- Everyday AI
- Posts
- Google's Gemini Gems - What they are and how they compare to custom GPTs
Google's Gemini Gems - What they are and how they compare to custom GPTs
Google Gems explained, Big tech and US Gov. partner, California AI bill advances and more!
Everyday AI
August 29, 2024
👉 Subscribe Here | 🗣 Hire Us To Speak | 🤝 Partner with Us | 🤖 Grow with GenAI
Sup y’all 👋
Over the past few months, there’s been a major shift toward AI-powered coding and development.
Thanks to AI, of course, people with about zero technical knowledge can create working applications in minutes.
Wondering — would you want us to cover anything like this in future episodes?
What type of technical coding and development episodes, if any, might you wanna see?Click to see live results |
✌️
Jordan
Outsmart The Future
Today in Everyday AI
7 minute read
🎙 Daily Podcast Episode: Google just released Gemini Gems, its own version of custom GPTs. So what are they and how do they work? We break it down. Give it a listen.
🕵️‍♂️ Fresh Finds: NVIDIA earnings shake AI world, OpenAI’s new funding round and Apple’s new AI image feature. Read on for Fresh Finds.
đź—ž Byte Sized Daily AI News: Apple Intelligence lets publishers opt out, OpenAI and Anthropic partner with US Gov and California passes AI bill. For that and more, read on for Byte Sized News.
🚀 AI In 5: There’s one major feature in Google’s new Gems that is a game changer for LLMs. BUT it might not work how it should. See why here
🧠Learn & Leveraging AI: So what exactly is Google Gemini’s new Gems and how does it stack up to OpenAI’s GPTs? We break it down. Keep reading for that!
↩️ Don’t miss out: Did you miss our last newsletter? We talked about an EY leader dishing on AI adoption, Google Gemini 1.5 update and Microsoft Copilot gains traction. Check it here!
Google's Gemini Gems: What they are and how they compare to custom GPTs đź’Ž
Google silently released its answer to custom GPTs with its Gemini Gems.
In short, Gems allows users to create a smaller or more customized version of their big model, Google Gemini.
↳ But how does it work?
↳ Is it any good?
↳ And how does it compare to ChatGPT's custom GPTs?
Join the conversation and ask Jordan questions on Google Gemini here.
Also on the pod today:
• Demo of Google Gems 🧑‍💻️
• Features of Google Gems ✨
• Comparison between GPTs and Gems 🥊
It’ll be worth your 46 minutes:
Listen on our site:
Click to listen
Subscribe and listen on your favorite podcast platform
Listen on:
Here’s our favorite AI finds from across the web:
New AI Tool Spotlight – Creatify creates viral videos in one click, CompanyGPT is a search engine for companies and people and Develyn is an open-source AI agent for developer relations.
Trending in AI – NVIDIA’s recent earnings report has shown that AI spending continues to boom, despite NVIDIA’s stock dropping slightly.
New AI Models — Magic dropped a new language model with a very impressive 100 million context window?! Whoooa.
OpenAI - OpenAI is looking to close a new funding round at a $100B+ valuation.
Apple – Apple’s new iOS developer beta lets you remove objects from pictures using AI.
Meta – Meta released a statement talking about how its Llama AI models are being used by banks, tech companies and more.
AI Chips - IBM Cloud will offer Intel’s Gaudi 3 AI chips next year.
1. Apple's AI Tool Sparks Publisher Showdown 🍎
Apple's latest web-crawling tool, Applebot-Extended, is causing quite the stir among publishers, allowing them to block their data from being used in Apple's AI training. With major players like The New York Times and Vox Media saying "no thanks" to Apple's bot, it's clear there's a growing tension over who controls the digital content.
As publishers scramble to update their robotstxt files, it seems many are still in the dark about this option.
2. AI Giants Partner with U.S. Gov. for Enhanced Safety Measures 🛂
OpenAI and Anthropic have announced a significant partnership with the U.S. Artificial Intelligence Safety Institute, granting early access to their advanced models. This collaboration aims to ensure the safe deployment of powerful AI algorithms as governments worldwide focus on evaluating the risks associated with these technologies.
The Biden administration’s initiative emphasizes the importance of responsible AI development, showcasing a commitment from these companies to prioritize safety alongside innovation.
3. Google Imagen 3 to Allow AI-generated Images of People đź‘Ą
Google’s Gemini Imagen 3 generator is set to allow users to create images of people while steering clear of past blunders! After a hiatus due to some eyebrow-raising historical inaccuracies, like multicultural German soldiers from 1943, the tool will now avoid photorealistic identifiable individuals and minors.
With improvements in tech and safety measures, Gemini promises a more responsible approach to AI-generated imagery.
4. California Advances AI Regulation with SB 1047 ⚖️
The California State Assembly has passed the Safe and Secure Innovation for Frontier Artificial Intelligence Models Act (SB 1047), marking a significant step in regulating artificial intelligence. This important legislation mandates that AI companies implement safety measures, such as the ability to swiftly shut down their models and protect against unsafe modifications.
While some industry leaders, including OpenAI and Anthropic, express concerns over its focus on catastrophic risks, Senator Scott Wiener argues it’s a necessary framework to address foreseeable dangers.
5. Google’s GameNGen AI Revolutionizes Real-Time Gameplay 🎮
Google researchers have achieved a significant milestone with their new AI system, GameNGen, which generates real-time gameplay of the classic shooter Doom without a traditional game engine. This innovative technology operates on a single Tensor Processing Unit (TPU) and delivers impressive performance at 20 frames per second.
The potential impact on the $200 billion gaming industry is profound, as it could reduce development times and costs, making game creation more accessible to indie developers and individual creators alike.
Google Gems: The one fatal flaw you need to know
There’s one huge feature inside of Google’s customizable version of Gemini, Google gems, that could change the game.
It’s the ability to use dynamic documents.
But does it work? We dive in.
Check out today's AI in 5.
🦾How You Can Leverage:
You ever make your own AI?
It’s kinda magical. (When it works.)
That’s essentially what Google Gems is — a customizable LLM that anyone who can type can build. 💎
After announcing their customizable version of its Gemini LLM back in May, Google finally rolled out the bite-sized, customizable version yesterday.
So obvi we gave it a few test drives to save you time.
So, what are Google Gems?
How do they work?
And are they better than ChatGPT’s custom GPTs?
We answered all those questions live today, so go give today’s show a watch/listen.
Enough chit chat.
Here’s what you need to know about Google Gems.
1 – What Google Gems are 💎️
Google Gems are like AI assistants you can whip up without breaking a sweat.
Think of them as customizable versions of Gemini that anyone who can type can create.
But here's the catch – you'll need to shell out $20 a month for a Gemini Advanced account if you wanna be a builder, Bob.
If you're rolling with a free account, you're stuck with just 5 pre-built Gems.
Also, don't get too excited about the latest and greatest AI tech here.
Gems are actually running on the older Gemini 1.5 Pro model, not the shiny new updates Google's been bragging about.
And unlike their GPT cousins, you can't upload specific documents to Gems. More on that in a bit.
So in short, Gems are:
- Customizable versions of Gemini
- Require a $20/month paid account to build/create
- Are based on an older version of Gemini 1.5 pro
- Very easy to create
Try this:
Yesterday, we did a quicker review of Gems.
If you want a more bite-sized hot take on Gems, go check this video out.
2 – How Gems Work 🛠️
Building a Gem is easier than making instant ramen.
Well, assuming you can toggle them on. It’s as simple as 1-2-3:
You give it a name, write some instructions (or let the "magic wand" expand your simple ideas into AI-speak), and then take it for a test drive.
Sounds great, right?
Well, pump the brakes shortiessss.
Our tests showed that the whole “Google integration” thang to your live data is a bit finicky.
One of Gems biggest selling points?
They’re supposed to tap into your Google Docs, Drive, and Gmail like a pro. But in reality, they're more like that intern who misfiles everything.
(Gosh dernit Billy!)
We tried to make a Gem that would only search specific emails, and it went rogue, pulling info from newsletters we forgot we even subscribed to.
And forget about using your Gem across multiple chats or mentioning it in regular Gemini conversations – that's a no-go.
Try this:
Gems are super easy to create, fosho. But, the real perceived value is the ability to connect with live docs. Right now, that’s an advantage no one else has.
Claude Projects? Have to upload static files.
Custom GPTs? Have to upload static files.
Gems?!?
The promise of live/collaborative files is super exciting. But after our show today, we actually dug in deeper to do additional testing.
The result?
Not great.
3 – How They Compare to GPTs 🏆
Alright, let's not sugarcoat it – GPTs are still wearing the crown in the custom AI kingdom.
Here’s our line-by-line breakdown of important features between Google’s Gems and OpenAI’s custom GPTs.
And let’s just put this out there.
We’re not biased.
We just like what’s good and don’t like what’s …. Not good.
And right now, Gems are…. Not good.
GPTs are running on the beefier GPT-4o model, they let you upload specific files, play nice with third-party APIs, and you can even build them by just chatting.
Plus, you can share GPTs with your team or the whole world if you're feeling generous.
Gems?
They're like that kid at the party who doesn't know how to mingle – stuck in their own little world.
The one ace up Gems' sleeve was supposed to be that live document access, but right now, it's more of a wild card than a reliable feature.
Try this:
Take a task you've successfully set up as a GPT and try to recreate it as a Gem.
Pay attention to the differences in the setup process, the types of instructions you can give, and the end results.
Then, try using both for the same task and see which one makes you pull out less of your hair.
Bonus points if you can find something Gems do better!
What do you think? Let us know your thoughts!
⌚
Numbers to watch
$150 Million
GitHub Copilot competitor Codeium has raised $150M at a $1.25B valuation.
(Or your fave LLM like Claude, Gemini, Copilot, etc)
Now This …
Let us know your thoughts!
Vote to see live results
How much trust do you have in our reporting? |
Reply