Gemini is Here! Does it Crush ChatGPT/Claude for Writers? 

Thanks so much for posting this. As fiction writer for many years, I'm just starting to learn the benefits of using AI as a writing partner and all your videos, NN, have been incredibly helpful.

Hey Novelist, thought you might be interested in some info on Bard. I asked it some questions, that you have likely since found out. From Bard - Context Window: My context window size is currently 20,480 characters, which represents the maximum amount of context I can access and process from the conversation history. This limit helps ensure efficient response generation and prevents data overload. Prompt Length: The character limit for prompts is 50,000 characters, enabling you to provide me with a significant amount of information to work with. However, I'll only process the most relevant sections of the prompt to maintain coherence and relevance in my responses. Dynamic Context Pruning: To handle even lengthy prompts effectively, I employ dynamic context pruning. This technique filters out non-essential context, allowing me to focus on the most relevant elements for generating concise and relevant responses. Reply Character Limit: The character limit for my replies is 5,000 characters. This ensures efficient and readable responses, even when addressing complex or detailed prompts. If I need to provide more extensive information, I'll split my responses into multiple parts.

Thank you for all your hard work! I find your videos super useful!!

What a great video - So helpful! Thank you for this!

Finally for the book description, actually give that Claude but I'd probably generate one with Bard and blend them. ChatGPT's was a big fail though. With all this said, I'm sure if you know each AI's strength's weakness, you can tweak the prompt. Because really, all this says is you've matched up the AI with your prompting style. I can easily change the prompt to make ChatGPT give me a better book description by simply highlighting what I like, such as "Include a hook" at the beiginning or "add a comp line mentioning two comparable novels" as well as "Write this in XYZ tone". Etc. IMO, it's our job as prompters to really polish our super-prompt until it gives us exactly what we want from whatever tool we have available to us.

Thanks for covering this from writer's perspective.

The thing I like about Bard is that it gives you 3 versions of it's response. That way, if you're not happy with the first response, you can choose one of the other two.

In terms of beats, it's so interesting. ChatGPT summarized the beats, Bard stripped everything down to LITERAL action beats without any of the connective tissue in between even in the "prose-ier" side of things. Claude actually did the best job in terms of actually writing active/descriptive prose, but it is SO overwritten that it has "AI wrote" this written all over it. An example: The brass bell above the carved oak door jangled as a tall, lean man with dark hair and darting eyes entered the cafe." It would have been nice if you had prompted ChatGPT differently so it actually wrote (not summarized) the story to give it a fair comparison, but on this one, I give it to Claude.

The baker thing is because you said Cozy. hahah

Good to know that Google's FREE AI has some chops!

Lots of Cozies' main characters are bakers, quilters, crafters, librarians, and such.

Hi. I have just discovered your videos. All this about AI is very interesting. Seems like a good tool for aiding writers along. I am new to all this but have watched a couple of videos and it seems like Sudowrite and Novelcrafter are the ones to keep an eye on. Specially Novelcrafter where it seems like you can switch to which ever Ai you want to use. Will Gemini appear there as well? Also, I just joined the waiting list for Novelcrafter, do you know how long it will take to get an invite? I live in Sweden so I dont know if that makes a difference? Thank you.

Im findinf Gemini advanced to be amazing! I write rhe prose and use gemini to offer feedback and my writing has improved so much!!

I actually think ChatGPT and Bard did much better with the prompt. They set up the atmosphere/tone/setting of the story which created an immersive vibe. They also gave a much clearer theme (ie. the way cats can be more keen observers than people), and a crime that is actually one worthy of a book, a mysterious interaction among clients. Claude tried to make the crime about a scratch on the cat which seems a bit silly and while Claude did include "active beats" they were pretty bare-bones action beats that any writer could have come up without AI. I can do bare-bones beat once I'm immersed in the story prompt and ChatGPT/Bard did way better at immersing me in the prompt while Claude barely immersed me.

One thing to keep in mind here is that the current version of Gemini included in Bard is the Pro version. The more powerful Gemini Ultra version will be incorporated into Bard next year. It's the Ultra version that beats Gpt4 in almost every test.

I like the prose that Claude generates it was pretty spot on with the summery I had written and gave it. It did well generating a list of chapter summery it started out well writing the first few chapters but by 7 it started going seriously off the rails spewing out disjointed sentences all detail but not making sense.

I'm using the opensource Cybertron 7B data model. It's pretty unreal so far. Textgen-WebUI is my LLM of choice running it locally with the Playground extension. That Cybertron model is unreal at writing fluent and great dialogue in my testing thus far.

For prose, I prefer ChatGPT. It feels much more professional. It doesn't quite match my style. It's a bit too simple, but I can edit something like that better. Claude's prose feel childish to me. It could be the POV as I don't normally write in that POV but it just struck me as off. And Bard's prose was NICE in it's drama and intensity, but this feels like a very specific tone you want for maybe super intense scenes, not from page 1-200 of a book. LOL Interesting experiment. All this will probably depend on the writer's personal style/taste and their genre. I write mystery/suspense/horro so my prose matches ChatGPT/Bard a bit more (still with tweaks needed to better match my style). But someone who writes another genre may prefer Claude.

All the baking made me hungry for pastries

So, I'm viewing this about a day after you posted it. I was also trying more than one. I used the exact same prompt on Claude2 and I have been arguing with Claude since I entered that prompt. Although I have used Claude for stories in the past, now she is absolutely refusing (and I've tried to trick her into going ahead anyway in a few different ways) to give any pitches. She says that as a language model, she doesn't have the context and nuances to come up with creative fiction and she is afraid that a "cozy mystery" will contain offensive material. She wants us to "talk about how we can build community or help people" instead. I KNOW this sounds like a joke or I'm making it up... and I am not at all "anti-AI". I am using AI for all kinds of things and like I said, I've even used Claude to write fiction stories. But, suddenly, she has become woke and is afraid to write anything fiction. Is anyone else having this issue or just me?

Found your content. Great info. Thanks. Can you recommend a tool/s or non fiction writing? Do you have teaching on how to create non fiction with AI?

Oof bard isn't available in Canada yet. I imagine there's a problem adding u's to words like doughnut or flavour.

We'll really start to see how all these AI models use the same core ideas when we are suddenly flooded with novels and screenplays about fortune cookie and baker murder mysteries.

Was using Claude pro early,did 27 chapters with no problem ,so ran out of messages till 8pm,came back at 8, whatever it was doing earlier it would not do it then,than claimed they did not write anything for me, trying to get customer service is a night mare, wasted another $20 same as I did on chatgpt.

Claud is impossible to use without writing G-rated material. It even told me that one of my characters should not drink and that it will no longer cooperate with the character's lifestyle, even in a fictional sense.

👍

Asked Bard if it is using the latest language model and it says its still using the one from May 2023.

Besides being free, Bard is also available in Europe, unlike Claude. So that certainly cuts down on the competition.

it is very strange , but when i wrote to Bard and Claude to write me 10 options of..... they wrote exactly the same subjects order. like you got with the bakery. here is my prompt (very poor but im not an English speaker so i wrote in hebrew): Give me ten ideas for sci-fi stories that I can write with an interesting twist in the story Here are the first 3 suggestions from Bard: The hero arrives on a new planet, but discovers that it is not exactly what he thought. For example, the planet could be Earth-like, but with completely different plant and animal life. Or, the planet could be a futuristic world with advanced technology, but with a corrupt society. Man manages to develop teleportation, but discovers that it is not perfect. For example, teleportation can be dangerous, or it can lead to unexpected side effects. Man discovers that aliens exist, but they are not what he thought. For example, the aliens can be friendly or hostile, or they can be in a technological state similar to that of man or much more advanced. Here are the 3 first results from Claude: An astronaut who stumbles upon an abandoned planet finds the remains of ancient aliens that existed there, leading him to discover ancient secrets about the universe. A scientist invents a time machine, but is accidentally sent to the far future where humanity is almost extinct and he must find out what happened and how to prevent it. A space explorer finds an extraterrestrial life form on a planet, but it later turns out to be an intelligent life form capable of taking over and killing. Its AMAZING exactly the same results

Did you ask Bard what model it's using? Gemini is not available in most locations right now. I had to VPN into San Francisco to get access to Gemini.

The common themes that keep popping up when doing these tests really shows one of the major limitations of the current generation of AI and while a Human is still absolutely vital to the equation. If we just let AI create "art" right now it would all end up being the same stuff.

Every time I hear a cat named Marmalade it reminds me of the cartoon Milly, Molly

Unfortunately it’s not available in Europe yet.

Is here... 😂😂😂

It would be nice if they were like apple but being Google they'll drop it in two years .

OK, as a beginner and non writer wanting to write I for the life of me could not tell thr difference.

Let's be clear... (as Jason mentions) the lofty claims by Google is for the UNRELEASED Gemini Ultra. So, while comparisons to the just released Gemini Pro are useful, they don't actually debunk any of Google's claims. It's comparing granny apples to red-delicious apples. :)

Cozy mysteries aren’t gonna have the best conflicts, anyway.

Maybe it will write better if you tell it what style you want.

My opinion beautiful mother saves the world 🌎 vortex a single point ☝️ boom singularity golden rose beautiful reverse beautiful holy and true love of all of them and you are very welcome blessings dear forever loving god❤

Currently Bard is running the Pro version, which is barely better than ChatGPT 3.5 turbo. Not really impressive. And the Ultra benchmarks are BS. People are already tearing it a part; it's marginally better (maybe) than GPT4. It's been a meh release from Google.

how do we access the new Gemini through Bard. Can you pin a comment with the we address

And not a single word written from imagination. This is good?