I've watched several of your videos and I really enjoy them. I appreciate how you take complex topics and make them so much easier to understand. Please continue to make more!
Great. Just Great! I am so surprised to see you have "only" 1.4k views. It takes quite a lot to get me to click on a video that is almost half and hour long, but this one was well worth seeing through. You are so good at making this clear! And you are so balanced and transparent. I am actually very interested in what open source type of agent builders might be out there?
@@andrewHutchison-v8n thanks a lot 🙏 i will keep making more! If youre technical you can check out Crew AI which is an agent framework, but if youre non technical then there’s not too many right now!
I have seen a lot of agentic ai videos. This one was good, very good as far as a technology explainer goes. Ranks among the best i have seen so far. Nuch appreciated👍
This was very clear and very well done. Much appreciated. I’m looking for a no code solution involving either Agentic AI or these “middle-steps” to create a SaaS prototype. As a B2C prototype, it straddles a consumer and enterprise approach as you’ve described. Where do I go next to explore what I need?
Interesting, So If I create a Custom Agent on any of the LLM's using the already available infra will these agents be saved for long term usage? Or would I have to start from scratch every time I start?
Great video!!! One question: Does the System Prompt affect token use? If it's being inputted with each prompt, does that mean shorter Custom Instructions take up fewer tokens per prompt than longer Custom Instructions?
Hi There! I kind of liked your other video 'AI Agents Explained' more. However, I must say that I am getting too confused about how to make inroads to a career in AI. I don't even know where to start. There are NVIDIA NIM, Omniverse, AI Agents, RAG, list goes on . Can you give a pathway. How could I be employable in AI?
There's nothing stopping you from making them fully agentic now. They totally can. the problem is that the results at the moment are often subpar. Mainly because the current models capability to reason (i.e. know what steps to take) is inferior. Often because models use general knowledge to come up with reasoning which may not be accurate. For example, if the user asked an agent to book me a movie ticket, it could totally do that automatically by looking at your calendar, seeing when you're free, seeing what films are available at that time, and booking a movie. But the level of reasoning needed to know the right actions might be basic. The agent might say "To plan this movie I need to know when Alfie's free, and what films are available at that time", but the model may not know to reason with things like "Is it for a date with his wife, or a movie with his kids" or "Should I prioritize the cinema with the films he likes, or should i prioritize the cinema with the best popcorn?" The reasoning part is where they fall down today.
You use too many words in your video. It’s hard to understand you. Try to talk less about the details and a bit more about the bigger picture. Anyway, thank you for sharing.
titling the video building without building anything uhn you did not use an LLM to give you a better title ? ah i forgot we're on youtube & most of youtubrrs are just after views , while pretending to provide value , value start with being honest & zero tactic of click baity titles that are searched & the algo reward , you wasted my time hope i will see you next time in my feed with something really valuable
I've watched several of your videos and I really enjoy them. I appreciate how you take complex topics and make them so much easier to understand. Please continue to make more!
@@hanami27 thanks so much for the encouraging words!
Great. Just Great! I am so surprised to see you have "only" 1.4k views. It takes quite a lot to get me to click on a video that is almost half and hour long, but this one was well worth seeing through. You are so good at making this clear! And you are so balanced and transparent. I am actually very interested in what open source type of agent builders might be out there?
@@andrewHutchison-v8n thanks a lot 🙏 i will keep making more!
If youre technical you can check out Crew AI which is an agent framework, but if youre non technical then there’s not too many right now!
@@AlfieMarsh CREW AI....thanks for the tip.
I have seen a lot of agentic ai videos. This one was good, very good as far as a technology explainer goes. Ranks among the best i have seen so far. Nuch appreciated👍
Excellent video. Thanks so much for posting.
Alfie you are great! Fantastic analysis. Liked and subscribed to your channel. Thanks.
Thanks for the support 🙏
Great video mn. 👍 Well explained and easy.
Keep up these great videos Alfie!
Thank you! Let me know what topics you want to hear more about!
This was very clear and very well done. Much appreciated.
I’m looking for a no code solution involving either Agentic AI or these “middle-steps” to create a SaaS prototype. As a B2C prototype, it straddles a consumer and enterprise approach as you’ve described. Where do I go next to explore what I need?
Did you ever find a good no code solution for this?
@@bryanalstoncoxing I’m leaning towards VectorShift. However, every two weeks another 100 apps/services pop up; bound to find something. :-)
Thank you for the overview!
Nicely done here bro.Thanks
A must watch again
ASI Prompt: "Achieve World Peace Without Harming Humans."
AI definition of human fit only to a master race....
Interesting, So If I create a Custom Agent on any of the LLM's using the already available infra will these agents be saved for long term usage? Or would I have to start from scratch every time I start?
I am interested too. Can agents be auto-update with the last versions of ChatGPT/etc and where the memory is stored?
Plugging in the LLMs to these agents isnt hard so you could update the model easily. But they may perform slightly differently
Great video!!! One question: Does the System Prompt affect token use? If it's being inputted with each prompt, does that mean shorter Custom Instructions take up fewer tokens per prompt than longer Custom Instructions?
Hi There! I kind of liked your other video 'AI Agents Explained' more. However, I must say that I am getting too confused about how to make inroads to a career in AI. I don't even know where to start. There are NVIDIA NIM, Omniverse, AI Agents, RAG, list goes on . Can you give a pathway. How could I be employable in AI?
Great content. If I'm creating a chatbot for a bank, can I use an agent for actions like reporting an issue or checking balance?
@@wesleymogaka you could. Any “function” or “action” can be called via the app. It can be anything related to any other software. So yes, absolutely!
"Hey there! As someone who's fascinated by technology, I totally get your excitement about its uses nowadays. #SmythOS #Technology
Co-Pilot is cool. We all know what that is. Why does Google have to create a new acronym that we have to learn and remember.
Why don't they just call GEMS, templates. We are all already used to this name and know what it means?
Why can't we make these Agents fully Agentic? It seems like any obstacles to this are artificial.
There's nothing stopping you from making them fully agentic now. They totally can. the problem is that the results at the moment are often subpar.
Mainly because the current models capability to reason (i.e. know what steps to take) is inferior.
Often because models use general knowledge to come up with reasoning which may not be accurate.
For example, if the user asked an agent to book me a movie ticket, it could totally do that automatically by looking at your calendar, seeing when you're free, seeing what films are available at that time, and booking a movie.
But the level of reasoning needed to know the right actions might be basic. The agent might say
"To plan this movie I need to know when Alfie's free, and what films are available at that time", but the model may not know to reason with things like "Is it for a date with his wife, or a movie with his kids" or "Should I prioritize the cinema with the films he likes, or should i prioritize the cinema with the best popcorn?"
The reasoning part is where they fall down today.
You use too many words in your video. It’s hard to understand you. Try to talk less about the details and a bit more about the bigger picture. Anyway, thank you for sharing.
titling the video building without building anything uhn you did not use an LLM to give you a better title ? ah i forgot we're on youtube & most of youtubrrs are just after views , while pretending to provide value , value start with being honest & zero tactic of click baity titles that are searched & the algo reward , you wasted my time hope i will see you next time in my feed with something really valuable