⚡️ Guess How Many U.S. Adults Have Tried ChatGPT

June 2023 is here and the AI news juggernaut isn't slowing down…thanks for reading!

Here’s what we got for you today:

• 🤖 Just 14% of U.S. Adults Have Tried ChatGPT

• 🦅 Move Over LLaMA 7B We Now Have Falcon 40B

• 🌐 OpenAI’s Not So Secret Method to Fix ChatGPT Hallucinations

• 📸 A.I. Powered Camera With No Lenses

• 📝 Another Statement On A.I. Risk

• 🖼️  A.I. Generated Image of the Day

• ⚡️ A.I. Meme of the Day

Just 14% of U.S. Adults Have Tried ChatGPT

It seems the folks ranting on Twitter threads were spot on. Not many people are using ChatGPT, or they're not using it properly. This article throws a lot of numbers at us, but the bottom line is... we're still in the early stages of A.I. adoption.

According to a survey, while over half of U.S. adults are aware of the chatbot ChatGPT, only a small fraction have actually used it, and they found it somewhat useful. Knowledge of this tool varies, with those having more education, higher income, of Asian descent, male, and under 30 being more familiar. Among the users, it's mostly used for entertainment, learning, and to some extent, work. Despite some criticism about its accuracy, about 40% found it moderately useful, particularly the younger crowd.

Move Over LLaMA 7B We Now Have Falcon 40B

Falcon is a custom-built AI model trained using high-quality web data, with a focus on data quality at scale. Its unique architecture makes it more efficient and faster than GPT-3 while consuming only 75% of the training compute budget and a fifth of the compute at inference time. Matching the performance of top AI models from DeepMind, Google, and Anthropic, Falcon comprises 40 billion parameters and was trained on 1 trillion tokens using 384 GPUs on AWS over two months.

The pretraining dataset, created from public web crawls, contains nearly five trillion tokens, supplemented with curated sources like research papers and social media conversations. Falcon has shown promising results on open-source benchmarks and can generate creative text and solve complex problems.

OpenAI’s Not So Secret Method to Fix ChatGPT Hallucinations

Large language models like GPT-4 are getting better at solving complex problems that require many steps. But these models can still make mistakes, called "hallucinations," where they come up with things that aren't true. So, OpenAI researchers are trying to figure out how to train these models to make fewer mistakes.

They test two methods: one is called "outcome supervision," where they tell the model if the final answer is right or wrong. The other is called "process supervision," where they guide the model step-by-step. Using a bunch of math problems as a test, they found that the process supervision method works a lot better.

One important aspect of this is "alignment." If a model is aligned, that means it's thinking in a way that humans approve of. Process supervision is better for alignment because it makes the model follow a step-by-step process that we humans can understand.

A.I. Powered Camera With No Lenses

Neat! Paragraphica is an innovative device that combines location data and artificial intelligence to create a unique "photo" of a specific place and moment. It exists as both a physical prototype and a virtual tool. The camera's viewfinder shows a real-time description of your current location, and when you press the trigger, the device uses data about your location, weather, time of day, and nearby places to generate a descriptive paragraph of the scene.

Another Statement On A.I. Risk

Hmm, yet another document loaded with signatures. I suppose these statements do make an impact since they're being discussed, but it makes me question if there could be a more effective way to communicate these messages than just endorsing documents.

The Center for AI Safety, also known as CAIS and based in San Francisco, is a group that works to make sure artificial intelligence (AI) is safe to use. AI can do amazing things, but it also has some problems that need to be fixed to make sure it's safe. CAIS is all about solving these problems. They do research on safety, help other researchers learn more about AI safety, and push for rules that make AI safer.

Cool A.I. Tools, News and Demos

25+ cool apps added to Daily Zaps AI hub. If you have any cool tools to share, feel free to submit them or get in touch with us by replying to this email.

What Are ChatGPT Shared Links? - Shared links are a new feature that allow users to generate a unique URL for a ChatGPT conversations. Shared links offer a new way for users to share their ChatGPT conversations, replacing the most common method of sharing screenshots.

Bard Just Got More Visual - Bard will now provide images in responses to illustrate concepts, bring recommendations to life, and give you a richer experience.

Nvidia Hits $1 Trillion Market Value - Nvidia’s market capitalization surged past $1 trillion for the first time in the chipmaker’s history on Tuesday.

Building Systems with the ChatGPT API - Learn how to automate complex workflows using chain calls to a large language model.

 Search Is Now Available In The ChatGPT Plugin Store - It's somewhat unexpected that this wasn't incorporated at the time of the launch, but better late than never.

Who would've thought Time magazine would hop aboard the 'A.I. will end humanity' express so soon. All aboard the panic train, I guess.

100% A.I. Generated Image of the Day

Midjourney Prompt - The lion king as a McDonald’s happy meal toy,warm tones, hyper realistic,highly detailed,saturated,medium contrast, highly sharpened detail --ar 6:4 --style raw

A.I. Meme of the Day

Politics aside…is it good or bad that the White House is focused in A.I. this much so soon?

Refer Daily Zaps

Can you do me a solid and refer a buddy, colleague, or family member to Daily Zaps? As a thank you, I'll send you an Airtable list of over 170 AI-focused startups complete with links to their sites, founder names, funding amounts, investors, and all that good stuff.

If you don’t have anyone to refer you can still help by doing below.

• Reply to this email with a simple “yes” or “no”…LOL..so that I know people are reading this tiny newsletter on the Internet.

• Desktop readers, drag newsletter email to your primary inbox. This will make sure you don’t miss any issues.

• On your phone? No problem. Hit the 3 dots at top right corner, click "Move to" then "Primary."

• Apple mail users: Tap on our email address at the top of this email (next to "From:" on mobile) and click “Add to VIPs”

Peace out,

Daily Zaps Team