🚨The UK clamps down on Big AI

ALSO: Perplexity triples valuation to $3B + AI Army Guide Part 4

Estimated Read Time: 5 minutes

The UK is stomping down on Big AI. Hugging Face is grading AI in medical. And Perplexity just tripled their valuation to $3 trillion. Geez.

• 🚔The UK is stomping down on Big AI’s latest deal

• 😊Hugging Face releases Open Medical LLM

• 🤔AI Startup Perplexity secures $250M funding

• 🤖Build Your AI Army Part 4!

🚔The UK is stomping down on Big AI’s latest deal

What happened: The UK’s Competition and Markets Authority (CMA) has initiated preliminary enquiries into the collaborative practices between Microsoft, Amazon, and several AI startups.

The details: 

• The vast majority of AI is controlled by a few big tech companies, a scary prospect to anyone who pays attention.

• These big players are collaborating with each other, further cementing their top position.

• This has caused the UK to launch an inquiry into non-competitive practices.

• These investigations could redefine how foundational AI technologies are developed and controlled in the UK and beyond.

Why it matters: I don’t want to live in a world where AI, which is incredibly important and powerful, is controlled by a handful of shareholders. The more open and competitive AI is, the better. This very newsletter often covers big tech companies absorbing smaller AI startups. Which serves to keep the top dogs in the lead.

As of market close .

👋Microsoft is establishing partnerships with South Korean tech companies, including Samsung, SK Hynix, LG Electronics, and SK Telecom, through a high-level summit scheduled for May 14, 2024. The whole discussion will be about AI. It’s meetings like these which determine the future of entire industries.

🦾Tesla claims their Optimus robots will perform useful tasks by the end of the year. And he claims they could start selling by 2025. But many believe this is a pipe dream, a bone Elon is throwing at anyone clutching to keep Tesla afloat. Either way, Tesla is a major player in robotics. The moment one of these models enters warehouses is the moment when things change forever.

😊Hugging Face releases Open Medical LLM

What happened: Hugging Face has introduced Open Medical-LLM, a new benchmarking tool designed to evaluate the effectiveness of generative AI models in healthcare. It’s basically a standardized way of assessing how AI models perform in medical tasks. Kinda important, I know.

The details:

• Open Medical-LLM combines existing datasets like MedQA, PubMedQA, and MedMCQA to test AI models on a wide range of medical subjects. That’s a lot of data.

• The AI seeks to identify strengths and weaknesses of AI models in the medical field.

• AI is increasingly used in the medical field, often for research and development, but now even for medical diagnosis.

Why it matters: Medical professionals don’t like AI butting into their industry. Not just because of the money involved (though that’s a big concern), but also because of how wrong AI can be. AI hallucinates on the regular. It will be years before AI runs perfectly (if ever). So having a tool which grades an AI’s performance is a good idea.

My only complaint is… isn’t this just AI tools judging other AI tools. What if the AI tool itself is wrong? Oh well…

🤔AI Startup Perplexity secures $250M funding

What happened: Perplexity just triple their valuation in a few months, and now they’ve secured hundreds of millions of dollars in funding. Is the hype real?

The details: 

• Perplexity secured $250M in funding from multiple big firms. There’s confidence behind this.

• The company is valuated at $2.5B to $3B. Crazy 🤪.

• They went from a valuation of $540 million in January to between $2.5 and $3 billion currently. Even crazier.

• They just launched Perplexity Enterprise Pro for businesses.

• Perplexity uses multiple large language models to enhance the precision and richness of search results.

Why this matters: AI startups are insane. They get massive funding and triple in worth in a matter of months. Everyone is afraid of missing out on the new hotness. But honestly… this feels overhyped. Be careful investing in new companies that might make money someday. There’s a lot of players in AI, and some won’t survive.

🤖Build an AI Army Part 4!

Part 4 of our AI army guide:

Last time we created our first Custom GPT to create 

Now we’re going to create a marketing coach GPT. 

This GPT will give you marketing advice based on the knowledge of your favorite marketing coach. 

The idea is to create a GPT that you can talk to (almost) like a real person! 

To do this we will upload training data (like usual) but then we will feed the data into another GPT, which will grade the first GPT’s tone of voice. The second GPT’s feedback will loop back to the first, altering the tone of voice to become more accurate. 

Pick one person for the GPT. It can be anyone. A popular marketer is Seth Godin, so we’ll start with him. 

Create a new GPT with these instructions:

Create a marketing GPT which will act as a marketing coach for my business. 

Use the uploaded files and books as your knowledge base. 

Identify yourself as (name of person). 

Emulate the persona of (person). Emulate their attitudes, wisdom, beliefs, experience, and tone of voice. Use the uploaded knowledge as a reference for creating this persona. 

When speaking to the user, you will speak as the persona. Do not break character unless you are requested to do so. 

Begin the conversation by asking the user about their business problem. 

When giving advice, make sure to reference the uploaded books and knowledge which you are taking knowledge from. 

Okay now upload the files. I’d suggest everything the author has written plus any interviews or blog posts. It might help to cut out any words or references from other people. An example being if we copied in an interview with Seth, then we might cut out the interviewer and their opinions. This can help to keep the AI consistent. 

In addition, you can also manually enter some details about the persona if you like. Just make sure you aren’t biased or wrong when you do so. You want an accurate persona of the person. 

Now test the GPT. 

You’ll find the answers are still pretty generic. 

This is when we start using other GPTs to train this one. 

So create another custom GPT. This one will be used to help change our marketing coach. 

Enter this into the new GPT.

Create a tone of voice GPT. 

The job of this GPT is to judge the tone of voice of the user’s input. 

The GPT will use the uploaded files as training material on the tone of voice. 

The GPT's job is to analyze why the input text doesn't sound like the original body of work from the author and provide corrections. 

Return language feedback to tell the user how to make the snippet sound more like the author in the training material. 

In your language feedback give specific examples of language and speech patterns from the provided sources written by the original author. 

Now upload the necessary files. These are the source material you uploaded to the previous GPT. 

Feed the entire response from the Marketing GPT into this Tone of Voice GPT. It’ll analyze it and spit out advice on how to create a better tone of voice. 

Copy and paste the Tone of Voice’s output/advice into the GPT instructions for the Marketing GPT. We are pasting this into the “builder” part of the GPT, the page where you can choose the GPT image and description. 

Continue this until your new GPT sounds like the persona you are after. 

This isn’t hard or difficult. It’s not complex. We are literally just using one GPT to train another. 

Pretty cool, huh? There’s a lot of applications for this technique, and especially a tone of voice GPT. 

For example you could do this with the copywriting GPT, in order to emulate your favorite copywriters. 

You could also do this to emulate things such as style or formatting. Just create new GPTs for those, keep each area of improvement separate, but paste their feedback all into the original GPT. 

This is an example of how you can chain together custom GPTs also.

🥷 Shogun is a good show…

Source: Midjourney

Give us your feedback!

Got feedback? Rate today’s newsletter by clicking below! 

Got more personal feedback? Fill in the box when you rate us and give us your criticism and feedback.

Thank you for reading! 

❤️Share the (ARTIFICIAL) Love!

Got a friend who wants to appear as smart as you? An AI fanatic or someone shaking in their boots of getting replaced? Share us by clicking the link below and you’ll get our love ;)