactly AI Agentic
actually, the opposite is true. its very difficult for an LLM to complete a "simple" task without it being pre-programmed in its training data
Real answer: its impressive because they built an effective proprietary tool to scan and act on their whole screen with somewhat good results. This is not yet ubiquitious in public tools yet
@@MrWizardGG except the usefulness of it at the moment is nearly zero. Yes I suppose you have to start somewhere so that it can advance further along.
wtf, they didn't need any special proprietary crap for this which makes bounding boxes and IDs, VRSEN made a video ages ago where he made some javascript that does that with a vision model and it worked.
well windows 12 will already have this ability so it does not matter. Its just a matter of time (1 year top ) to have this on your computer and do anything by voice etc
Clever Macro agents hehe
Could you make a video tutorial on how to turn a sketch into a consistent character? I need advice on which AI tools to use, and which offer the best value. Sometimes my prompts based on my descriptions don't match my sketch, and the AI tools that can handle sketches don't always create consistent characters.
-----
Q6
I don't have any experience with this, I'm just throwing an idea: Have you asked the AI to create a character concept from your sketch that covers many view angles of it in one image. Then pass that image instead of your sketch?
@@ahmadmanga That's exactly what I was thinking, but it might not be a long-term solution. I'll try your suggestion first, though.
Hi David, love your content! I have an Alienware m18 R2 with an Intel i9-14900HX, NVIDIA RTX 4090 (24GB), 64GB RAM, and 8TB storage, but I struggle to run LLaMA 70B models. Could you create a video for users like me on optimizing setups (8-bit quantization, mixed precision, etc.) to run large models efficiently? Your help would be greatly appreciated. Many Thanks!
Get the 3B model
We need much more complex voice conversations with Ai not AI imagery. I want to have complex debates with AI on my Alexa, at the moment it's very poor.
For convo i found chatgpt plus/paid new voice chat very natural.
your headphones aren't straight.
"Proprietary" whatever -- hard pass. No AI will ever touch my personal computer unless it's 100% open source. People are insane giving access to their personal systems and information to blackboxes controlled by third parties.
Poor chubs stuck in their bedrooms 😂. What is wrong with yall go to the damn office.
you have very good videos but this is not ok, it does not offer any value to people.
first
Don't insult our intelligence man. AI agents un on top of LLMs, you don't know that ?
I've got to say...i turned this podcast off after a few minutes because only one of the speakers could pronouce the "th" sound which is foundational to speaking English. One guy in the podcast speaking like a child is bearable, but any more and it just destroys the message. i.e. why should I listen to people who are not careful about what they say?
Because they speak, at the very least, two languages while you're still struggling to understand plain English.
This is such a bad idea towards humanity! Watch how this all unfolds! Don't trust them!
first