In this blog, I hope to do two things:
Share Insights on the benefits of AI-Enhanced Development Workflows
Share a simple Step-by-Step on how to ...
Some comments have been hidden by the post's author - find out more
For further actions, you may consider blocking this person and/or reporting abuse
Interesting article. Thanks for sharing your experience, and I definitely agree with asking multiple different LLMs the same question. It's kinda like getting opinions from more than one doctor 😊.
A friendly suggestion for you: Your article had quite a few grammatical/spelling errors which you could easily fix by using.... AI! 😁
Seriously, though, I had to reread a few parts to make sure I understood what you were trying to say. You could do something as simple as pasting the text into a Google doc to immediately see the errors. I actually prefer this to asking an LLM to fix errors because I prefer to think about how I want to fix the errors (and sometimes they aren't really errors), but LLMs are obviously a viable option.
I can see the thought you put into your article – conceptually it's great! But if you want to grow your readership, you'll want to make your writing as clear as possible so that people don't have to work to understand what you're saying.
Nice work, and best of luck to you!
Thanks @squidbe ! I really appreciate your kind words and your comments and feedback really resonates with me! I do use Google Docs for early drafts but I DO need to use it more often on my final drafts. This is a great point and something I will start to make sure I do in future posts.
But because you brought up the topic of AI in writing, this might be a great avenue to share something pretty personal to me. 😊
I actually purposefully do not use AI tools when it comes to my writing here or anywhere else I share my writing.
Here are my reasons:
TLDR; I want to grow in my writing, speaking, and editing skills to become a stronger natural communicator and to keep my writing voice. I avoid AI because it does not help me to reach my personal goals.
I want to continue to develop my writing voice and grow without being biased by any specific LLM. I have found in the past that LLMs will alter my voice and approach. I have often fought with AI chat agents to work with me and my goals. Haha!
I want to grow in my editing and proofreading skills. As I do a LOT of writing and sometimes find myself in scenarios where AI tools are not available, I need to harden my writing skills.
Building on point 2, I do public speaking quite often and my speech often reflects my writing. I have found as I write, proofread, and edit more often, my public speaking communication has improved.
(Building on Point 1) I'm from the Southern Unites States and I tend to use AAV (African American Vernacular) when I speak. Though I understand that most people don't understand that in written form, it is important to me to keep some of "who I am" in my writing. AI often deletes my cultural identity when it edits my writing.
Finally, I'm okay with the mistakes. In a way, I'm happy the mistakes exist so I'm never mistaken me for an AI when I write. I don't desire to be perfect but I do desire to be better. I'll also add that I spend a lot of time reading blogs on this and other sites and I can almost always identify when it's been written by or highly edited by AI. I personally do not enjoy overly AI produced writing pieces.
P.S. I will be re-reading through this blog again as I appreciate your feedback as it likely needs some more TLC. haha
Once again, I appreciate your thoughts. It will make me fight to be even more thorough with my re-reads and edits. I clearly need to tighten up on that skill.
Side Note: I do have a crazy goal to train an LLM that writes like me to give me a blueprint on how I can write my blogs in the future but, for now, I more or less deliberately avoid running my writing through AI tools.
That is excellent! I wish more people felt the way you do. I get why some people just want to pop their text into some chatbot and accept what it gives them, but if too many people do that, writing will become homogenous and bland. So, kudos to you for acknowledging the importance of having your own writing voice. 😎
That's why I just use Google Docs. It gives simple cues on possible errors, and that's all I want.
You mentioned fixing grammar using AI, but to be fair you don't need the "AI" to do it. There are great tools for this, Grammarly is a famous one but LanguageTool is a great alternative (which I use) which is open source (correct me if I am wrong about this).
I haven't looked into how these tools work exactly, but definitely not the LLM type of AI but old school text parsing pro max. And they work fantastically.
Not every automation needs AI, which is a great thing.
You know, I really need to look into Grammarly. I have been avoiding it (for the same reasons I mentioned in my other comment) but I should give it a go.
Definitely. That's why I mentioned Google Docs. Spell check and grammar check have existed for decades – I remember using them when I used to use MS Word a looong time ago.
I can see why some people want to use AI to check spelling/grammar, but I prefer to be the one making the decisions on what I do/don't change, which is why I like the simple cues of red and blue underlines to point out potential problems so that I can quickly decide whether to make a change.
Great writeup, very insightful - this post should have been in "Top 7 Featured DEV Posts of the Week" (dev.to/devteam/top-7-featured-dev-...) !
Just a question (funny how I realize I could ask this question to ChatGPT or another LLM and probably receive a capable answer, haha):
What are the pros and cons of using a local LLM running on your own machine, versus one hosted "in the cloud" and accessed online?
One advantage I can think of is that, even when I'd have no working internet connection for whatever reason, I'd still have a capable "local search engine" at my disposal?
Another potential advantage, for the environmentally conscious among us: we're not contributing to the already ballooning (largely due to AI) energy consumption of data centers (but maybe we're just shifting the energy consumption to our own machine ...)
Drawbacks that come to mind are the resources this uses on my local machine - disk space, memory, CPU, energy/electricity consumption ...
Thoughts?
First, thanks you so much for the kind words and reading my blog. I TRULY appreciate it! And I'm not sure if you saw, but i did end up being spotlighted in the "Top 7 Featured" blog so that was a true honor! :-)
As far as pros and cons, I think I echo the pros and cons you mentioned. I think energy consumption and privacy are huge pros for local LLMs. I tend to spin down my LLMs when I'm not using them but cloud LLMs have to be essentially accessible at all times so there's a continuous need for them to be running (thus the HUGE energy consumption).
so what I'll do is share some of my additional thoughts on some of the pros and cons that stand out for me.
Pros (Local LLMs)
Privacy: The ability to work with a Chat agent and to know that the information will never be stored in the cloud is great. Technically, I know most large cloud LLMs do not share out info (at least not intentionally) but there is peace of mind knowing that my conversations, especially ones with intimate details, remain on my machine.
Curation : I think one nice element of tools like Ollama is the ability to curate the LLMs that best suite your needs for a specific task. You could use Llama3.2 for general queries, you can use llam3.2-vision for vision based tasks, and use codegemma for code generation. I think the ability to specialize is powerful and incredibly helpful for accomplishing you personal and dev goals.
Personal Learning : The ability to experiment and learn about how LLMs work is one of the biggest pros to me. Though I am not training these models from scratch, I am learning a ton about their capabilities and how models differ and about proper prompting. Cloud models tend to be more of a black-box until they are released to the public (if they are open-sourced). Just being able to build upon models and having the ability to investigate more deeply into them has taught me so much.
Cons (Local LLMs)
Less 'Accurate': Open-Source models can sometimes be older models trained on older data and sometimes are still working on previous generational bugs. I have run into issues with Llama3.2 where it would give me "hallucinations" to certain questions I ask but if I ask a more modern Cloud model, like Gemini, it wouldn't have issues. When I need information that will be heavily impacted by recency or need it to be more consistent and less likely to contain hallucinations in its answer, I tend to use a cloud model.
Accessibility: Hosting your own models is not accessible to everyone because of system requirements. I hope to update my blog with this note (as some people have pointed out I haven't) but not every machine can run Llama or any other model. Cloud LLMs truly have made AI accessible to almost anyone w/ an internet connection
Model Selection: This isn't the biggest deal but I find discovering and working with different models somewhat cumbersome. There are a lot of open-source models that are available and it's truly hard at times to differentiate between their individual strengths and weaknesses. I'm highly invested and interested in available models but it's hard to keep with all the old and new models and versions. I find myself primarily investigating models that have already been said by others to be "good". Cloud LLMs tend to be the best versions of their existing models, tend to be pretty comparable, and there are far less of them; making it easier to choose a model.
As I continue to work with more local and cloud models, I'm interested to see how this pro-con list evolves.
This was a great question. Thanks for asking!
You really have me thinking even more deeply about all of this. haha
Thanks, yes I think most of that makes sense ... and yes, now I see that you're being featured in the "7 posts of the week" blog - not sure how I missed it, but that's great!
Last night, I built two GPTs for my personal use to help with my farm's social media and marketing practices and assist with creating content for my brand on LinkedIn. I cracked myself up when I told my wife last night while going to bed that I had just hired our first two employees. Their names are "K3bert Marketing Maven and Copy Catalyst" and "Kunekune Copy Editor and Marketing Muse."
I'm seeing more people, not just technologists, discussing ways they use AI to "supercharge" their skills as leaders, coders, marketers, etc. If there is any takeaway from your post, there is much to learn to build custom-built AIs for your specific needs, either through one of the services out there or by installing them on your machine and working from there. But, it's a journey I encourage everyone to go on, regardless of where you are in your career journey.
Love this, Thanks for the write up. I’m going to get this running for myself soon! It’s a great tutorial and actual real human experience.
Thanks @obeykinobi ! It's really been great for me. You should check out Open WebUI when you get comfortable with Ollama. I'm using it with Llama, because it gives you a nice interface similar to ChatGPT and Gemini. Hoping to write up a blog about it soon!