Greetings,

Elon Musk’s generative AI startup, xAI, officially announced its long-awaited fundraising.

The new funding values the company at $24 billion. This is less than OpenAI's $86 billion valuation but more than the $18 billion valuation of the AI startup Anthropic after its recent funding.

… anyhow,

Welcome to another issue of Weekly.Ai, where we're on a mission to turn you into the AI Yoda of your office—wise, indispensable, and a bit mysterious!

What you’ll learn today? How to create the best Prompts for Chatbots.

Plus, enjoy a front-row seat to Google AI's latest bloopers - seems even robots have sick days!

• Google’s AI lies. And It lies a lot.

• GPT-4o features are now open to free users

• OpenAI confirms that they are currently training the next ChatGPT

• An updated guide for exceptional Prompting

• Cool Ai tools

• What’s cooking with BotBuilders

Let's dive in, shall we?

Google’s Ai search lies. It lies a Lot.

Even Liz Reid, Google’s head of search, admitted in a blog post that the company had made adjustments to its new AI search feature after screenshots of its errors went viral.

Reid’s post directly referenced two of the most viral, and wildly incorrect, AI Overview results.

One saw Google's algorithms endorse eating rocks because doing so “can be good for you,” and the other suggested using nontoxic glue to make everything stick to the pizza.

It’s probably best not to make any kind of AI-generated dinner menu without carefully reading it through first.

Here are some more wildly absurd answers:

• “1919 was 20 Years ago”

• “Yes, it’s possible to train eight days a week”

• “Yeah, Dogs Played in the NBA”

• “Some recommend setting your thermostat to 600-670 degrees Fahrenheit while sleeping”

Kinda ironic, that not long ago, Google started paying Reddit $60M/year to access its content archives for training its AI models.

Luckily, Google also made it possible to remove the Overviews completely. Here’s a quick tutorial.

What we could learn from this:

Google has one of the most powerful Ai’s, and even it still messes up. In short, make sure to always check the Sources.

You have to understand, Ai doesn’t speak our language. And while it’s making huge improvements, if you want an exceptional output, you need to make it easier for the Ai to understand you.

What a coincidence that we’ll go over exactly that later in this issue. 😉

GPT-4o features are now open to free users

After OpenAI already hinted at it in mid-May, we now all (yes, even free users) have access to the new model GPT-4o.

More specifically, we now have access to the:

• Browse feature

• Vision Feature

• Data analysis

• GPT Store (specialized versions of ChatGPT)

Yes! We as free users now have access to the Store. We can use GPT’s.

But before you cancel your subscription, here’s why it’s still worth it.

The free plan is super limited, and free users can’t Build GPTs. Plus the new voice feature (coming soon) will only be available for Plus users.

OpenAI Establishes Safety and Security Committee and Confirms Training Next-Gen ChatGPT

OpenAI has established a Safety and Security Committee led by directors Bret Taylor, Adam D’Angelo, Nicole Seligman, and CEO Sam Altman.

As OpenAI trains its next-generation model, it aims to advance capabilities towards AGI while maintaining industry-leading safety standards.

The committee's first task is to evaluate and improve OpenAI's processes and safeguards over the next 90 days, after which their recommendations will be reviewed by the full Board and publicly updated.

OpenAI will also consult with other experts, including cybersecurity advisors Rob Joyce and John Carlin.

A detailed guide on best practices for working with AI

These are 5 advanced Prompting tips you could use today, to skyrocket the effectiveness of the output. Every. Single. Time.

1. Context Learning - Provide the model with several examples of the task so it can better match your expectations.

2. Chain-of-Thought - Ask the model to explain its thinking before giving the final answer.

3. Retrieval-Augmented Generation - Give the model useful formats it can use in its answers to make them more accurate and reduce mistakes.

4. Structured Input and Output - Organize your inputs and expected results clearly and neatly.

5. Small, Focused Prompts > One complicated prompt - Break down complex tasks into smaller, easier steps, with each step focusing on one specific part of the task.

Most importantly, don’t forget to be detailed. It will f*ck up, wherever you let it to.

This was a *huge* article. But if you truly want to learn more about these techniques, and bring your Prompts to the next level, we recommend you give it a quick read.

Here are some tools for the week …

• Doly - Create amazing 3D product videos quickly and easily on your iPhone.

• Rewind - Personalized AI for automated notetaking. It memorizes everything on your laptop

• QA.tech - AI that tests your web apps, providing bug reports and actionable feedback

• Sider - Sidebar for assisting in reading & writing on any webpage (with GPT-4o, Claude 3, Gemini 1.5, etc.)

We’ll send our Newsletter Weekly.ai on Tuesday morning at 7 am MST from now on.

Sorry that we had to skip the last week …

That’s all for today.

If you’d like to give us feedback, just reply. We read our emails daily.

P.S. If you're still reading this, you're officially a genius. You know brilliant stuff when you see it ;-]

Weekly.Ai of 06-12-24

You have story ideas or feedback, email us at [email protected]