When Google's Bard entered the AI world, it was making a grand entry, full of promise. But then, things took a rather dramatic turn. During a demo, Bard made a blunder that was let's say messy. The aftermath? Google's stock dipped by 8%. Ouch.📉😬
Despite the fanfare, Bard hasn't been able to weave the same magic as ChatGPT in winning the market and conversations. With ChatGPT's plugins leading the race, Bard seemed like it was trailing behind in a marathon, huffing and puffing.
But here's a plot twist - Bard hasn't thrown in the towel yet. It recently announced 'extensions' (definitely not calling them plugins, because that's so been there, done that).
What's interesting is that Bard isn't just about catching up, it has a few tricks up its sleeve. There are some cool features that could make even the naysayers sit up and take a second look. 🚀🔮
🧑✈️ Bard's Features
It's time to roll out the red carpet for Bard's feature lineup. And they are quite interesting.
1️⃣ Power-Packed Performance: Bard operates on the PaLM 2 language model, which flexes almost three times the parameters as GPT-3.5. As for GPT-4's actual parameter numbers, they're still a mystery so we cannot compare with PaLM 2.
2️⃣ Voice Command: Bard is all ears with its built-in voice input. Just speak your instructions, and it swings into action. It's like having your very own virtual assistant who never goes on tea breaks.
3️⃣ Web Support: While ChatGPT relies on a plugin to surf the web, Bard is a natural, natively supporting web access.
4️⃣ Choose Your Draft: Bard doesn't believe in one-size-fits-all. For every response, it generates multiple drafts, giving you the liberty to pick the one you fancy. No more need to give a separate prompt and ask for multiple versions.
5️⃣ Docs and Gmail Integration: Now, here's where Bard truly shines. Any tabular data it generates can be effortlessly copied into Google Sheets. Compare that with ChatGPT, where you'd have to ask it to export the data to csv. Moreover, you can export any result to Google Docs or Gmail, and continue working on it from there.
While Bard may have had a rocky start, it seems like it's setting the stage for a promising second act with these exciting features. Who knows, Bard might just have what it takes to keep us on the edge of our seats, eagerly anticipating what it does next. 🍿🎬
🧩 Bard Steps up its game in reasoning
Google's AI chatbot, Bard, is leveling up.💪 Recent reports from Google indicate that Bard is getting better at tasks that require logic and reasoning.🎯 This improvement comes from a technique called "implicit code execution," which has enhanced Bard's performance in areas like math and coding.📊💻
Here's the deal: Large language models (LLMs) like Bard are essentially designed to predict what comes next in a conversation or a sentence.🔮 This skill is great for writing tasks like crafting emails📧 or essays📄, but when it came to coding, Bard was a little wobbly.😅
"But what about coding experts like GitHub's Copilot and Amazon's CodeWhisperer?" you may ask.🤔 These models are different. They focus mainly on coding, unlike Bard and its peers (like ChatGPT), which are trained on a variety of text samples from all over the web and other resources.🌐📚
So, Google decided to help Bard step up its game in coding and mathematics. Thanks to the new technique of implicit code execution, Bard can now write and even run its own code.💡 It works like this: Bard finds prompts that need a touch of code, it writes the code quietly in the background, checks it, and then uses the results to provide a better response.🎈
With this new upgrade, Bard is expanding its skill set and becoming more versatile in handling diverse tasks.🏋️♀️ It's a promising development, and we're keen to see how this will shape Bard's future capabilities.🚀🌟
😃 AI Humour
Tool of the Day
Snipd
Snipd is a cool new tool you can use to discover highlights in podcasts. Listen to the best highlights from the podcasts, mark the portions you love and share it with others also. This is good for people who binge on knowledge podcasts and deep dive interviews.
💬 Prompt of the Day
I'm currently learning about [email marketing]. Ask me a series of [five] questions, one at a time, that will test my knowledge. Wait for my response before proceeding to the next question, ask the next question after your explanation of the answers. Identify knowledge gaps in my answers and give me better answers to fill those gaps. When finished show me the number of correct answers and the quantity of failed answers.
Reference: Rodolfo Azevedo
This prompt can help you have your own AI study partner.