• Practical AI
  • Posts
  • Practical AI is back [and ChatGPT is under fire…]

Practical AI is back [and ChatGPT is under fire…]

Claude and Elon takes on OpenAI

Hello!

I am back!

I am sorry for going off the grid, going dark, going silent…

To the 100+ new subscribers who signed up for the Practical AI newsletter in the last two months of silence, welcome.

Please bear with me, as today’s newsletter will probably be slightly different from previous newsletters.

But don’t worry; there will still be Memes!

Lots of Memes!

So what happened, and why did I go silent for two months?

The short answer: I got scared!

The longer answer: When I started this newsletter and went down the AI rabbit hole, it was more like opening Pandora’s box than a rabbit hole...

I cannot unsee what I have seen.

I started out super positive about AI but turned quite negative and felt like doomsday was upon us.

Basically, all I could see was how AI was going to ruin society, ruin businesses, ruin the future of my children, and ruin life as we know it.

I think this negative perspective started to reflect in the newsletter.

And I don’t want to write a newsletter that is all about the negative sides of AI so often regurgitated in mass media.

I started this newsletter because I wanted to share the practical sides of AI and how businesses and business professionals could utilize AI to enhance and improve (but also be ready for the changes that inevitably will come).

As most of you know, this newsletter is not about “the 20 new AI tools released this week [no. 13 is shocking]”.

As much as there is a market for that, I want this newsletter to be the most actionable and valuable from a business perspective.

I spend a crazy amount of time researching this newsletter - easy 20 hours a week or more.

It is not sustainable, especially if, in the end, it “ruins my life” and freaks me the h*ll out.

So I stopped.

Cold turkey.

So what’s next?

I still believe AI is a disruptive technology that will change society.

But also, not all of it will be negative or evil.

I read the article “Why AI will save the world” by Marc Andreessen, co-founder of Netscape and VC firm Andreesseen-Horowitz.

In this 7000-word article, he argues that AI is good and will enhance human life.

And while I am not sure I agree 100% with everything he says, I agree with much of it.

So having read the positive perspectives on AI, I have decided to return to my original mission, which is to educate about AI and how it will affect business and life.

Thus, for my own mental health’s sake, and maybe yours too, I will focus the positive sides of AI (no more negative stuff), and having a much stronger educational profile - i.e., sharing practical and actionable advice.

In short, I will be writing in a positive or neutral tone, AND I will have a much stronger focus on “How to’s” and “This is what you can do with AI tool X.”

Sounds good?

I hope so!

With that out of the way, let’s get on with the regular programming.

But first, a quick question…

I am looking for questions people have about AI.

I want to try to answer some of them in this newsletter.

Therefore, do you have any questions about AI? 

I would love you to reply to this email with any AI questions.

Big or small, please don’t be shy!

They can be anything from “What can I do with ChatGPT’s new code interpreter,” to “How can I create multiple images with Midjourney using the same avatar/person,” to “In my job/business, we do X. Do you think there is a way we can use AI/ChatGPT/etc to solve this problem”?

The newsletter will NOT be a Q&A style newsletter; instead, I will write the newsletter with those questions in mind.

And I will not mention your name unless you want me to! (feel free to add URL and/or Twitter).

BUT HEY, no questions are stupid!

No question is too small!

If the question is “too small” for the newsletter, I might just answer your email directly! Free consults, anyone? 🙂

So, please email me any questions you have!

I warned you there would be memes…

Anthropic releases their “ChatGPT killer” Claude 2

Earlier this week, the AI company Anthropic released their new AI, Claude 2.

Claude 2 is only available for those in the UK or the USA, just like Bard…

But, as always, there is a workaround 😀 

So I “worked around” and gave Claude 2 a spin.

The media have already dubbed it as the “ChatGPT killer.”

While I am not sure I agree with that sentiment, it is very promising and a solid contender for the throne of publicly available LLMs.

Curious how it looks (or how to use it outside of the USA/UK)?

I made this (not so) short video of how it looks inside and how it works.

I intended to make it a few minutes long, but as the “blabber month” that I am, it turned out quite long...

Claude 2 in four sentences

It has 100,000 tokens available, allowing it to process much larger prompts, such as analyzing lengthy documents, technical guides, or even books (or generating longer outputs).

Claude 2 has been fine-tuned to deliver an improved user experience with explanations, less harmful or offensive output, and greater conversational abilities (touted by some as a “safer” AI).

Able to upload five documents (max 10MB each) to process and work on/with.

Great at programming, reasoning, and its training data is new, vs. ChatGPT’s data from 2021.

Obviously, there is more to an AI of this scale. But what strikes me is the ability to upload five files and 100,000 tokens; best of all, it’s free!

You need to pay $20 for ChatGPT Plus for the ability to use the code interpreter, which only allows one file upload.

And ChatGPT 4 gives you 8000 tokens to work with (despite its capability to handle 32K).

At this moment, you have access to an AI that is newer, “better” (more tokens), and can handle five files at the same time, FOR FREE.

That’s pretty amazing!

But hurry, I am sure Anthropic will introduce a paid version soon. It is not exactly cheap to run these gigantic LLMs.

So what can Claude2 do for me?

It is still early, and I am sure we will start to see use cases pop up over the next few weeks, just as we did with ChatGPT.

If you watched the video I included above, you would have heard about the millionaire that could have saved $100,000 by using Claude 2 to analyze their insurance contracts. I suggest you watch it, I am sure it is relevant for most of us…

So far, I have experimented with Claude 2 for copywriting tasks.

Having 100K tokens available is incredible, but I found ChatGPT 4 to be better at creating content.

I did a test where I put Claude 2 up against ChatGPT 4 and Bard.

ChatGPT came out on top, Claude 2 came in second, and Bard third.

I gave each AI the same prompts and compared the output.

Now, I am no genius writer, and English is not my first language. But from what I could tell, ChatGPT 4 is still the king of content creation.

But that is my limited experience telling me so.

I am sure Claude 2 is insanely good at content. It has to be…

Jasper.AI, the largest AI copywriter tool out there, switched from OpenAI/ChatGPT to Anthropic/Claude 2 as their AI.

I don’t imagine a billion-dollar company like Jasper would switch their AI brain “just like that.”

But I did not stop at just comparing the output…

No, I had more (nerdy) fun with Claude 2 and ChatGPT.

I discarded Bard entirely and did not include it in this next experiment.

But here is what I did.

I took the content created with ChatGPT 4 and Claude 2, fed it to ChatGPT 4, and asked it to create the optimal version based on them.

Interestingly enough, this new output was even better than the initial winner from ChatGPT 4.

Isn’t that interesting???

Unfortunately, I did not record this, so I cannot show you how I did it.

But if you want me to create a video and walk you through my process, I am more than happy to do so.

All you have to do is reply to this email and tell me you want to see it.

NB! 
The way I create content using AI is not only by “giving a prompt and press enter.” It is much more complex than that and takes time and multiple iterations. However, compared to creating content “the old way,” I save lots of hours each week (and I don’t need to hire expensive copywriters - sorry guys and gals…)

Enough about Claude for today!

I need to fit in a few lines about Elon Musk as well…

Elon Musk announces xAI

The week I chose to “come back” turned out to be one of those crazy weeks in AI.

Stuff always happens in AI, but this week was particularly crazy.

I will limit myself to Claude 2 and Elon Musk’s new AI company, xAI.

It is not surprising that Elon Musk would launch an AI company. It has been “in the cards” for a long while, and I have also mentioned it several times.

So here it is, the cat is out of the bag: xAI.

We don’t know much about the new startup, but its purpose is to “understand the true nature of the universe.” 

Pretty lofty stuff, in other words…

But we know that the AI will be trained by data from Twitter (and probably a bunch of other sources).

Elon Musk has put a lot of restrictions on access to Twitter for third parties in recent months. This is to prevent (AI) companies from scraping information from there.

Many of us understood what was coming.

He wants the data “for himself.”

I understand that - he paid $44 billion for it, after all.

At least his AI will have access to a source the others don’t….

(If the Twitter users will appreciate it is a whole other question)

Elon Musk has accused the major players in AI (Google, Microsoft, and OpenAI) of developing the technology without considering the risks to humans.

xAI wants to build a good AGI.

Musk fears that the others will not.

Is Musk a knight in shining armor, or is he offering us a Trojan horse?

If I dare offer a perspective…

Elon Musk already owns the largest AI company in the world!

Huh?

Say what now?

Tesla!

About 2 million Teslas are currently rolling on the streets around the globe.

They are all full of technology, including Tesla’s AIs (plural).

So here is a brain twister for you…

Is Tesla a car company or a data company?

Elon Musk has already said that xAI will work closely with his other companies, including Tesla.

Elon Musk already owns Tesla, Space-X, SolarCity/Tesla Energy, Neuralink, Twitter, The Boring Company, and now xAI.

Let me rephrase: Elon Musk already owns Electric cars with AI and data collection capabilities, Space Ships and satellites including internet via satellites/data transfer, Solar panels and energy creation and storage, brain chips for humans, a social media platform with 400 million users, an infrastructure company, and an AI with the stated aim to develop AGI.

It is beyond doubt that Elon Musk is a smart guy - a genius, no doubt.

Let’s hope he is one of the good guys!

I think he is!

(See what I did there, I am talking and thinking positively about AI 😀)

This is all I had for my comeback issue.

It turned into many memes, but hopefully, you found it interesting, or at a minimum, entertaining...

I want to remind you to send me questions - hit reply 👍

Until next week…

AI’ll be back (on Thursday).

Thomas

PS! 
This is where I always finish off with a dad joke created by AI. Today there is a slight twist… ChatGPT vs. Bard vs. Claude 2. Who wins? Hit reply and let me know…

ChatGPT

Claude 2

Bard AI

This newsletter took 18 hours to create 😯

Reply

or to participate.