Justin's AI Newsletter

The most interesting content, news, and insights in AI each week.

--

Hey Friends. Happy Sunday. Welcome back to another newsletter. Let’s jump right in.

If you like this newsletter, tell your friends to subscribe here: justinfineberg.com. It’s the number one way to support the newsletter!

---

🤖 Don't fight AI, embrace it.

Recently, Italy made the decision to ban ChatGPT, and it seems that some schools and companies have also chosen to prohibit the technology. This approach is off the mark, and it could spell trouble for the futures of employees, individuals, and businesses alike.

Just last week, Goldman Sachs made a bold prediction: 300 million jobs could be lost to AI. But here's the twist – only 7% of US jobs are expected to be entirely replaced by AI. In contrast, a whopping 63% of jobs will actually be "complemented" by AI technology.

So, it's not really about AI taking away your job; it's more about losing your job to someone who can use AI to do it better than you can.

For businesses, it's time to think about how to incorporate AI into your everyday operations and long-term plans. If you've dabbled with ChatGPT, you know that this technology can save you some serious money. Invest time into trying new AI products, understanding what’s currently possible, and experimenting with using these products in different workflows.

For everyone else, keeping up with the latest developments is key. While it may be hard to stay on top of all the fast-moving AI news, do your best to try out as much as you can. Don't stress if you miss some stuff – the AI landscape will continue to evolve, and more opportunities will keep emerging.

Whether you're a business owner or just someone curious about the world of AI, you want to be on the right side of history. Make the most of being an early adopter. Embrace the technology, put it to use, and accelerate your growth as an individual or organization. History favors those who adapt and innovate.

---

🔑 The Key to Better ChatGPT Prompting

One concept that I think will become increasingly important for ChatGPT users is “Few-Shot Learning”.

Few-Shot Learning involves providing an AI model with a small number of examples to more accurately produce your ideal output.

This is an important concept in prompt engineering. Let’s go through a specific example.

Imagine you’re a skincare brand trying to write Instagram captions.

Without Few-Shot Prompting, you might prompt ChatGPT with:

Now, the result is fine. However, it lacks knowledge of our ideal writing style or our unique business specifications.

Here is the prompt with few-shot learning:

​

Few-shot learning involves providing examples of an input, and using those examples to teach ChatGPT what it should return.

For instance, in each example I provided, I started the caption with “Hey Skin Fam” and ended with a call to action and a link to “Example.com.

In other words, it learned the details and style of our ideal output.

This is just one simple example, but few-shot learning is incredibly powerful. If you find yourself doing repetitive tasks in ChatGPT, this is the best way to ensure that ChatGPT’s output meets your needs.

For a more in-depth video on Few-Shot Learning, check out my recent video.

💡 This Week's Recommendations

​Brainstorming ChatGPT Business Ideas With A Billionaire: Dharmesh Shah serves as the CTO of Hubspot and is highly active in the field of AI. It's fascinating to hear his perspective on what excites him about AI and where the real opportunities lie in the AI revolution.

​GPT-4 Is a Reasoning Engine: This piece explores the idea that GPT tech is more like a clever reasoning engine rather than just a knowledge base. Why does it matter? Well, it means we need to provide these models with ample information to reason effectively. The author also discusses "vector databases" and how they're a game-changer in the AI world. They help give these large language models access to unlimited information beyond what they've been trained on. I highly recommend this article because it'll open your mind to the many use cases of AI with vector databases. Trust me, there's a whole lot of untapped potential waiting to be discovered.

​Our approach to AI safety: If you're into the whole AI safety discussion, this one's for you. OpenAI shares their thoughts on AI safety, making the case that we can't get AI perfectly safe just by working in a lab. We can't predict every situation, so they say we need to test AI in the real world, step by step. Their plan? Keep releasing AI, but do it slowly and carefully to stay on the safe side.

​AI is entering an era of corporate control: Stanford released a 385 report on AI. The Verge breaks it down in this article. Some interesting things discussed in the article:

• AI can potentially help reduce emissions
• Chinese population are more hopeful about AI than Americans
• The number of incidents concerning the misuse of AI is rapidly rising.
• The demand for AI-related professional skills is increasing across virtually every American industrial sector.

---

Thanks for tuning in,

Justin

​

P.S. Any feedback on the newsletter or any other thoughts? Feel free to reply to this email – I read and appreciate every response!

​