This is NOT Will Smith. Interesting!

Deep Fake meets Hollywood

Hello!

Here is a fresh edition of the Practical AI newsletter, the newsletter that shows you what your future looks like today.

Here are the headlines:

  • From South Park to Hollywood Deep Fake

  • Let’s get practical with ChatGPT and GPT-3

    • Topic research and article outlines

    • Help programmers debug code

    • Product research on Amazon

    • Ryan Reynolds

  • The craziest AI experiment of 2022/2023

  • AI art of the week

  • Quick links - interesting AI news

Let’s jump in!

From South Park to Hollywood Deep Fake 

Have you ever watched South Park?

In case you haven’t, this is how ChatGPT describes it in one sentence (it was better than mine, so I used it):

South Park is an adult animated sitcom that follows the absurd adventures of four foul-mouthed grade schoolers in the fictional town of South Park, Colorado, satirizing a wide range of topics and societal issues.

It was first aired in 1997 and was recently renewed until 2027.

But what does South Park have to do with AI or Deep Fake, you ask?

Well, technically, nothing.

However, Trey Parker and Matt Stone, the creators of South Park, are the new Deep Fake AI favorites of Hollywood.

In 2019 and 2020, the duo worked on creating a feature film utilizing various synthetic media tools, like face-swapping.

They were scheduled to start filming Monday 16th of March 2020.

As many Americans will remember, on the 15th of March, the country shut down due to a certain virus, and the film project was suspended indefinitely.

They decided to spin out the synthetic media technology, and Deep Voodoo was born.

During the lockdowns, the viral success “Sassy Justice” was born.

They utilized their deep fake technology to create a silly video series making parodies of public figures like Donald Trump and Mark Zuckerberg.

5+ million views later suggest people were impressed.

Here is the full 15-minute video:

If you want to see the individual clips (basically the 15 min cut into smaller videos), check out their youtube channel: https://www.youtube.com/@SassyJustice 

The breakthrough success came in the middle of 2022 when their technology was used in a music video by the rapper Kendrick Lamar.

Cue the title of this week's newsletter: “This is not Will Smith.”

A screenshot from Kendrick Lamars music video

This video went viral thanks to the deep fake technology from Deep Voodoo.

In the video, we can see Kendrick Lamar's face (on the right of the image above) morphed into several other famous faces.

Here is the video:

At the end of the video, the song gets a bit out of sync, but that is a video editor issue, not an AI issue (basically, a poor job by the video editor for not syncing the soundtrack and the video track better).

In case you do not recognize all the six faces (I didn’t), these are the people Kendrick Lamar turns into:

  • OJ Simpson

  • Kanye West

  • Jussie Smollett

  • Will Smith

  • Kobe Bryant

  • Nipsey Hussle

  • NME (the magazine) claim to have identified the face of Kid Cudi too

Pretty amazing, right?

In the previous issue where I discussed Deep Fake, “This is NOT Morgan Freeman,” I spoke about the creative agency Revel AI that produces commercials and other videos for the entertainment industry using AI technology.

Deep Voodoo is giving them a run for the money, though eying the biggest prize of them all - Hollywood.

In December, the company secured a $20 million investment, its first outside capital.

The investment will accelerate Deep Voodoo’s developments in deep fake technology and offer more cost-effective visual effects services to the film industry.

Will 2023 be the year we see a major Hollywood production using AI and deep fake technology?

I think yes…

Actually, without knowing what technology is behind it, the new Indiana Jones movie scheduled for release in June 2023 uses some amazing visual effects to make Harrison Ford both older and younger.

I just don’t know if they used AI for the effects or more traditional CGI (I think the latter).

Either way, imagine what will happen when CGI and AI have a baby…

I think deep fake technology is going to be huge in Hollywood, and I think it is a great use case for AI and deep fake.

The implications of deep fake technology used by the wrong people with an agenda is the scary side of this technology.

My big question is, at what point will we (or technology), be unable to distinguish between the two?

And when that happens, then what…

Let’s get practical with ChatGPT and GPT-3

The “rest of the world” has finally discovered ChatGPT and the social media gurus are flooding Youtube, TikTok, Instagram, and Facebook with their videos.

ChatGPT is what wet dreams are made of for lazy people looking for shortcuts to making money.

Since I am no guru, you guys will just have to settle for some use cases that have real applications in business (not that some of the stuff the “gurus” are pushing aren’t useful - it is just the packaging and hype I hate!)

In order not to make this newsletter 100 pages long, I will only include three use cases in the newsletter.

I write more about AI, use cases, and AI tests over at my two blogs as well:

Use case 1 - topic research and article outlines

As ChatGPT is churning out text, obviously, I had to have content as the first use case.

But too many people use ChatGPT to generate and publish content “as is.”

This is not the way!

You should regard ChatGPT as a personal assistant that can aid you in the creation of content.

What does that mean, exactly?

Instead of asking ChatGPT to write you an article on “The Benefits of Daily Meditation,” ask it What are the topics an article on "the benefits of a daily meditation routine" should cover?

This way, you use the AI to aid you in understanding what an article on this topic should cover, and then you use that information to guide your article research and production.

The next question you may want to ask, is to create an outline for said article:

The next step is NOT using ChatGPT to write each of the paragraphs.

I mean, you could, but I would advise against it.

This tweet from Andrea Bosoni is spot on, in my humble opinion.

You want to create quality content that people want to read, which is helpful.

And despite Google officially saying they do not punish AI content, the keywords here are “helpful” and "people first"!

So, remember, ChatGPT is an assistant, not a writer!

NB! There are use cases for ChatGPT as a writer. Still, in this example, I am specifically talking about content for websites that are meant to rank or build the overall authority of a website in the eyes of Google.

Use case 2 - help programmers debug code

This use case I found on Twitter.

I get that most people reading this email will not understand much of the below screenshots.

I used to be a programmer, so let me try explain in simple terms.

When you create software products, etc., you “program” or write the computer code behind those solutions in a language the computer can understand.

When you run the programs (or the code), and there are errors in the code (like logical errors), the “programming program” will give you errors in the form of error codes.

Sometimes those error codes make no sense at all.

Debugging is the process of finding errors, and is an integral part of a programmers workday.

I hated debugging more than my ex-wife’s mother…

Having a tool like ChatGPT helping you (assistant, remember), is a game changer!

And this is what Joe has discovered.

Thus, there is a real use case for using ChatGPT to support you in debugging and error correcting for programmers.

Imagine the time savings you can achieve having an assistant like this.

And that is basically what Joe says - who needs coworkers when you have ChatGPT?

Here are the screenshots from ChatGPT (from the tweet)

This is AWESOME!

Ask any coder you know, they will agree.

NB! I don’t think ChatGPT will replace coders entirely. You still need to know to code and understand what you are doing and the scope of what it can do.

Use case 3 - use GPT-3 to do product research on Amazon

This is another use case I found on Twitter.

However, this is more technical than ChatGPT, and it uses GPT-3, the more versatile and available “brain” (v3) behind ChatGPT (you can buy access to GPT-3 on openai.com).

This is the tweet:

I got in touch with Jon and had a conversation with him about his test and asked if I could talk about his use case and use his screenshots.

He was cool with it all and asked if it would help me if he did a write-up of what he did.

Then he went silent.

Turns out he had doubts about publishing his findings.

For good reasons I think.

It is one heck of a use case of AI he has dug up.

If I was to speculate, he is probably working on something where he is using this himself - albeit probably in another niche than dog treats (Sorry, Jon).

Eventually, Jon decided to publish his article.

Here is a quick summary of what he did:

  • Get product reviews for dog treats from Amazon. He used a combination of programming and available tools to get these reviews from Amazon.

  • Once he had all the reviews in a file, he sorted them and deleted all the 3, 4, and 5-star reviews.

  • Then he created code to give GPT-3 the input together with the goals he wanted to accomplish and a request he wanted to be answered.

  • Here are the goal and the request that he fed the AI

    • Goal: Start a new dog treat business on Amazon

    • Request: Give me 5 things I should focus on in order to get ahead of the current competition.

  • The AI gave him 5 things he should focus on to ensure success (basically the opposite of things the 1 and 2 star reviews complained about with the competing treats).

Here are the five things a dog treat business should do, according to the AI:

  1. Ensure high-quality ingredients and avoid artificial additives.

  2. Ensure the treats are of an appropriate size and density for the type of dog they are intended for.

  3. Provide clear and accurate product descriptions and nutritional information.

  4. Enure the product is packaged correctly and securely to prevent spoilage and contamination.

  5. Offer a satisfaction guarantee and/or return policy.

Personally, this is one of the best use cases I have come across in the last six weeks since the AI craze started.

NB! I think it is possible to perform this experiment in a non-technical/manual way as well:

  • Find the products on Amazon and sort the reviews

  • Copy/paste the results in excel

  • Sort by stars, delete 3,4,5 stars

  • Copy and paste as many reviews into ChatGPT as possible and provide it your goals and request (similar to what Jon did). There is a limited input size with ChatGPT, so you may need to do this a few times.

  • Ask ChatGPT to make a summary of the results it gave you each time.

I have not tried it, but I think it would work.

Use case 4 - a bonus, because “why not”?

Ask ChatGPT for life advice, as you would your 90-year-old wise grandfather or grandmother:

I know, not exactly a use case for business or work, but still a few good ones for the road….

Use case 5 - a bonus bonus, because Ryan Reynolds…

A timely little tidbit from one of the greatest marketers in Hollywood…

(Ryan Reynolds is one of the owners of Mint Mobile.)

If you want to learn more use cases for ChatGPT, I am currently working on a range of articles discussing the various practical applications of ChatGPT, similar to what I wrote above.

The craziest AI experiment of 2022/2023

This AI experiment is so crazy I could not call it the AI experiment of the week.

Also, it was not performed by me (I wish).

I was going to include a sentence about this in the “quick links” section, but I realized that I have not read about this anywhere, and it baffled me!

Maybe it drowned in all the “noise” about ChatGPT, but this is crazy bananas to me!

ChatGPT passed the United States Medical Licensing Exam (USMLE) without any training in a recent research experiment!

Yes, you read that correctly.

It passed the medical exam.

Isn’t this crazy??

Sure, it was an experiment, but I believe this gives a glimpse into the future of what AI can do and what powers it holds.

So while everyone is writing about how ChatGPT can help students cheat on their exams, or how you can use it to write articles in 3 seconds or ask it to generate a business plan for your new biz, the media sits silently about this?

Personally, I am blown away!

This should be headline news on all the big news outlets.

But it is not (or maybe it was, but I did not see it since I do not even own a TV??)

AI art of the week - illustrate AI in the style of Caravaggio

This is a bit meta, but I basically asked the AI to illustrate itself in the style of Caravaggio.

Created with PlaygroundAI.com

Quick links - interesting AI news

Microsoft takes another big step towards integrating ChatGPT and the OpenAI tools into Microsoft products with the introduction of Microsoft Azure OpenAI Service. MAOS will be available to all developers, not only Microsoft, meaning they can access OpenAI’s APIs and get enterprise features and start building new and exciting AI tools.

Are we about to have a “Napster moment” with generative AI? Getty Images is taking generative AI company Stability AI to court in the UK, claiming they have “unlawfully copied and processed millions of images protected by copyright.”

CNET, one of the large publications that recently got a lot of attention as they were “caught” using AI content, is now revising its AI-written content after being notified of serious errors. Read this article and revert to my first use case of ChatGPT above 😀

This is not AI related, but at the end, I wanted to shout out another newsletter that I love and read every week.

If you are into newsletters as a busienss, run one yourself (I see you guys), newsletter growth, or curious about the creators behind the successful ones, you need to check out the Growth In Reverse newsletter. I genuinely love it (and I am not paid to say so!).

That’s all for this week.

I have a really interesting story I wanted to share today, but it is simply too big to add to an already long newsletter.

Therefore, I will leave you with a question until next time:

What is a bottleneck in AI that nobody really talks about?

If you have any questions about AI or any feedback, just hit reply or tweet me @thomassorheim 

Al’ll be back,

Thomas

PS! Why was the math book sad?

.

.

.

It had too many problems.

This is not the end. It is where the fun begins!

Reply

or to participate.