How Data Will Be AI’s Bottleneck

OpenAI's Advanced Voice Model is live, Microsoft AI earnings take a dip, how to properly implement data with AI and more!

Author

Everyday AI
July 31, 2024

Outsmart The Future

Today in Everyday AI
7 minute read

🎙 Daily Podcast Episode: Running into bottlenecks while implementing AI? It could be your data. We break down how to tackle AI and data. Give it a listen.

🕵️‍♂️ Fresh Finds: Google drops new Gemma small language model, Runway is going turbo, Meta apologizes for AI mishap, Brazil’s $4B AI plan and Google Photos AI editing tool now available. Read on for Fresh Finds.

🗞 Byte Sized Daily AI News: OpenAI launches Advanced Voice Model, Microsoft’s AI earnings take a hit and AMD shares skyrocket. Kaboomy. For that and more, read on for Byte Sized News.

🚀 AI In 5: Google JUST released its new AI image model, Imagen 3. We dive in and see how it stacks up against the competition. TBH, we’re kinda digging this. See it here

🧠 Learn & Leveraging AI: Having issues implementing your data with AI? Here’s how to make sure your data is set up properly. Keep reading for that!

↩️ Don’t miss out: Did you miss our last newsletter? We talked about Meta unveils SAM2, Apple uses Google TPUs for its AI, Meta adds AI Studio to social platforms. Check it here!

How Data Will Be AI’s Bottleneck 📊

Yeah, AI is cool. 

But have you tried AI WITH good data?!

If you're running into AI implementation bottlenecks, it could be your data to blame.

Matthijs de Vries, Founder & CEO of Nuklai, joins us as we tackle one of the elephants in the AI room -- data.

Join the conversation and ask Jordan and Matthijs questions on AI and data here.

Also on the pod today:

• Data and Large Language Models 🤖
• Practical Data Strategies with AI 📊
• Data Quality Issues ❌

It’ll be worth your 31 minutes:

Listen on our site:

Click to listen

Subscribe and listen on your favorite podcast platform

Listen on:

Upcoming Everyday AI Livestreams

Thursday, August 1st at 7:30 am CST ⬇️

Friday, August 2nd - AI can (almost) do your CEO's job

Here’s our favorite AI finds from across the web:

New AI Tool Spotlight – Simply Draw gives you AI feedback to enhance drawing skills, Lingoedit is an AI-powered translator and Creatr is an AI design tool to build products.

Trending in AI – Meta has apologized for its AI saying that the Trump rally shooting never happened.

AI Video Production — Runway says its newest ‘Turbo’ model produces video 7X faster. Whooooa buddies.

Google - Google’s Circle to Search feature now includes an “About This Image” tab that can check if there’s a Google AI watermark. Speaking of images, Google Photos AI editing tool is now available to all Google Photo users.

Money in AI – Brazil is proposing a $4 billion AI investment plan.

Small Language Models — Google just released a 2 billion parameter update to its small language model, Gemma, in Gemma 2. Annddddd…. it’s looking pretty good.

Check This Out – Wix’s AI can now write entire blog posts for you. Can’t wait to delve into them! lolz.

AI Hardware — There’s a new AI-powered hardware device called, Friend, and it’s giving off very strong vibes.

1. OpenAI Unleashes Advanced Voice Mode To Select Users 🗣

OpenAI has just rolled out its Advanced Voice Mode, giving a lucky few ChatGPT Plus users a taste of the mind-blowing GPT-4o audio responses. Initially showcasing a voice that was eerily similar to Scarlett Johansson's, OpenAI quickly backtracked after she raised concerns and even hired legal counsel to protect her likeness.

Now, testing is ramping up with four new preset voices, leaving the drama behind but keeping the excitement alive as they monitor user feedback.

2. Microsoft’s Earnings Disappoint as Azure AI Growth Slows 📉

Microsoft's latest earnings report revealed a 29% growth in Azure, falling short of the previous quarter's 31%, disappointing investors. The Intelligent Cloud segment generated $28.5 billion, yet shares dropped 7.8% following the announcement.

Despite this dip, Microsoft continues to invest heavily in AI, spending $19 billion in capital expenditures this quarter.

3. AMD's Shares Skyrocket as AI Chips Drive Growth 🚀

Advanced Micro Devices Inc is on fire, with shares up 11% after the chipmaker dropped a bombshell revenue forecast of $67 billion for Q3, outpacing analyst expectations of $66.2 billion. CEO Lisa Su is buzzing about their MI300 AI processors, predicting a whopping $4.5 billion in sales this year—up from the previous $4 billion target!

Meanwhile, AMD's second-quarter revenue soared by 89% to $5.84 billion, making it clear that they’re ready to take on NVIDIA in the AI accelerator race.

4. OpenAI Teams Up with U.S. Congress to Shape AI Regulation 📜

OpenAI is making serious moves in D.C. by backing three major Senate bills, including the Future of AI Innovation Act, which could officially establish the U.S. AI Safety Institute to regulate AI standards.

In a LinkedIn post, VP Anna Makanju emphasized the importance of government involvement in ensuring AI is safe and accessible, while also hinting at OpenAI's desire to cozy up to lawmakers as they face potential scrutiny. The bills also aim to allocate federal scholarships for AI research and create educational resources

5. Meta Pulls the Plug on Celebrity AI Chatbots 🔌

Meta has officially discontinued its celebrity chatbot feature, which had generated excitement since its introduction at last September’s Connect event. The company is now prioritizing its new AI Studio, allowing creators to design their own chatbots instead of relying on celebrity personas.

Despite investing millions in celebrity likenesses, Meta has shifted focus to more customizable options for a broader audience.

Is Google's Imagen 3 BETTER than MidJourney?

Google JUST released a new image generation model - Imagen 3.

Is it better than DALL-E 3 or Midjourney?

We show you how to access Imagen 3 and give a head-to-head comparison.

🦾How You Can Leverage:

Picture this: You're at a fancy restaurant, ready to dig into a gourmet meal. 

The chef's got Michelin stars. Waiters that look like they drive BMWs. The place is legit. 

But there's a problem - your fork's the size of a toothpick, and your knife's duller than a corporate memo. 

That's what it's like trying to do AI with messy data — just showing up won’t get ya fed. 

We talked about the data elephant in the AI room today with Matthijs de Vries, the CEO of Nuklai, and he's got some ideas that'll flip your data strategy on its head.

 Forget what you've heard about data being the new oil. 

Matthijs says data is your company’s oxygen. 

Without it, our AI dreams are gonna flatline faster than a startup's funding in a recession.

 Let's dive into how we can give our data strategy some serious CPR.

1, 

2, 

33333333

1 –Break Down Those Data Silos 💣

Imagine you're a detective trying to solve a complex case.

You've got witnesses, evidence, and a hunch. But here's the kicker - each piece of information is locked in a different room, and you've only got keys to a few. 

Well, you’re not even sure if you have keys. But Billy in IT might. 

Frustrating, right? That's what's happening with company data right now.

Your sales data's lounging in Salesforce, customer support logs are having a party in Zendesk, and who knows what secrets are hiding in those dusty Excel files on Bob from accounting's desktop. Did he upload those in OneDrive? 

Matthijs says we need to tear down these walls. The goal? Create a data democracy where every bit of information is accessible, regardless of its origin.

Try This:

Channel your inner data detective.

  • Pick three key data sources in your company.

  • Maybe your CRM, support ticket system, and financial software.

  • Now, set up a simple data lake create a unified access point. 

Start small and get your data talking. And when your data is talking, that’s when you can start putting GenAI to work. 

And while you’re at it, Inspector Gadget, go peep this article on 4 Data Cleaning techniques to improve LLM performance. 

2 – Data Context is King 👑

Let's play a quick game, ya’ll.

I'll give you a number: 42.

What does it mean?

  • Is it the answer to life, the universe, and everything?

  • The number of pancakes Jordan ate last month? (Don’t judge. Prolly accurate TBH)

  • Or is 42 your company's latest sales figures for thingamabobbers? 

Without context, it's just a number floating in space.

This is the problem Matthijs says we need to tackle head-on. Raw data is like handing someone a bunch of random ingredients and expecting them to whip up a gourmet meal. (Assuming we have a real knife and fork now.)

Ingredients need a recipe and a chef like data needs labels an context. 

Think of adding context like giving your data a compelling backstory. It's the difference between "Number of Sales: 1000" and "Number of Sales: 1000 (Enterprise clients, Q2 2024, Before new feature launch)". 

Suddenly, that number has a rich history and implications for future strategy. And when you combine the LLM powers of working with unstructured data AND structured data, you’re now driving a faster car than that waiter. 

Try This:

Pick your most-used dataset and play "Data Archaeologist". Dig into its past, uncover its secrets, document its quirks. Create a "Data Biography" that answers:

  • What's the source of this data?

  • How often is it updated?

  • What does each field represent?

  • Are there any known issues or biases?

  • How has this data been used in the past?

  • What decisions have been made based on this data?

Share this biography with your team. Now your data’s got a story shorties.

And when you give you data context, a LLM can really start connecting the dots.

3 – Crowdsource your company’s AI oxygen 🌬️

If data is oxygen, shouldn’t we all be…. collecting all that yummy internal oxygen inside our own doors? 

Yuuuuup.

When it comes to data, our own companies, humans, processes and docs are legit drippin with more golden bars than Simone Biles in Paris. 

Jordan and Matthijs agreed that a future business boon is companies turning their IP and domain expertise into structured and unstructured data. 

And you wanna get real meta here? 

You can use LLMs to help collect, transcribe and categorize all that internal domain expertise and then feed all that deliciousness back into….. you guessed it! …. fuel for your own LLM. 

In the corporate world, this means tapping into the collective knowledge of your entire organization. From the CEO to the newest intern, everyone's got insights that could be turned into valuable data. 

Try This:

Launch a company-wide "Data Discovery" challenge. Here's how:

  • Start recording literally every internal meeting

  • Over communicate everything, and include more context, details and processes in decision making

  • Once you have a few hundred hours of high quality content, take all them juicy transcripts and LLM the snot outta them, and you’ve got the basis for a great AI assistant. 

Need help on what that looks like? 

Or maybe you need a Copilot mentor or ChatGPT trainer? 

Hit us up, and we’ll get you on the right path y’all. 

⌚

Numbers to watch

$11 Million

Sybill has raised $11 million for its sales AI assistant.

(Or your fave LLM like Claude, Gemini, Copilot, etc)

Now This …

Let us know your thoughts!

Vote to see live results

Do you attend our livestreams?

Every weekday, we bring you fresh AI insights, exclusive interviews, and breaking news with our Everyday AI livestream.

Login or Subscribe to participate in polls.

Reply

or to participate.