Americans quickly embraced AI

Read time: under 4 minutes

Welcome back, Superhuman

Happy Friday! It must be a very happy Friday indeed for employees at Databricks — who just shipped the world’s leading open-source AI model for just a tiny fraction of the cost it took to build GPT-4.

TODAY’S MENU

• How to prompt better with OpenAI’s official guide

• DBRX is the new sheriff in open-source town

• How quickly Americans embraced AI

• Chart: Open-source expectations

• 5 new AI tools to boost your productivity

• AI Generated Images: Neon Caves

NEWS

Today in AI & Tech

• Hard Time: Sam Bankman-Fried, the founder of the ill-fated cryptocurrency exchange FTX, was sentenced to 25 years in prison for defrauding investors.

• Hi-Tech Specs: Meta announced it’s adding AI features to its Ray-Ban smart glasses next month, including the ability to identify objects and translate text.

• Network & Chill: LinkedIn is reportedly trying to increase engagement with short-form videos and games in a bid to attract Twitter and TikTok users.

• Flavor Architects: Researchers in Belgium used AI to improve the taste of the country’s world-famous beers, finding that even small changes to a brew’s molecular structure could affect its aroma.

AI AT WORK

Learn how to prompt better with OpenAI’s official prompt engineering guide

There a ton of guides and courses out there on how to prompt ChatGPT the right way. But if you want to sharpen your prompting skills, who better than OpenAI to show you the way?

The ChatGPT maker released a prompt engineering guide with strategies to get the most out of your prompts. Some of the most valuable strategies from the guide include:

• Include details in your prompt

• Split complex tasks into simpler subtasks

• Specify the steps needed to complete the task

• Instruct ChatGPT to think through the answer slowly

• Provide examples and specify length of the desired answer

You can find the full list of prompt engineering, along with specific examples, in the OpenAI prompt engineering guide here.

NEXT IN AI

Databricks’ new model is another major step for open-source AI

Silicon Valley startup Databricks announced its open source AI model – DBRX – beats out competitors across 12 benchmarks, including reading comprehension and logic.

It scored higher than Meta’s Llama 2, Mistral’s Mixtral, and X’s Grok — three of the biggest open-source models currently on the market. DBRX still can’t beat new top dog Claude 3 or the tried-and-true GPT-4 from OpenAI. But those are not open source.

What makes DBRX unique? Sure, Meta allows others to build on top of its Llama model. But critics point out that it’s not totally an open book, keeping its training methods close to home. Databricks says it’s one-upping Meta in the transparency department: It wants to give users complete access to the building blocks it used to create DBRX.

The value of open-source: Say you’re a company that wants to use AI to learn more about its customers. You might not want to hand over sensitive customer data to major companies like OpenAI. Databricks says its model fixes that, giving you the ability to study proprietary information in private, without the fear that it could fall into the wrong hands.

Photo finish? Meta might be taking a pit stop, but it will soon be coming in hot. Reports suggest Llama-3 could be released as soon as July and is about as powerful as GPT-4.

AI & US

How quickly Americans embraced AI

A new Pew study shows that people with postgraduate degrees are the most likely of any education level to use AI at least a few times a day. Nearly a third of postgraduates in the US say they use AI several times a day or “almost constantly.”

Age also appears to play a role: 57% of 18-29-year-olds say they use AI at least a few times a week, compared with 36% of people 65 and over.

Overall, a surprising proportion of Americans have quickly embraced AI since ChatGPT was first introduced to the public in Nov 2022. Across all demographics, half of Americans say they use AI multiple times per week.

CHART

Open Source Expectations

In 2023, between 80-90% of the market share in AI went to closed-source models — those whose source code and training data are kept private. But according to Andreessen Horowitz, that’s rapidly changing.

Today, 41% of business leaders say they currently use open-source AI tools or plan to do so soon. Another 41% say they’ll make the switch as soon as open-source models catch up to the likes of GPT-4 and Claude, which are both closed-source.

Just 18% say they have no plans to increase their open-source usage.

PRODUCTIVITY

5 AI Tools to Supercharge Your Productivity

Gem: Helps recruiters find the best candidates, personalize communication at scale, and hire talent faster.

Retell: API that helps developers build human-like voice agents.

Heyday: An AI copilot that transforms your documents, notes, and conversations into quotes, shareable content, and a queryable database.

Symbl: A platform that helps you stream live calls, extract conversation insights(questions, topics, sentiment), and generate contextual responses.

Recast: Turn your read-later list into bite-sized audio summaries with AI.

^{* indicates a promoted tool, if any}

AI GENERATED IMAGES

Neon Caves

Prompt: a vibrant huge neon mountain with a cave that has vines hanging down the entrance of it

Acquire new customers and drive revenue by partnering with us

Superhuman is the world’s biggest AI newsletter for businesses and professionals with 600,000+ readers working at the world’s leading startups and enterprises. Companies like Amazon, Calendly, and Notion feature their products in Superhuman. Main ads are typically sold out 4 weeks in advance. You can book future ad spots here.  

🧞 Your wish is my command

Reviews of the day

Thanks for reading.

Until next time!

Zain

p.s. if you liked this newsletter, share it with your friends and colleagues here. For questions and feedback, email us at [email protected]. We always listen and respond