TOGETHER WITH

Good morning, human brains. Welcome back to your daily munch of AI news.

Here’s what’s on the menu today:

• Your ChatGPT just got way better. Here’s how 🤖 🤙

  Access all of GPT’s functionalities without switching models and more.

• NVIDIA’s AI: the new “chicken or the egg?” 🐣❓🐓

  ChipNeMo is an AI model created to accelerate AI chip design.

• AI safety regulations: A topic that induces warm fuzzies 🏛️ 🤡

  You must share your AI’s results with Uncle Sam before releasing it.

 Sponsor Bot Eat Brain | New here? Subscribe!

MAIN COURSE

ChatGPT can read your PDF 🤖 🤙

In August, we reported on OpenAI’s updates to ChatGPT. It got fine-tuning and API updates for GPT-3.5 Turbo.

In September, we covered ChatGPT’s multimodal updates. This allowed it to see, hear, and speak.

Earlier this month, we covered ChatGPT’s DALL-E 3 integration. This gave you text-to-image capabilities directly inside of ChatGPT.

The suspense is killing me.

ChatGPT’s been juicing. OpenAI released an “All Tools” and a PDF/document analysis tool.

The “Tools” update allows you to seamlessly access all of GPT-4’s functionalities without needing to switch to your preferred model.

The “Documents” update allows you to upload and analyze several different document types, including PDFs.

We checked, no cap:

Let’s start with one about the tools.

Roger that. Instead of manually selecting modes like “Browse with Bing” from the GPT-4 dropdown, the chatbot now predicts your need based on the context of your text input.

Rather than going to the dropdown and selecting “DALL-E 3,” you can just tell ChatGPT to generate an image of a gorilla joining the circus.

Behold:

Okay, cool. What about the documents?

With its new enhanced document handling feature, you can upload and engage with PDFs and other documents.

Wow, this with “plugins” is going to be a game-changer.

Sike. As of now, GPT-4’s “All Tools” feature doesn’t include ChatGPT plugins.

SPONSORED BY GAMMA

Ideas are beautiful. Present them that way.

Ditch the design drama.

Instead, transform even the most intricate of your ideas into stunning slide decks in just minutes with Gamma.

😰 Too many ideas, too little time?

Say goodbye to late-night slide struggles. Let Gamma turn your ideas into visual masterpieces.

🚀 Speed up your design process.

Quickly create captivating presentations.

📱 Go beyond static slides.

Every slide is interactive and mobile-responsive by default.

🎨 No design degree? No problem.

Gamma does all the heavy lifting, so you look like a design pro.

And it’s a web app - so no downloads.

The best part? It’s free.

BUZZWORD OF THE DAY

SIDE SALAD

AI designs AI chips that design AI 🐣❓🐓

Is that you, Bot? 😱

Last week, we reported on NVIDIA Research’s Eureka. It’s an AI agent that creates algorithms to teach robots new skills.

On Friday, we covered NVIDIA Research’s HITL-TAMP. It’s a robotic learning method that combines two standard approaches to train robots more efficiently.

So, we’re covering Krispy Kreme?

Close. It’s NVIDIA again.

Yesterday, they published a paper on ChipNeMo. It’s an AI model that’s designed to accelerate semiconductor creation for AI chips.

So, what the hell is ChipNeMo?

NVIDIA researchers took LLaMA and trained it on 24 billion tokens of chip design data.

Then, they fine-tuned it using over 1,000 real-world examples from NVIDIA’s designers. This new LLM is ChipNeMo.

Here’s how it holds up:

What do you do with it?

NVIDIA is using ChipNeMo as a chatbot for engineers. It specializes in software script generation, and summarizing bug reports.

They’ve also made a code generator to assist designers in writing software for chip design.

So, long story short, it will streamline chip design.

That’s correct. Bill Dally, NVIDIA’s Chief Scientist, said, “This effort marks an important first step in applying LLMs to the complex work of designing semiconductors.”

A LITTLE SOMETHING EXTRA

More AI safety tomfoolery 🏛️ 🤡

A couple of weeks ago, we reported on the state of AI safety in 2023. We took a comprehensive look at the recent history of AI policy-making.

More regulations, perhaps?

Indeed, detective.

Yesterday, The White House issued an executive AI order. Allegedly, it’s to set new standards to ensure the safe and responsible development of artificial intelligence.

I’m not going to read that.

That’s why we’re here, friend.

Enjoy this easily scannable list of the highlights:

• The order calls for non-discrimination in AI algorithms to prevent bias, especially in areas like housing and job hiring.

• The federal government will adopt best practices for using AI, ensuring data protection and fairness in public services.

That’s not so bad.

Here’s some more:

• Developers must test powerful AI systems and share results with the government before releasing them to the public.

• The government will create tools to detect disinformation made by AI.

• The U.S. will form a National AI Research Resource and fund projects using AI to tackle big challenges like healthcare and climate change.

Great, I feel safer already.

MEMES FOR DESSERT

YOUR DAILY MUNCH

Think Pieces

What is this ChatGPT thing, anyway? A look at how to use AI, how it’s stealing your content, and how to steal it back.

A TED talk about how to think about AI. Bonus points if you’re a Hitchhiker’s Guide to the Galaxy fan.

How AI affects content moderation. Spoiler alert: it makes it easier and harder, simultaneously.

Startup News

OpenAI leases two buildings from Uber’s Mission Bay headquarters. It’s San Francisco’s largest office lease since 2018 and houses 2,000+ employees.

Abridge, a clinical documentation AI startup, secured $30 million. Investors include CVS, Kaiser Permanente, Mayo Clinic, and more.

Staff from Google, Microsoft, OpenAI, and more join the UN’s AI advisory board. It’s called the AI Advisory Body on Artificial Intelligence

Research

EdgeLlama — a peer-to-peer network of LLMs that are fine-tuned by a community of users (Available next week).

MimicGen — an AI system that generates entire datasets by observing and mimicking human demonstrations.

FP8-LM — an automatic mixed-precision framework for training LLMs

Tools

MagicPost — an AI content creator specifically for LinkedIn posts.

Kimchi Reader — an AI tool for reading and learning Korean.

flowRL — automatically updates your site’s UI based on the user’s behavior.

Chatmind — create and edit maps with text prompts.

RECOMMENDED READING

If you like Bot Eat Brain there’s a good chance you’ll like this newsletter too:

👨 The Average Joe — Market insights, trends, and analysis to help you become a better investor. We like their easy-to-read articles that cut right to the meaty bits.

TWEET OF THE DAY

We could explain this, but Ethan Mollick, the Wharton University Professor, prompted ChatGPT to explain itself.

Tag us on Twitter @BotEatBrain for a chance to be featured here tomorrow.

AI ART-SHOW

Until next time 🤖😋🧠