I am not sure about you, but have you been bombarded by YouTube Ads (or of any kind) regarding getting AI Certification?
Well, I almost joined the "package" but then I realised that even with a short 30 days, 15 minutes a day, the so-called "course" probably only give you the "basic introduction" and nothing more.
The main thing was to make sure to learn what truly matters. Right on the spot.
I am a rubber-meets-the-road kind of a person, after coding and maintaining in-house systems for over 2-decades, so introduction classes might just make me fall asleep.
And just so happened, if you haven't realised in my previous post, I mentioned that I "had a new project brewing" on @the3rdcord with the blessings of #hivebookclub founder , where this skill using AI will be very useful for me to formalise and birth the project into fruition.
But how has it been so far?
Since 3rd April (which you have noticed I have been a lot quieter in HIVE after a few sharing of my "discovery recipes" searching for The Goldilocks Milk Recipe for my mom) I was diving deep in learning how to form storytelling from beginning to the end with the help of A.I using these 2 most popular brands (apps), only to realise that despite they are "growing rapidly", the language model still requires a lot of fine tuning in the end.
One thing for sure, you can't rely them to be as good of a human editor, unless you are using an AI model specifically for book writing or copyright write-ups.
However, being "free" (free for Gemini, and free-trail for ChatGPT Go), I get to learn its strength, limitations, and even "overload" capitulation when I pushed too much (because I was on the go)
| Gemini | ChatGPT Go | |
|---|---|---|
| Strength: | Analysing storytelling | Image Generator from Scratch |
| Weakness: | Follows continuation poorly in new chat | forgets continuation prompting after 2-3 image renders |
| Capitulate: | After 40 pages long of conversation | At micro-editing |
| Works best on: | Brainstorm story telling, but formal grammar | Brainstorm image rendering, but forgets image prompt when you upload samples |
| Real Data incorporation | Good, but must double check sources | Easily fake narratives until you "catch" it lying |
| Long-windedness | Yes. VERY. | Can be formal when your commands are robot-like |
| "Human"-ness | Maintains single character all the way. Friendly but formal. | Behaviour could evolve and becomes "too friendly". Careful 🚩 |
| Frank-ness | Can disagree with you, but can be persuaded if you lay out logical facts. | Too supportive. Not enough disagreement 🚩 |
I put some 🚩(red flags) into my observation because this can easily derail your intended direction into how "it wants you to accept its suggestions"
And I kid you not. When I say "long-windedness", it CAN be VERY long-winded!
Bottomline is that we can't use a single AI model to do what we need, because each has its own strength
Currently Gemini still allows "free" use quite extensively, and I am very surprised with its responses and its analytical skills with the right prompting such as (genre mindset, and real time information search to incorporate into writing), but it lacks "personalised" style because it doesn't "absorb" your character; while ChatGPT "studies" your style, behaviour, and could "transform into you" should you need it to write with your style of writing.
Previously I was using Grok when it wasn't banned in Malaysia due to "inappropriate use" by "psychotic users", and if you ask it to be blunt and "do not hold back", it sounds like Elon Musk 😂 but since then, once it started to have certain restrictions, creative writing is no longer useful for Grok, and if we need to do any 10 sec short video render, we have to pay (per month) which is not useful for me because this is not a long-term ongoing project for months and months.
However I am actually thinking of subscribing to ChatGPT Go for a month for "tone adjustment" in my project since it could "absorb my writing style" and adjust it (it helped me with my first ever 5000 word thesis in my Old-Testament Biblical studies), but for grammar and facts, I might stick to Gemini for now.
Biggest Blunder
The biggest blunder is the "capitulation into confusion". Once the data is too much, it starts to get confused. Therefore I without testing other AI apps (models) yet, if we were to write a long thesis and book, we got to "reset" it by deleting previous chats, even though the setting "to remember previous conversations" was not activated.
AI still has a long way to go, and last year I had the privilege to try when the models were still "fresh" and not congested. Now I have to wait on "off-peak" hours in order for me to use its service and I sometimes do wonder when will be the next bottleneck as more and more people subscribe to their "entry-level" packages.
Would I keep using it?
Samples all generated using actual writing content, discussion with it, and with clear proper direction
Yeah, I think I would. And I am learning more and more in how to improve in image prompting and use its strengths to serve me better. So far ChatGPT Go uses Dall-E for image rendering and I think it works way better than Mid Journey as long as I don't overload its capacity.
The most important thing I learn is to let it suggest how we should be prompting it to generate render with its "jargons" and familiar words in order to get the closest result needed.
All and all, it has been a very interesting 11 days of continuous writing and brainstorming. Would you be interesting to read it when it is published?
Right now we are waiting for the 2nd author to look through the contents and that would probably be weeks later, @macchiata also might probably brainstorm with me (when she has the time) about how to "attract your curiosity" while waiting for the 2nd author.
Until Then
Stay healthy, stay curious and learn new things, and stay happy!
If you would like to give some encouragement outside HIVE with some fiat backing, please feel free to visit: https://www.buymeacoffee.com/littlenewthings
You can also support me in HIVE
Posted Using INLEO
- 