• Everyday AI
  • Posts
  • OpenAI releases specs on how ChatGPT works

OpenAI releases specs on how ChatGPT works

Expert: How to combine AI + data, OpenAI releases Model Spec framework, TikTok to label AI content and more!

Author

Everyday AI
May 09, 2024

Outsmart The Future

Today in Everyday AI
6 minute read

🎙 Daily Podcast Episode: You have an abundance of data AND the power of AI but still not leveraging it? You’re not alone. An expert tells us how to avoid AI’s data-driven decision paradox. Give it a watch/listen.

🕵️‍♂️ Fresh Finds: Zapier Central AI now in beta, Runway’s AI film festival and Cisco expert weighs in on AI disinformation and Stanford researchers find AI-powered solution to give glasses superpowers. Read on for Fresh Finds.

🗞 Byte Sized Daily AI News: OpenAI releases Model Spec framework on how ChatGPT works, TikTok to label AI-generated content and former Google CEO gives insights into the future of AI search. For that and more, read on for Byte Sized News.

🚀 AI In 5: We're checking out a GPT inside ChatGPT that lets you convert a file to almost any format you need! See it here

🧠 Learn & Leveraging AI: Here’s how to use AI to actually take advantage of that data, and not just collect it on a digital shelf. Keep reading for that!

↩️ Don’t miss out: Did you miss our last newsletter? We talked about GenAI and the U.S. labor market, OpenAI's AI image detector, AlphaFold 3 release and ChatGPT vs consultants. Check it here!

AI’s Data-Driven Decision Paradox 🔓

We have so much data at our disposal. But it’s not being used to its full potential.

Or sometimes, not at all.

Sure, AI can do data crunching wonders when we pair AI and data together.

But, what challenges and conflicts should we be aware of before doing so? And will that help us finally use AI and data to grow our companies?

Aakash Indurkhya, Head of Product at Virtualitics, joins us to discuss AI and data analysis.

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

Also on the pod today:

• Data Collection Challenges 📉
• Generative AI and Data Visualization 👁
• Overcoming the Data-Driven Decision Paradox 😵‍💫

It’ll be worth your 34 minutes:

Listen on our site:

Click to listen

Subscribe and listen on your favorite podcast platform

Listen on:

Upcoming Everyday AI Livestreams

Friday, May 10th at 7:30 am CST ⬇️

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

New AI Tool Spotlight – AlphaBot is a real-time AI chatbot for customers, Replica is voice AI and text-to-speech for creators and RateMyJD helps improve job descriptions.

AI breakthroughs — Stanford researchers created an AI-powered solution to turn any glasses into 3D holographic headsets.

Trending in AI – Zapier has announced its Zapier Central AI workspace is now in beta

AI in Society – A cyber expert from Cisco says that GenAI disinformation threats are exaggerated.

Read This – Taoping, a smart cloud platform, has launched an AI-powered smart terminal.

AI Video - Runway is hosting an AI Film Festival tonight in NYC,

1. OpenAI Shares Guidelines on Model Behavior 🧐

OpenAI's latest move shakes up the AI world as they unveil the Model Spec, a comprehensive document outlining how AI models like ChatGPT should behave in various scenarios. From emphasizing user assistance and benefiting humanity to navigating complex ethical considerations, these guidelines mark a significant step toward responsible AI development.

Good chatbot.

2. TikTok To Label AI-Generated Content 📲

TikTok collaborates with Adobe's Content Authenticity Initiative and C2PA to introduce Content Credentials for AI-generated content verification. This initiative will revolutionize content authentication standards across social media platforms. TikTok's phased implementation includes automatically labeling AI-generated uploads and effects, ensuring transparency and trustworthiness in digital content.

3. OpenAI's Considers Allowing AI-Generated 'NSFW' Content 😳

OpenAI is under a bit of fire for considering allowing the creation of not-safe-for-work (NSFW) content, including explicit and extreme material, with its AI tools. This decision has raised concerns among campaigners about the company's mission to produce safe and beneficial AI. Joanne Jang from OpenAI emphasized the importance of discussing the responsible creation of erotic content while maintaining control and legality.

4. Former Google CEO hints that AI search won’t have any links

Former CEO of Google, Eric Schmidt, predicts a future where Google's mission of organizing the world's information will eliminate the need for traditional websites and advertising will become more effective. This aligns with Google's past actions and their current efforts in AI technology. However, Schmidt's views may not reflect the current thinking within Google.

5. French Startup Mistral AI Skyrockets to $6 Billion Valuation 🚀

Mistral AI, led by CEO Arthur Mensch, is gearing up to secure a whopping $600 million in funding, driving its valuation to $6 billion – nearly tripling its value in just six months! Sheesh.

With big investors like General Catalyst and Lightspeed Venture Partners onboard, Mistral AI is primed to take on the AI giants of Silicon Valley.

ConvertAnything with this GPT inside of ChatGPT

We're checking out a GPT inside ChatGPT that lets you convert a file to almost any format you need!

ConvertAnything lets you convert files to video, text, audio, compressed files, and more.

What good is all that LLM work if you just spend an hour copying and pasting and reformatting to save it?

🦾How You Can Leverage:

97% of data goes untouched. 

Yikes.

If data is the new oil (or gold, or whatever) then why are so many companies failing to put their data to use?

Especially when Generative AI makes the process so much easier? 

Tough questions. 🤔

Aakash, is the Head of Product at Virtualitics, an AI-driven analytics platform for interactive data exploration and informed decision-making.

And he broke down a common problem plaguing so many businesses —

  • Companies have the data. They all want to use it to make better decisions. But the majority of companies fail to do so. 

Whether you’re an enterprise leader wrangling terabytes of data or a small business owner side-eying a mountain of spreadsheets building up on your desktop, today’s convo is for you.

  • Data’s easy to find. 

  • It’s hard to leverage. 

  • Let’s change that. 👇

1 – Make your data agree 🤝

Data can legit be anything. 

  • The number of visits on your website vs. the number of sessions on your website. 

  • The customer satisfaction ratings vs. customer retention rates. 

With so much easy access to so many (or too many?) data points, Aakash emphasized the importance of making sure your data agrees with itself. 

He highlighted the significance of ensuring data alignment within an organization to facilitate accurate decision-making.

Before you even sprinkle any AI magic, you’ve gotta make sure you have clean, clearly labeled and harmonious data sets. 

Try this:

Cleaning, organizing and classifying your data aint always easy. 

Here’s a great walkthrough from Unito on properly preparing different types of data to work with each other. 

(And yes, you can actually use AI to help prepare your data to be used by AI.) 

2 – Talk with AI for data answers 🗣

Halt, who goes there? 🖐

When we talk data and large language models, there’s a few pretty big misconceptions and low-hanging fruit that hardly anyone’s picking. 

Aakash broke it down, breaking down both sides of it. 

  • First — LLMs allow you to use plain English (or your language of choice) to figure out what the heck your data is. 

  • Once a model has explained what the heck the data is, you can then have it actually compute or manipulate the data however you need to. 

Try this: 

*disclaimer — never upload confidential, sensitive or proprietary information into LLMs unless you’ve been told it’s OK.*

Let’s say you’re staring at a spreadsheet with more unidentifiable rows than a boat race in another country. 

  1. Just ask ChatGPT what the data is and what it means.

  2. Give it context on what your business is and what (you think) the data is about.

  3. Then, ask it questions about trends or outliers in your dataset.

Watch the magic.  

And just like that — you’ve gone from confusion, to on your way to making numbers in a database work for you. 

We broke down some of the basics (and advanced ways) you can use ChatGPT for quick data analysis here. 

Click to watch the video

3 – Humans connect the data points 🧠

When you can get clean data and combine the best-practices of using LLMs for data analysis, you’re starting to realize some golden opportunities. 

Yet, it begs the question — If Generative AI systems are getting better and better at making sense of complicated data in a fraction of the time, what does that mean for the humans? 

You know … the smart data scientists who are normally the ones deep-diving into BI Dashboards and wizarding Excel formulas? 

Aakash envisioned a different role in the AI-human data journey — as a creative data connector. 

He gave the example of HVAC data:

  • Maybe you have one dataset showing when HVAC systems break, and another that just shows the number of visits that a technician makes to repair HVAC systems. 

But, those two things aren’t related. And scenarios like this are common in data science and analytics. 

As LLMs can carry more of the traditional data-crunching load, data scientists may be able to step out of the cell-by-cell process and build better data pipelines and connectors in the real world. 

Try this:

No one has a crystal data ball.

But, the gap of that 97% of data being untouched will likely start to shrink as LLMs become more widely used for repetitive data decision-making work. 

So what does the future of human work in data analysis look like?

➡️ This article from Geeks for Geeks takes a look. 

⌚

Numbers to watch

60%

60% of leaders worry their company lacks a vision and plan to implement AI.

Now This …

Let us know your thoughts!

Vote to see live results

When it comes to AI and data, what do you wanna learn more about

Login or Subscribe to participate in polls.

Reply

or to participate.